Want to dive inside the full archive? Through this special offer, you can start a 14-day free trial to unlock everything behind the paywall, including every framework and tool I’ve published since I launched Intelligently Human.

Start Your Free Trial

Note: Substack requires credit card details to start your trial, but you won’t be charged anything today. You can cancel instantly in your settings at any time and still keep full access for the entire 14 days.

If you read something useful in the archive, reply and share your thoughts. It shapes what I build next!

Warmly,
Kim

THE PATTERN

The Blind Spot in Marketing Dashboards

Across the marketing organizations I observe, the strongest AI performance numbers share a hidden risk most leadership teams haven’t yet named: ungoverned AI in the trust layer. On the surface, content velocity is up, review volume is climbing, and social proof looks stronger than it did eighteen months ago. Yet serious risk is accumulating within the workflows producing those stellar results.

The problem is that AI has quietly entered the human trust signals buyers mostly depend on to inform a purchase decision: reviews, testimonials, and ratings. In the organizations I observe most closely, leadership treats this critical layer as a high-volume production workflow, rather than a high-stakes reputation workflow.

What I’m seeing is a specific failure of organizational self-awareness. Teams can track their content output metrics to the decimal, but they cannot tell you which AI-assisted workflows operate closest to buyer trust. That dangerous gap is exactly what recent regulatory enforcement trends make visible in specific, documented cases where the workflow looked efficient right up until it didn’t.

THE BREAKDOWN

When Metrics Overwrite Reality

The trust layer in B2B marketing is built on a simple premise: reviews reflect real experience. When AI enters the workflow that produces those reviews, that premise becomes an assumption. And assumptions that go untested at scale become liabilities.

In January 2025, the FTC approved a final order against Sitejabber, an AI-enabled consumer review platform. The agency’s core finding was operationally familiar: Sitejabber hadn’t fabricated reviews from scratch, but had instead exploited the timing of automation. By triggering review prompts at the exact moment of purchase, before customers received or experienced the product, they captured premature ratings that inflated review counts and average scores. These distorted signals then filtered directly into Google search results, steering buyers toward vendors based on unverified trust.

For B2B marketing leaders, this maps directly to standard review generation programs on G2, Capterra, or peer review platforms. Any automated review prompt that triggers before a customer experiences meaningful product value introduces the same systematic risk. While the AI workflow successfully hits its performance objectives (more reviews, stronger ratings), it fails the fundamental governance test: Does this review validate a real user experience?

Your marketing dashboard rewards this setup by showing increased review volume and stronger average ratings. But a strict risk lens asks a completely different question: Has the customer experienced first-hand what they appear to validate?

Navigating Velocity and Deception

The underlying driver of this trend is simple: generative tools inverted the natural economy of content production. Creating authoritative collateral used to be the bottleneck; managing its distribution was the simple part. Today, production is instantaneous and cost-free, meaning the operational bottleneck shifts entirely to validation.

In December 2024, the FTC approved a consent order against Rytr, which sold an AI service that gave subscribers the means to generate testimonials and reviews. This was not generic content. It was detailed, specific, persuasive customer proof, with no connection to a real customer experience. The FTC’s concern was the workflow itself. AI was producing the social proof buyers depend on to make purchasing decisions.

In December 2025, the FTC reopened and set aside the Rytr order, concluding the original complaint did not satisfy the legal requirements of the FTC Act and that the order unduly burdened AI innovation.

This story remains useful for a different reason. The workflow Rytr offered still exists across dozens of tools now embedded in marketing operations: AI that can generate first-person customer language, detailed product experiences, and conversion-ready social proof at scale, with no verification that any of it reflects a real buyer or a real outcome. Whether or not a particular tool has crossed the FTC’s current threshold, the organizational question is the same: who decides which forms of proof AI should never generate? In many organizations I’ve worked with, the answer is usually paired with a blank stare: “no one has decided that yet.”

Mistaking Automation for Oversight

The exposure compounds when teams have relied on a structured technical workflow for genuine human oversight. The question no one asks aloud is whether running content through an automated checker is the same thing as reviewing it.

A stark example of this is the FTC’s final order against Workado (formerly Content at Scale AI). The company marketed an AI content detector designed to verify whether copy was generated by AI or by a human. However, the FTC found the company’s accuracy claims were unsupported because the model was trained primarily on academic text and failed to generalize to marketing copy. Teams that relied on it to certify content as “sufficiently human-reviewed” were substituting a broken tool for actual human judgment.

This is the pattern drawing the least internal scrutiny today. A team runs marketing copy through an automated checker. The checker passes it, but no human evaluates the claim’s quality, the source’s integrity, the brand risk, or whether the content should exist in the market at all. The tool’s output becomes the governance record.

The danger goes beyond inaccurate detection; it’s an organization that relied on “passed the tool” as a substitute for “reviewed.” That’s exactly the judgment that AI cannot perform on its own behalf.

THE CONTRAST

Building a Blueprint for Verifiable Trust

The Content Authenticity Initiative (CAI) and its Content Credentials standard take a different design premise. Adobe, as a founding CAI member, has built Content Credentials directly into its Creative Cloud suite. Rather than treating provenance as an afterthought, the standard embeds verifiable attribution data into creative files, allowing platforms and buyers to see exactly how content was created — whether it was AI-generated, human-edited, or built entirely from scratch.

Adobe’s design decision treats disclosure as a workflow function, not a remediation step. Governance no longer occurs after the content ships; it’s baked into what ships. Marketing teams using Content Credentials can produce AI-assisted content while maintaining a verifiable chain of attribution that external platforms and buyers can independently inspect.

That design principle provides the model that modern marketing organizations need. The core question it answers is simple yet foundational: Before this AI-assisted content enters the trust layer, does anyone explicitly own what it says and where it came from?

THE EQ INSIGHT

Asking the Right Questions Many Dashboards Miss

The core pattern connecting Sitejabber, Rytr, and Workado is not a failure of technology. In each case, the software performed as designed. The true breakdown occurred in leadership, specifically the absence of a conscious decision to ask whether the workflow should produce that output in the first place.

This inquiry, should this exist?, is a question about judgment boundaries. Answering it requires organizational self-awareness: the capacity to recognize that a workflow producing a strong KPI can simultaneously erode the brand trust that the KPI is meant to represent. In leadership, self-awareness is the baseline competency that precedes every other form of sound judgment. You cannot regulate what you refuse to see.

What an EQ lens surfaces that dashboards miss is the gap between productive metrics and accountability metrics. Review volume and content velocity are production metrics—they measure output. Determining whether a review reflects a real buyer’s experience or tracing an AI claim back to a verified source are accountability metrics—they measure truth. Recent regulatory enforcement actions are a public record of organizations that measured production while ignoring accountability.

The marketing leaders navigating this shift successfully are the ones who identified exactly which workflows operate closest to buyer trust and assigned a human owner to protect each one. Governing the trust layer requires a definitive mandate, not another technology configuration.

If this pattern sounds familiar in your organization, I work with marketing leaders and founders to build the judgment boundaries that govern AI before a consequence makes them urgent. I’m offering a limited number of EQ-i 2.0 coaching sessions at a discounted rate for Intelligently Human subscribers this quarter. Reply to this email if you’d like to learn more.

Sources

FTC v. Sitejabber — Final Order, January 2025. ftc.gov/news-events/news/press-releases/2025/01/ftc-approves-final-order-against-sitejabber

FTC v. Rytr — Final Order, December 2024; Order set aside December 2025. ftc.gov/news-events/news/press-releases/2024/12/ftc-approves-final-order-against-rytr | ftc.gov/news-events/news/press-releases/2025/12/ftc-reopens-sets-aside-rytr-final-order

FTC v. Workado / Content at Scale AI — Final Order, August 2025. ftc.gov/news-events/news/press-releases/2025/08/ftc-approves-final-order-against-workado-llc

FTC Consumer Reviews and Testimonials Rule — Final Rule, August 2024. ftc.gov/news-events/news/press-releases/2024/08/federal-trade-commission-announces-final-rule-banning-fake-reviews-testimonials

Content Authenticity Initiative - https://contentauthenticity.org/

Content Credentials (as adopted by Adobe) - adobe.com/products/creative-cloud/content-credentials.html

About Kim

Kim Celestre is a strategic advisor and executive coach who helps B2B marketing leaders navigate AI transformation without eroding judgment, trust, or human value. Her work is grounded in AIGP-certified responsible AI expertise, executive coaching, and 25 years of Silicon Valley marketing leadership, including 4 years as a Forrester industry analyst.

.

Marketing teams are heading into third-quarter business reviews, celebrating the strongest AI performance they have ever reported. Production is up, costs are down, and AI now touches almost everything they ship. For years, marketing leaders worked hard to prove that AI delivers results, and their dashboards now confirm this. But their celebration may be short-lived. In the background, a harder question looms as the conversation quickly shifts from AI results to AI risk.

THE PATTERN

A dashboard built for one question

Marketing built its AI dashboards to measure performance because the business relentlessly demands proof. The metrics that survive a quarterly review track output, cost, speed, and pipeline contribution. Each metric answers a single question: did the work get better, faster, or cheaper? Because these dashboards only answer this specific question, they can report a strong quarter while hiding the risks that quarter created.

The harsh reality is that AI raises marketing performance while generating new risks. AI-drafted claims, algorithmically personalized messages, and mass-produced competitive comparisons all carry brand or legal exposure. These risks never appear on the performance dashboard because no one created a metric to track them.

WHY IT MATTERS NOW

The board moved AI into the risk column

While marketing measured AI by performance, the board moved it into a different category. The Conference Board found that 72% of S&P 500 companies disclosed at least one material AI-related risk in their 2025 annual filings, up from 12% two years earlier. Reputational risk topped the list, cited by 38% of firms. Marketing didn't write that risk disclosure, but marketing's output is where much of that exposure originates.

Welcome to the new quarterly review. When AI was an experiment, the board wanted proof of adoption. Now that AI poses a disclosed risk, the board demands proof of governance. A dashboard that reports only performance answers the old question while ignoring the new one.

THE MISMATCH

Ambition outran governance

Marketing is chasing the AI leadership position, but it currently lacks the structure to hold it. Gartner’s 2026 CMO Spend Survey found that 70% of CMOs aim to become “AI leaders” in 2026, yet only 30% report mature AI readiness capabilities. This readiness measure goes beyond tools. It demands governance and the decision rules that dictate when AI operates independently and when humans must intervene.

This governance shortfall leaves claims unreviewed and exposure unowned. When the marketing team publishes AI-generated content without a human checking the underlying claim, brand, legal, or revenue risk quietly enters the workflow. It rarely surfaces in the same quarter the work shipped. Rather, the risk shows up months later in a legal review or a customer complaint, by which point nobody can trace it back to the source, and the person answering for it wasn’t the one who shipped it.

THE NEW TEST

The next test is risk-adjusted performance

For two years, marketing built credibility by proving AI could drive the numbers. That free ride is over. The board sees the exposure, public filings now broadcast the threat, and CFOs are stepping in to ask a harder question: “How much risk did you generate to hit these numbers, and what will it ultimately cost the organization?”

Marketing leaders who can answer both halves of that question will control the review; those who can’t will be at its mercy. Taking control requires a second layer of measurement, one that tracks exposure with the same discipline marketing already applies to AI output. That risk-adjusted framework is where this July series goes next.

Until then, confront one brutal question:

How much of this quarter’s performance relies on risks you aren’t even measuring?

Sources

The Conference Board and ESGAUGE, “AI Risk Disclosures in the S&P 500: Reputation, Cybersecurity, and Regulation,” October 2025. conference-board.org/press/AI-risks-disclosure-2025

Gartner, “2026 CMO Spend Survey” (press release), May 11, 2026. gartner.com newsroom

About Kim

Kim Celestre is a strategic advisor and executive coach who helps B2B marketing leaders navigate AI transformation without eroding judgment, trust, or human value. Her work is grounded in AIGP-certified responsible AI expertise, executive coaching, and 25 years of Silicon Valley marketing leadership, including 4 years as a Forrester industry analyst.

Each month, this series follows a fictional composite leader through a real professional challenge. The situations are composites drawn from patterns I observe across B2B marketing teams in AI transformation. The names and companies are invented. The failure modes are not.

THE AUDIT

What the passing deployment revealed

Three weeks after the incident, Nadia ran the AI Scope Map against both AI deployments. As expected, the customer retention deployment failed on the first criterion. The demand gen system passed all three gates, but blindsided Nadia with a revelation.

The demand gen deployment succeeded for reasons completely unrelated to the agents. From the start, Nadia built an informal governance architecture around it: scope, owners, and review points. But she failed to document those constraints.

When she approved the second customer retention deployment, she left those implicit AI agent guardrails behind. The agents didn’t fail, but she removed what made them work in the first use case.

WHAT SHIFTED

The informal structure, formalized

Nadia spent the following week formalizing those boundaries into two mandates:

Mandate 1 (Demand Gen): She formalized the boundaries she originally set for the demand gen system. Writing them down required no changes to the active system.

Mandate 2 (Customer Retention): She designed the mandate that the deployment always needed: qualifying accounts, escalation triggers, and communication authorizations.

When she brought both documents to the CEO, the discussion shifted into a governance audit. The CEO requested that she run the framework against every active deployment in the marketing stack. Not as a correction. As a baseline.

WHAT DIDN’T SHIFT

The limits of hindsight

The framework clarified the past but couldn’t fix it. When Nadia expanded the AI agent audit to four other active deployments, the governance exposures became clear.

Two deployments passed the audit cleanly. The other two carried governance exposure: one without a written mandate and one without a kill switch or incident response process. Both produced results that Nadia measured as successful in her AI agent dashboard for months.

WHAT LEADERS TAKE FROM THIS

Governance as an enabler, not a brake

The audit revealed what output metrics miss: a deployment without governance carries risk regardless of how well it performs today.

Governance cannot live implicitly in the minds of individual leaders. Formal mandates, explicit guardrails, and rapid response mechanisms must become core components of production readiness.

WHERE WE LEAVE NADIA

She wrote the AI mandate, but the audit is still open.

Nadia left the review with two clean mandates and a CEO who wanted the framework applied across the entire organization.

But one thing remains unresolved.

The incident report documented what the customer retention agents sent to active renewal accounts. It also revealed that the other two agent deployments ran without governance for months. Without a governance layer, they produced no audit trail. She is still unaware of what happened in those deployments before the governance existed.

The mandate is written. The audit is still open.

This series, in four parts:

Part 1 — The Leadership Brief: The mandate that was never written (published)

Part 2 — The Framework: The Agent Scope Map (published)

Part 3 — Real-World Examples: Where strategic constraint held, and where it didn’t (published)

Part 4 — The Debrief: What Nadia decided, and the question she’s still carrying (this post)

Sources

No external sources cited in this post. Supporting research referenced in Parts 1 and 2 of this series.

About Kim

Kim Celestre is a strategic advisor and executive coach who helps B2B marketing leaders navigate AI transformation without eroding judgment, trust, or human value. Her work is grounded in AIGP-certified responsible AI expertise, executive coaching, and 25 years of Silicon Valley marketing leadership, including 4 years as a Forrester industry analyst.

The Agent Scope Map from week 2 of the “Exercise Strategic Constraint” series makes a specific case: governance architecture should precede agent deployment, not follow the first incident that makes its absence obvious. The four examples that follow, JPMorgan Chase, Maersk, ServiceNow, and Atlassian, bear that out.

THE PATTERN

The successful deployments share a structure

JPMorgan Chase, Maersk, and ServiceNow look like an unlikely comparison: a global bank, a shipping company, and an enterprise software firm. Each operates under completely different business pressures and deployment models.

Yet their structural similarity is worth noting. Every successful deployment relied on a clear foundation that included:

• Named KPIs and written mandates

• Defined kill switches to halt out-of-scope agent actions

• Clear accountability assigned to specific individuals before agents went live

In every successful case, the governance framework preceded the deployment. This sequencing is what separates them from companies that scaled first and were forced to explain their reversals later.

Atlassian sits in a different category. The company made a structural bet on AI’s future performance rather than its demonstrated results. That distinction is where strategic constraint either holds or falls apart.

JPMORGAN CHASE

Governance at scale: 450 use cases, one operating principle

JPMorgan Chase is the playbook for enterprise AI done right. The bank runs more than 450 AI use cases in production, 200,000 employees utilize its proprietary “LLM Suite” daily, and the firm publicly tracks annual AI-driven business value at $1.5 to $2 billion, growing at a clip of 30 to 40 percent year-over-year.

The scale is impressive, but the discipline behind it is the more instructive story.

JPMorgan didn’t arrive at 450 use cases by approving every pilot project and seeing what survived. Instead, they treated every single deployment as a governed, independent bet. Before external AI tools were ever permitted near their ecosystems, leadership built internal capability first, a deliberate sequencing decision that prioritized governance architecture over deployment pressure.

Chief Analytics Officer Derek Waldron described it to McKinsey as managing the distance between what the technology can do and what the enterprise can actually capture.

JPMorgan closes that distance administratively. Every use case requires three things before it goes live:

• A named owner accountable for its output

• A measurable outcome tied directly to its output

• A defined mechanism to halt the agent if it deviates

At JPMorgan, governance is not a technology decision. It is an operating system decision made before any agent goes live.

MAERSK

Two use cases, three years of groundwork

Maersk’s AI story is less about what the company deployed and more about what it chose not to. While competitors rushed to announce company-wide AI transformation initiatives, Maersk spent three years, from 2021 through 2024, doing the unglamorous work: cleaning foundational data and building the infrastructure required to support commercial-scale AI.

When they finally deployed, they launched two domains: vessel route optimization and predictive maintenance across their global fleet. Both domains share three traits that make them well-suited for agents:

• Errors are recoverable. A suboptimal route or missed maintenance signal creates cost, not an irreversible relationship or brand damage.

• Success depends on data that agents can access. Weather patterns, fuel consumption rates, sensor readings, and voyage history are all machine-readable inputs; No relationship context is required.

• Outcomes are measurable against an objective standard. Fuel consumption percentages and vessel downtime are among the most trackable metrics in industrial operations.

The BCG AI Radar 2025 finding maps directly to Maersk’s approach: leading AI companies concentrate on an average of 3.5 use cases while laggards spread themselves thin across 6.1, cutting their ROI in half. Maersk ran a shorter list. The results followed.

SERVICENOW

The kill switch as a design feature, not a contingency plan

Most organizations treat the ability to stop an AI deployment as something to figure out after an agent causes an unexpected outcome. ServiceNow built it into the product before anything went wrong, then turned that capability into a market position.

In a 2026 Fortune interview, CEO Bill McDermott described the company’s approach as an AI Control Tower: the ability to pause, redirect, or stop any agent anywhere in the enterprise in a single action.

This framing matters:

• A kill switch positioned as a contingency is a panicked reaction.

• A kill switch as a design feature is an architectural decision made before the first agent goes live.

By productizing the kill switch, ServiceNow gave its enterprise clients the psychological safety required to scale. Enterprise AI scales when organizations can control it, and ServiceNow made control the product.

Three governance elements define readiness for enterprise AI at scale: a defined kill switch, a human review point before live accounts, and a documented incident response process. For ServiceNow, these are not abstract ideals. They are product features the company sells. That alignment between internal governance practice and external product strategy is what makes ServiceNow the clearest operational model in this set.

ATLASSIAN

When a forecast becomes the mandate

On March 11, 2026, Atlassian announced the elimination of 1,600 roles, representing 10 percent of its global workforce. More than 900 of those cuts came directly from software research and development, to “self-fund further investment in AI and enterprise sales.”

CEO Mike Cannon-Brookes directly acknowledged the implications: “It would be disingenuous to pretend AI doesn’t change the mix of skills we need or the number of roles required in certain areas.”

The critical context here is that Atlassian was not cutting from a position of financial distress. Heading into the announcement, the company posted 26 percent cloud revenue growth. These cuts were a strategic bet, not a bid for survival.

Atlassian’s move represents a fundamentally different approach to AI than the guardrails built by JPMorgan Chase or ServiceNow. Instead of waiting for AI agents to prove their operational value before restructuring, Atlassian restructured first; the decision came before the evidence.

This is the case most likely to be misread as disciplined. Cutting headcount in the name of AI efficiency can appear like a focused, constraint-driven decision. But strategic constraint requires evidence before action. Atlassian made those cuts before that evidence existed.

Investors read the announcement favorably. The stock rose roughly 2 percent on the news. Whether the underlying bet proves right remains to be seen. Atlassian’s “forecast first, prove later” decision structure stands as the clearest test case for a new leadership thesis: can organizations successfully front-run AI disruption by restructuring before the results arrive?

THE SYNTHESIS

The sequencing spectrum

A clear spectrum of strategic constraint emerges across these four cases:

• The Disciplined (JPMorgan & Maersk): prioritized risk mitigation. They built rigorous governance and refused to scale or expand until their internal data and infrastructure were ready.

• The Enablers (ServiceNow): prioritized control. They built the “AI Control Tower” first, giving themselves and their clients the psychological safety to deploy at scale.

• The Bettors (Atlassian): Accelerated past traditional guardrails, trading immediate organizational stability for a head start in an AI-first economy.

Avoiding a public walk-back requires governance discipline. Leaders must ensure their governance architecture matches their appetite for speed. The Agent Scope Map—the three-gate framework for AI agent deployment decisions —does just that.

This series, in four parts:

Part 1 — The Leadership Brief: The mandate that was never written (published)

Part 2 — The Framework: The Agent Scope Map (published)

Part 3 — Real-World Examples: Where strategic constraint held, and where it didn’t (this post)

Part 4 — The Debrief: What Nadia decided, and the question she’s still carrying

Sources

• JPMorgan Chase & Co. Investor Day Presentation. U.S. Securities and Exchange Commission, 2025. https://www.sec.gov/Archives/edgar/data/0000019617/000001961725000483/investordaypresentation.htm

• JPMorgan Chase AI Strategy: $18B Bet Paying Off. AI News, December 16, 2025. https://www.artificialintelligence-news.com/news/jpmorgan-chase-ai-strategy-2025/

• BCG. “As AI Investments Surge, CEOs Take the Lead.” BCG AI Radar. January 15, 2025. https://www.prnewswire.com/news-releases/as-ai-investments-surge-ceos-take-the-lead-on-decision-making-and-upskilling-themselves-302661849.html

• Maersk 2025 AI Strategy. Industry analyst reports, 2025–2026. [Secondary reference]

• ServiceNow Q1 2026 Earnings. Fortune, April 23, 2026. https://fortune.com/2026/04/23/servicenow-earnings-forecast-blistering-growth-ai-product-sales/

• ServiceNow at ATxSG 2026: Autonomous AI, Enterprise Workflows and the Future of Productivity. The Fast Mode, May 2026. https://www.thefastmode.com/expert-opinion/48645-servicenow-at-atxsg-2026-autonomous-ai-enterprise-workflows-and-the-future-of-productivity

• Atlassian to Cut Roughly 10% Jobs in Pivot to AI. Reuters, March 11, 2026. https://finance.yahoo.com/news/atlassian-lay-off-1-600-212610757.html

• Atlassian Follows Block’s Footsteps and Cuts Staff in the Name of AI. TechCrunch, March 12, 2026. https://techcrunch.com/2026/03/12/atlassian-follows-blocks-footsteps-and-cuts-staff-in-the-name-of-ai/

About Kim

Kim Celestre is a strategic advisor and executive coach who helps B2B marketing leaders navigate AI transformation without eroding judgment, trust, or human value. Her work is grounded in AIGP-certified responsible AI expertise, executive coaching, and 25 years of Silicon Valley marketing leadership, including 4 years as a Forrester industry analyst.

Each month, this series follows a fictional composite leader through a real professional challenge. The situations are composites drawn from patterns I observe across B2B marketing teams in AI transformation. The names and companies are invented. The failure modes are not.

THE FRAMEWORK

Three gates make the authorization mandatory

After the customer renewal incident, Nadia had a clear view of her agents’ behavior. The harder question—the one that should have preceded the deployment—was whether agents were the right call for that domain at all. The Agent Scope Map makes that question mandatory.

The tool runs three sequential gates. Each gate holds three binary criteria, and every criterion must pass before advancing. Binary means exactly that: pass or build. An organization that clears two of three criteria in a gate still carries a governance exposure, and the framework treats that as a mandate to build before deploying.

The gate sequence is deliberate. Domain suitability comes first because an agent in the wrong domain carries risk that strong mandate and governance in the later gates cannot fully address. Mandate clarity comes second. Governance readiness follows. Each gate builds on the work of the previous one.

View the Full Map

Read more

Each month, this series follows a fictional composite leader through a real professional challenge. The situations are composites drawn from patterns I observe across B2B marketing teams in AI transformation. I invented the names and companies. The failure modes, however, are real.

Nadia approved the AI expansion in February, during a quarterly business review where her demand gen results were the one bright spot in a flat pipeline quarter. Eight months earlier, she built an agentic AI system to handle prospecting sequences, content personalization, and pipeline qualification. The results held, and the team gained capacity. Her CEO asked where they could apply it to other use cases, and Nadia thought about customer renewal.

Her reasoning was defensible. The same personalization logic that interpreted buyer signals in demand gen, she reasoned, could do the same in retention. She moved forward, setting firm guardrails on volume and frequency. But she didn’t define what the agents should never be allowed to decide.

In week three of the new deployment, an agent generated a renewal sequence for three enterprise accounts in active contract renegotiations. The system generated the sequence on schedule. But the tone read like a script, completely misaligned with a delicate, multi-million dollar renewal. No one on Nadia’s team flagged those accounts as outside the agent’s mandate, because no such mandate existed. Two account executives received escalations from their contacts. One account went quiet.

Nadia was in a pipeline review when her VP of Sales texted her about the incident.

THE SITUATION

Eight months of results made the next decision feel inevitable

Nadia’s demand gen system didn’t fail once during the eight months it operated. That track record shaped the February decision more than the capability analysis. Consistent performance at one layer of the customer journey created a specific kind of confidence: that the agent understood the work as well as executed the task.

Nadia tightly governed the first deployment, with named KPIs, a defined scope, and human review at key points. The second deployment inherited the first one’s results, not its architecture.

The agents ran as designed. But the design was the problem.

WHY THIS MATTERS NOW

Agentic AI is scaling faster than the criteria to constrain it

Right now, B2B marketing leaders are operating under an intense executive directive to scale automation. According to BCG’s AI Radar 2026, roughly 90% of CEOs expect measurable ROI from AI agents this year, and companies dedicated nearly a third of their AI budgets to agentic deployments. CMOs aren’t expanding agent access in a vacuum; they are responding to massive top-down pressure. The conviction, however, is running ahead of the evidence.

McKinsey found that 62% of organizations are experimenting with AI agents, but only 39% report any impact on enterprise-level EBIT. A March 2026 BCG analysis found that 60% of companies have seen minimal or no business value from AI despite significant efforts, and nearly two-thirds report uncontrollable scaling expenses. Gartner projected in June 2025 that more than 40% of agentic AI projects will be canceled by the end of 2027, citing escalating costs, unclear business value, and inadequate risk controls. Gartner’s analyst named the root cause directly: most projects are “mostly driven by hype” and “often misapplied.” The mandate gap is where the failure lives.

The technology performed; the agent generated the renewal sequence exactly on schedule. The failure was operational. Because executive pressure demands fast deployment, leaders are scaling AI without building the governance architecture that defines where it should stop. The agents ran as designed. Nadia never built the constraint layer.

THE GAP

What she governed, and what she forgot to define

Her initial demand gen deployment succeeded because it had three structural pillars that the customer retention expansion did not: a tightly defined scope, a clear human owner for every automated action, and a mandatory review checkpoint before any text reached a live account.

Nadia didn’t actively strip these guardrails away for the second rollout. She assumed the governance structure had transferred.

But continuity is a dangerous assumption when changing customer motions.

Unlike cold prospecting, enterprise accounts live in a high-stakes relationship layer. Here, slight changes in timing, tone, and context carry massive revenue consequences. A conversion-optimized AI script simply isn’t equipped to read those nuances.

The three accounts currently in contract renegotiations went completely unflagged because the system lacked any concept of relationship states. Tracking a sensitive, ongoing negotiation requires human context and judgment, but Nadia had never designated anyone to feed that critical insight to the AI.

She wasn’t alone in this governance vacuum.

According to a 2026 industry survey by SmarterX, only 13% of organizations have all four foundational governance elements in place for AI deployment; nearly a third have none.

Nadia’s customer retention expansion fell into the latter camp. The incident report documented what the agents sent. Nadia never defined what should have been off-limits.

WHERE WE LEAVE NADIA

The incident report closed. The governing question stayed open.

Her VP of Sales spent three weeks in direct conversations to repair the damage and bring back the account that had gone quiet, conversations that a proper mandate would have prevented. They resolved the escalations, but the structural gap remained. Nadia acknowledged the mistake in a leadership meeting and committed to reviewing her AI expansion criteria.

The underlying problem was systemic: no one asked her what those criteria were before she approved the deployment. She hadn’t asked herself.

The question Nadia struggled to answer in that meeting is the one this series picks up: which decisions should an AI agent never make? She can document what the agents sent. The mandate that would have prevented it was never written.

This series, in four parts:

Part 1 — The Leadership Brief: The mandate that was never written (this post)

Part 2 — The Framework: The Agent Scope Map (paid)

Part 3 — Real-World Examples: Where strategic restraint held — and where it didn’t

Part 4 — The Debrief: What Nadia decided, and the question she’s still carrying

Sources

• BCG. “As AI Investments Surge, CEOs Take the Lead on Decision Making and Upskilling Themselves.” January 15, 2026. https://www.prnewswire.com/news-releases/as-ai-investments-surge-ceos-take-the-lead-on-decision-making-and-upskilling-themselves-302661849.html

• BCG. “How Leaders Build an AI-First Cost Advantage.” March 27, 2026. https://www.bcg.com/publications/2026/how-leaders-build-an-ai-first-cost-advantage

• McKinsey & Company. “The State of AI in 2025.” November 2025. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai

• Gartner. “Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027.” June 25, 2025. https://www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts-over-40-percent-of-agentic-ai-projects-will-be-canceled-by-end-of-2027

• SmarterX. “2026 State of AI for Business.” April 2026. [Industry survey — secondary reference]

About Kim

Kim Celestre is a strategic advisor and executive coach who helps marketing leaders navigate AI transformation without eroding judgment, trust, or human value. Her work is grounded in AIGP-certified responsible AI expertise, executive coaching, and 25 years of Silicon Valley marketing leadership, including 4 years as a Forrester industry analyst.

Each month, this series follows a fictional composite leader through a real professional challenge. The situations are composites drawn from patterns I observe across B2B marketing teams in AI transformation. The names and companies are invented, but the failure modes are not.

This is the final installment of our four-part series following Marisol, a composite B2B SaaS marketing executive, as she navigates the fallout of six months of operational decisions made without narration. While each post stands on its own, this week focuses on the reckoning: what she found when she finally asked, what she can’t undo, and the conversation she still owes.

WHAT CHANGED

The exercise she rewrote the morning of the meeting

Marisol did not run the session as planned. She originally prepared to walk her team through the Cultural Honesty Map herself, plot their collective position, and outline her forward-looking adjustments. Drawing inspiration from a Klarna case study she read the week prior, she intended to lead with a firm commitment: identify exactly which tasks she would never ask AI to take away from the team.

She abandoned that script the morning of the meeting. Instead, she handed out printed copies of the map and asked her team to complete three silent, solo tasks:

1. Locate yourself on the map.

2. Locate me on the map.

3. Write a single sentence describing the gap between those two points.

She gave them ten minutes. The room fell into an uncomfortable silence. When she asked her senior content strategist to share first, the strategist hesitated, then cut straight to the core: “You put yourself in Aligned. I put you in Exposed. That’s the gap.”

Marisol had walked into the session believing she was naming the change. Her team had walked in after six months of trying to decode the silent signals she didn’t realize she was sending. The mismatch wasn’t a minor communication hiccup; it was the chasm between a leader who believed her narrative was landing and a team left to write their own anxious version of reality.

The Cultural Honesty Map gave the meeting a forcing function that previous sessions lacked. Working through the axes together, Marisol explicitly defined what would end in the marketing function, what would evolve, and what would remain strictly human-led, no matter how advanced the tooling became. As the senior strategist documented these commitments in real time, a junior marketer asked the question the team had quietly carried for months:

“Which of these decisions did you already make without us?”

The true penalty of leadership silence isn’t what the leader mismanages; it’s what the team stops contributing while they wait for clarity.

WHAT SHE COULDN’T RECOVER

The six months her team spent waiting for a starting point

Marisol could fix the communication gap moving forward, but she couldn’t erase its history.

Two of her strongest contributors built private exit strategies. One took two exploratory coffee meetings with an outside recruiter; the other updated her resume the week before the all-hands. While both ultimately decided to stay, the baseline of trust they held in October—before the platform consolidation began—was severely depleted. They had spent half a year mentally detached, constructing a future where their work was no longer central. That defensive posture doesn’t unbuild itself just because a leader finally delivers the context she owed them months ago.

The cost manifested in smaller, compounding ways, too. A brilliant campaign concept that the content lead conceived in February surfaced in a brainstorm three weeks after the mapping exercise. When Marisol asked why it was held back for six months, the answer was telling: the lead didn’t think it was worth pitching in a department she assumed was being quietly dismantled. The idea was excellent, but it was six months late.

The true penalty of leadership silence isn’t what the leader mismanages; it’s what the team stops contributing while they wait for clarity. Marisol waited to perfect her strategy before sharing it, forgetting that her team wasn’t demanding perfection. They just needed a starting point.

WHAT REMAINS UNRESOLVED

The discipline she has not yet proven

By the end of May, the marketing team possessed a clearer sense of direction and a shared vocabulary for their evolving roles. The senior strategist who first identified the alignment led a session on maintaining cultural honesty with cross-functional partners. The retention risks stabilized. These were meaningful, rapid cultural corrections.

Yet a harder reality remained: Marisol had not yet faced the version of cultural honesty that actually tests a leader.

She had not had the conversation with her CMO. The CMO still believed the AI rollout was a textbook success. The baseline performance metrics looked healthy, and the original all-hands slide deck was still circulating in board updates as proof that marketing was adopting automation responsibly. That narrative wasn’t inherently false; it was simply incomplete. The CMO was operating from the exact same surface-level data Marisol had relied on back in March. The difference was that Marisol now knew the human cost behind those metrics.

Correcting the record upward meant vulnerability. It meant relitigating her own executive execution in front of the leader who approved her rollout strategy. It meant documenting that the all-hands presentation hadn’t landed well, despite what the congratulatory Slack threads suggested. It required owning a blind spot she only discovered because her team was brave enough to point it out.

Cultural honesty is not a conversation. It is a practice. The exercise Marisol ran with her team will be tested every time the pressure returns: the next board update, the next earnings cycle, the next time a campaign underperforms, and the easiest move is to soften what she tells the people above her and below her. Marisol had begun the practice. She had not yet proven she could sustain it.

That is where Discipline #4 reaches its limit. A diagnostic tool can map the gap between what a leader says and what a team hears, and running the exercise can bring that reality to light. But no framework can force a leader to manage upward with the same vulnerability she demands from her team.

June’s discipline begins where this one ends—with a leader facing a harder, more strategic question: Once you have named what is happening, what do you explicitly choose not to automate, even when you have the power to do so?

Thank you for following Marisol’s story through May. The EQ in Action series continues in June with Discipline #5: Exercise Strategic Restraint.

If this series has been useful, share it with a marketing leader navigating the same terrain.

About Kim

Kim Celestre is a strategic advisor and executive coach who helps B2B marketing leaders navigate AI transformation without eroding judgment, trust, or human value. Her work is grounded in AIGP-certified responsible AI expertise, executive coaching, and 25 years of Silicon Valley marketing leadership, including 4 years as a Forrester industry analyst.

Each month, this series follows a fictional leader through a real professional challenge. The situations are composites drawn from patterns I observe across B2B marketing teams in AI transformation. The names and companies are invented, but the failure modes are not. This month’s real-world examples — Google, Cloudflare, and Klarna — are drawn from documented public reporting.

Last week, Marisol ran the Cultural Honesty Map. The gap she found between what she said and what her team heard wasn’t hidden; it had been widening for six months while she focused on operational execution. She wasn’t alone. These organizations confronted the same question. Some answered it with deliberate design. One discovered the cost of leaving it unanswered.

THE PATTERN

What teams do when leaders are silent about AI

The failure mode in Marisol’s story was structural, not interpersonal. She didn’t withhold information from her team intentionally. She held it back because she believed her AI strategy wasn’t ready to share. During the six months she spent perfecting it, her team built their own version from what they could observe: a platform consolidation, a workflow integration, a hiring pattern that quietly shifted. By the time she presented the strategy at her all-hands, the version her team had written had already replaced the one she was about to deliver.

Every industry where AI is reshaping the work itself reveals this pattern. McKinsey’s January 2025 Superagency in the Workplace study surveyed 3,613 employees and 238 C-suite executives across six countries and found that employees use generative AI at work three times more than their leaders realize. C-suite executives estimated four percent of employees use AI for at least 30 percent of their daily work. Employees self-reported 13 percent. The same study found that 47 percent of employees believe AI will replace 30 percent of their work within a year, against only 20 percent of leaders who believe the same.

Those numbers describe a perception gap, but the gap itself is the diagnostic. When leaders are operating from one understanding of AI’s reach inside their organization, and employees are operating from another, the silence between the two is where the cultural honesty failure lives. Teams aren’t waiting for permission to use AI. They’re using it. What they’re waiting for is leadership to acknowledge what’s happening, what it means, and what comes next.

On the Cultural Honesty Map, this is the Inferred quadrant: leaders silent on the substance, teams already in motion. The team is constructing the future their leader has not yet named, from the operational signals available to them. The team writes the strategy. The leader is no longer the author of it.

THE BREAKDOWN

How silence reveals major decisions

Google signed a contract with the Pentagon to provide cloud and AI infrastructure for military operations. Employees learned about it through media reports. According to internal accounts published by Fortune in May 2026, the company never clearly communicated to employees that it was negotiating the contract or that it had signed one. Leadership’s closest response to the resulting concern was an internal memo about responsible AI and military partnerships that did not explicitly acknowledge the agreement.

A Google researcher quoted in the Fortune piece described the lack of transparency as “pretty indicting” and said the deal felt as if it had been done “in the dark.” The internal pushback that historically defined Google’s response to military contracts, most notably the 2018 employee revolt that killed Project Maven, was largely absent this time. The absence wasn’t agreement. It was the simpler fact that the company never gave employees the option to push back against the decision.

The breakdown wasn’t the contract itself. The breakdown was the company’s choice to let employees discover a consequential decision through external reporting, and to respond with adjacent language instead of direct acknowledgment. The cultural honesty failure was not that Google made a controversial decision. It was that leadership treated the decision as something communicated through absence, which is the most expensive form of silence a leader can choose.

When leaders communicate a consequential AI decision through implication rather than direct statement, it does not stay implicit. It becomes the dominant narrative. By the time leadership decides to address it directly, employees have already accepted a version of the decision that the leader did not write and cannot easily revise.

THE NAMED MODEL

Two companies, two cultural honesty moves

In May 2026, Cloudflare announced a workforce reduction of more than 1,100 employees. The announcement came as a blog post from co-founders Matthew Prince and Michelle Zatlyn, titled “Building for the Future,” published the same day every affected employee received a personal email from one of the two founders.

The post was specific in three ways that mattered. It named the cause: Cloudflare’s internal AI usage had increased by more than 600 percent in the prior three months, with employees across functions running thousands of AI agent sessions daily, and the company’s existing organizational structure no longer aligned with how employees actually did the work. It named the responsibility: Prince and Zatlyn wrote that the decision was theirs to own as founders, not something managers should communicate. It named the protection: severance packages that included full base pay through the end of the year, healthcare coverage through year-end for US employees, and equity vesting extended through August.

Clarity did not ease the transition. Cloudflare faced criticism in industry coverage for the timing of the announcement, which came alongside a strong earnings report. But the cultural honesty move held. Employees who were leaving knew why. Employees who were staying knew what had changed. The story was the company’s to tell, not the press’s.

Klarna made a different cultural honesty move, in a different format. In February 2026, CEO Sebastian Siemiatkowski spoke openly on the 20VC podcast about expecting Klarna’s workforce to shrink from 3,000 to under 2,000 employees by 2030, driven by AI absorbing white-collar work. He named the direction. Then he named what was not changing: the roles built on human connection.

“I have people in Portland talking to Nike. I have people in China talking to Shein. I have people in Amsterdam talking to Adyen,” Siemiatkowski said. “I’m still gonna argue that it’s going to be vital to offer a human connection there.” He opened the same conversation by saying “I want to be honest about the fact that I do think there’s going to be a very big shift.” Cultural honesty was not subtext. It was the move he named first.

Cloudflare and Klarna landed on different sides of the Cultural Honesty Map. Cloudflare made a moment legible through specificity and ownership. Klarna made a direction legible by naming what stayed human even as the headcount shrank. Both moves are versions of the same discipline: leaders making the transition visible while there is still time for the team to be part of it.

For marketing leaders, the equivalent moves are not theoretical. They are: naming the AI shift inside the marketing function before the workflow has fully changed; attaching a specific human owner to the consequences when AI-assisted work fails; and saying out loud what work will remain uniquely human regardless of how fast the tooling improves. These are not policy statements. They are structural decisions about who carries the weight of the transition.

WHERE WE LEAVE MARISOL

The conversation she scheduled for the following week

Marisol brought the Cultural Honesty Map to her team the following Tuesday. Rather than presenting her findings, she asked each person to locate themselves on the map first, in writing, before anyone spoke.

The room came back with answers Marisol had not expected. Most of her team did not place her where she had placed herself. The conversation that followed was the first honest one her team had about what the AI rollout had meant to them. Not the metrics. Not the workflow. The version of the future they had been carrying without language for it.

That is where the discipline of cultural honesty begins: not in the announcement, but in the conversation the silence has already shaped. That conversation is what Part 4 examines.

Here’s what’s coming in May:

Part 1: Marisol’s all-hands and the six months her team had been writing a different story. Published.

Part 2: The Cultural Honesty Map. A diagnostic for locating the gap between what you’ve said and what your team has heard. Published.

Part 3 (this post): What three organizations did with the same cultural honesty question, and what their choices revealed.

Part 4 (next week): Marisol revisits the all-hands with new information. What she’d say differently. What she cannot recover. The one question she cannot yet resolve.

About Kim

Kim Celestre is a strategic advisor and executive coach who helps B2B marketing leaders navigate AI transformation without eroding judgment, trust, or human value. Her work is grounded in AIGP-certified responsible AI expertise, executive coaching, and 25 years of Silicon Valley marketing leadership, including 4 years as a Forrester industry analyst.

Sources

• McKinsey & Company: Superagency in the Workplace: Empowering People to Unlock AI’s Full Potential, Mayer, Yee, Chui, & Roberts (January 28, 2025).

• Fortune: Google’s AI deal with the Pentagon has sparked employee backlash. But don’t expect a repeat of Project Maven, Beatrice Nolan (May 4, 2026).

• Cloudflare: Building for the Future, Matthew Prince and Michelle Zatlyn (May 7, 2026).

• Fortune: Klarna’s CEO agrees with Dario Amodei. He thinks his white-collar workforce will shrink by a third by 2030, (February 17, 2026).

The morning after

The day after the all-hands, Marisol opened her laptop and pulled up six months of meeting transcripts. She searched for the words she thought she had said. What she found was the language she used to talk about the AI rollout — adoption rates, platform consolidation, workflow integration, Q1 numbers — without ever saying, out loud, what any of it meant for the people doing the work.

Her team spent the same six months searching for the missing words. By the time Marisol stood in front of them with her AI-Forward vision deck, they weren’t interpreting what she said. They were interpreting what she still was not saying. The gap between those two things is what the Cultural Honesty Map is designed to measure.

The Cultural Honesty Map

Cultural honesty is the discipline of being transparent about how AI decisions are made, where accountability sits when AI is wrong, and what happens when systems fail. It is also the discipline of making the emotional transition visible; naming what is ending, what is evolving, and what remains uniquely human.

Most leaders frame cultural honesty as a communication strategy. They think the work is finding the right words, the right cadence, and the right venue. The Cultural Honesty Map starts from a different premise. The work is not what a leader says in any one meeting. The work is the cumulative meaning a team is building from everything a leader has said and everything a leader has not.

The map locates a leader and her team on two dimensions: what the leader said about the AI transition, and what the team heard. The gap between them is where most cultural honesty failures live.

Read more

The deck her team waited for

Marisol opened the all-hands with the slide she had labored over most. A single line, centered in the brand’s deep purple, read:

Where Marketing Goes Next: Our AI-Forward Operating Model.

She had built the deck over four weekends, working in the quiet hours while her organization, forty-three people at a Series D B2B SaaS company, waited for a signal. She had waited for the right moment: strong Q1 numbers and a CFO who was finally using the phrase compounding returns.

She presented for thirty-eight minutes. She thought she was being responsible. While she was solving for the business by waiting for perfect data, her team was solving for safety by assuming the worst. When she finally opened the floor, the questions hit the nerves:

• Does AI-Forward mean Human-Optional?

• Is efficiency a proxy for headcount reduction?

• Whose judgment is being replaced first?

Marisol leaned on her talking points. She said no decisions had been made. In her mind, she was being factually accurate. In the vacuum of her six-month silence, accuracy felt like an exit strategy.

By the next morning, a senior manager Marisol had been quietly developing for a director role requested a private fifteen minutes. His opening sentence reflected her intuition about how the presentation landed:

I want to understand what I should be telling my team, because right now they think the all-hands was the warning before the layoff.

The strategy she finally shared, and the version her team wrote without her.

Marisol hadn’t been hiding the strategy; she had simply been waiting until it was ready. The platform consolidation came first, in October. The workflow integrations came second, through January and February. The vision deck was the third, designed to explain why the first two had happened. She built it carefully because she believed her team deserved a fully formed version, not a half-formed one she would have to walk back later.

She did not account for the six months between move two and move three. In that silence, her team did what people do when their leader goes quiet during a major change. They built their own theory of the case from the only signal available: the operational patterns Marisol kept executing without explaining.

By the time she stood in front of them with the vision deck, her team had already set the theory. The platform consolidation was the prelude. The workflow automation was the rehearsal. The all-hands was the announcement they had been bracing for since the consolidation went live. Marisol’s deck did not introduce a strategy to the team; it confirmed the one they had already written: that the company was preparing to reduce the marketing org and that the AI-forward operating model was the framework that would justify it.

The cost of the honesty gap

The pattern Marisol walked into is documented. Edelman’s 2025 Trust Barometer found that employees who feel secure in their jobs because of AI are twice as likely to embrace its growing use, at fifty percent, than employees who feel their job security is decreasing, at twenty-one percent. By staying silent to avoid alarmist talk, Marisol triggered the resistance she was trying to prevent.

The silence has a second cost: leaders lose visibility into what their teams are actually doing. McKinsey’s January 2025 Superagency in the Workplace study surveyed 3,613 employees and 238 C-suite executives and found that employees use generative AI three times more than their leaders realize. C-suite leaders estimated that four percent of employees use AI for thirty percent or more of their daily work. Employees self-reported thirteen percent. The same study found employees were twice as likely as their leaders to believe AI would replace at least thirty percent of their work within the next year. Marisol’s team had been making decisions about their own future inside that gap. The all-hands did not close it.

Why a polished strategy cannot undo six months of inferred ones

The failure was not the all-hands; that was simply where the failure showed its face. The actual failure was the six months of operational decisions Marisol made without narration, each one defensible on its own, but indistinguishable from a layoff plan when viewed from the outside.

She consolidated four tools into one. She integrated workflows. She tracked adoption metrics weekly. What she did not do, in any of the venues available to her, was explain out loud what this meant for the work itself. She didn’t name what was ending. She didn’t name what was changing shape. Crucially, she didn’t name what she would never ask AI to do for them.

Two years in the role had given her something dangerous: operational confidence. She had quietly removed certain conversations from her calendar because she trusted her team to follow her. She stopped explaining her reasoning. She stopped checking in on the unspoken because she assumed nothing was unspoken between them. That trust became an assumption, and the assumption absorbed the questions her team needed her to answer in language they could repeat to each other in the Slack channels she did not see.

Morale had thinned over the prior six weeks, slowly enough that it did not show up in the engagement survey that closed in March. Output had not dropped, but the energy underneath the output had. Two strong contributors had stopped volunteering for cross-functional projects. One had quietly started interviewing. The senior manager had held the information back because he did not know how to bring it up without sounding alarmist. The all-hands gave him the opening to say it.

Where we leave Marisol

The polish of the deck made the gap worse, not better. A leader who waits six months to present a strategy, and then arrives with a fully built operating model, communicates something her team will read whether she intends it or not: that she has already made the decisions, set the pillars, and mapped the roles without them.

A draft strategy invites input. A finished strategy, presented this late in a transformation, reads like the announcement her team had already been bracing for.

Marisol ends her week sitting with a question her confidence kept off her calendar: How do you lead a team that has already moved on without you?

She doesn’t have a framework for seeing that yet.

She will by the end of the month.

Coming in May:

Part 2: The diagnostic tool Marisol needs before her next all-hands. Paid subscribers receive the full one-pager.

Part 3: Stories from marketing leaders who navigated the same silence differently, and what each choice cost or protected.

Part 4: Marisol revisits the all-hands. What she would say differently. What she cannot recover.

Sources:

• Edelman (2025). 2025 Trust Barometer Flash Poll: Trust and Artificial Intelligence at a Crossroads. Edelman Trust Institute. edelman.com

• Mayer, H., Yee, L., Chui, M., & Roberts, R. (2025). Superagency in the Workplace: Empowering people to unlock AI’s full potential. McKinsey & Company. January 28, 2025. mckinsey.com

Recently, Reuters and CNBC reported that Meta is capturing employee keystrokes and mouse clicks across hundreds of websites, including Google, LinkedIn, Slack, and GitHub. This “Model Capability Initiative” aims to gather training data for the agentic AI systems—software capable of autonomous reasoning—that Meta is racing to build. Simultaneously, Fortune reported that the company is preparing to cut up to 20% of its workforce, with layoffs potentially starting in May. Both announcements arrived in the same quarter.

Employees described the program as “dystopian.” Whether that description is fair matters less than the speed of the reaction. The Meta workforce didn’t wait for leadership to provide context; they connected the dots themselves and ran with the most alarming conclusion.

The interpretation

When a company announces a consequential decision without providing context, the workforce doesn’t pause; it interprets. These interpretations compound, circulate, and settle into a shared narrative that becomes nearly impossible to unwind. At Meta, two facts became public within the same news cycle: The company is collecting granular workflow data from its employees, and it is preparing to cut roughly one in five of them. The official framing addressed the data collection. The employees connected the dots.

The interpretation spreading inside and outside Meta is that this isn’t a productivity initiative or a generic training effort. It is an extraction of knowledge. Employees suspect the company is capturing how senior engineers debug and how researchers navigate tradeoffs because that is the content AI cannot yet reproduce. The implication is that the people whose judgment is being captured are the same people the company is preparing to let go. The company hasn’t confirmed or refuted the plan. The workforce is reading the silence.

This is a cultural honesty failure, not a communications failure. Cultural honesty exists when a company’s story and its actions match so closely that employees don’t have to choose what to believe. When a gap opens, the workforce fills in the blanks. Like a game of Mad Libs, the version they write is almost always worse than reality. Once employees write this “missing half” of the story, it becomes the actual narrative. No subsequent clarification from leadership can easily overwrite a version that has already spread through the ranks.

The warning for marketing leaders

Meta is the visible case because of its scale, but this pattern isn’t unique. Any company that deploys automated AI tools while cutting staff or squeezing margins creates a similar silence. I have watched smaller versions of this play out in executive conversations over the past several months. In every case, the company shared facts but failed to explain what those facts meant.

The cost of this silence is high. Senior contributors stop speaking in meetings, Slack channels go quiet, and honest feedback vanishes. People replace candor with safe, performative work. The very human judgment that AI was supposed to augment disappears instead.

The warning is loud. When a company layers AI into operational decisions, transparency is no longer a luxury. It is the only thing that stops employees from reaching conclusions the company will eventually have to live with. Companies that cannot name what they are doing, why they are doing it, and what it means for their people are borrowing trust they cannot pay back.

Sources

• Reuters / CNBC, April 22, 2026. “Meta tracks employee usage on Google, LinkedIn AI training project.” cnbc.com

• Fortune, April 21, 2026. “Meta will start tracking employees’ screens and keystrokes to train AI tools.” fortune.com

• Gartner press release, October 23, 2023. “Gartner Survey Finds 83% of HR Leaders Are Expected to Do More Now Compared to Three Years Ago.” gartner.com

• Harvard Business Review, February 20, 2024. “Surveilling Employees Erodes Trust, and Puts Managers in a Bind.” Thiel, McClean, Harvey, and Prince. hbr.org

Each month, this series follows a fictional composite leader through a real professional challenge. The situations are composites drawn from patterns I observe across B2B marketing teams in AI transformation. The names and companies are invented, but the failure modes are not.

This is the final post in a four-part series following Lena, a composite VP of Marketing at a publicly traded healthcare SaaS company, through the consequences of a restructuring decision that looked right on every metric available. Each post stands on its own. This week: what she changed, what she couldn’t recover, and the conversation she’s been avoiding.

WHAT CHANGED

The session that reframed the work

Lena didn’t open the workshop with the assessment tool. She opened it with a question: What makes this team irreplaceable?

The room was quiet for a moment. Then her content lead said, “buyer fluency”. Her campaign manager said, “the ability to read a healthcare audience’s skepticism”. A junior marketer said she wasn’t sure she’d developed enough expertise to answer the question. That last response stayed with Lena.

The conversation that followed was the first honest one the team had about what AI changed in their day-to-day work, not operationally, but in terms of what they were being asked to learn. Lena brought on the new junior hires for content production. But they didn’t build the domain understanding that made production meaningful.

Lena used the Human Strengths Protection Map to give the conversation structure. The team worked through the eight capabilities together, identifying where expertise existed, where it thinned, and where the restructuring impacted its development. By the end of the session, each person named one strength they wanted to build and one way Lena could support their development.

This visual is drawn from the Five Disciplines framework. Paid subscribers receive the full Human Strengths Protection Map — a one-page assessment tool for identifying which capabilities your team needs to protect and develop before AI reshapes the work.

The session delivered something more durable than a governance process: a team that understood what it was trying to protect, and a leader who made visible commitments about how she would help them do it.

Two weeks later, one of the junior marketers flagged a case study draft before it reached final review. The positioning implied a clinical outcome that the product couldn’t deliver. The marketer caught it because she knew the question mattered, not because anyone told her it did.

WHAT SHE COULDN’T RECOVER

The cost of rebuilding from scratch

In March, Lena brought in a fractional healthcare content specialist for a three-month engagement. The scope was deliberate: not production support, but knowledge transfer. The specialist sat in on buyer conversations, debriefed the junior team afterward, and documented the judgment calls the original specialist made but didn’t record.

It was the right structural decision. It was also slower and more expensive than retaining the specialist who was laid off during the restructure. Lena struggled to explain this to her CFO without sounding like she was relitigating a closed decision.

The three enterprise deals that stalled in Q1 were still in motion. Two moved deeper into the evaluation stage. Lena’s team remained in contention, but she had no way to measure whether the relationship was genuinely recoverable or whether the competitor being evaluated had already established the trust that would eventually close the deal. Her pipeline reports tracked stage and deal velocity, but there was no metric to track a buyer’s confidence in the team’s expertise.

That was the cost of restructuring for efficiency. Now she had to rebuild her team from the outside in.

WHAT REMAINS UNRESOLVED

The conversation she’s been avoiding

By the end of April, the team had a clearer sense of what they were trying to develop and a structural path for doing it. The fractional specialist was three weeks into the engagement. The junior marketer who flagged the case study started asking pointed questions in sprint planning. These were big changes, and they happened faster than Lena expected.

What hadn’t changed was what Lena said out loud about the decision that made all of this necessary.

She hadn’t told her team that the October restructuring was optimized for efficiency at the cost of lost domain expertise. She hadn’t shared that with her CFO, who approved the layoffs. She hadn’t shared it with her CMO, who assumed all was well based on the performance metrics.

She knew what the restructuring cost. The team knew it too, through the daily friction of doing work without the deep expertise to do it well. But neither side addressed this in the same room at the same time.

That is where Discipline #3 reaches its limit. The Human Strengths Protection Map gives a leader the language to see what’s at risk. Running the session gives a team the structure to name the strengths they need. But these actions don’t require a leader to stand in front of the people who absorbed the consequences of a bad decision and say what she would have done differently.

Lena knows what the restructuring cost her team. She has yet to say this out loud, to them or to the leadership that approved her decision.

May’s discipline starts with a different leader facing the same question: what does it cost to say the truth out loud?

Thank you for following Lena’s story through April. The EQ in Action series continues in May with Discipline #4: Lead with Cultural Honesty.

If this series has been useful, share it with a marketing leader navigating the same terrain.

Each month, this series follows a fictional leader through a real professional challenge. The situations are composites drawn from patterns I observe across B2B marketing teams in AI transformation. The names and companies are invented, but the failure modes are not. This month’s real-world examples — the University of Bath research, Deloitte Australia, and Morgan Stanley — are drawn from documented public reporting and peer-reviewed research.

Last week, Lena ran the Human Strengths Protection Map. The gaps she found weren’t hidden; they were just unnamed. She wasn’t alone in needing that clarity. These organizations confronted the same gaps, some by deliberate design and some after a consequence made it visible.

THE PATTERN

What breaks before anyone names it

The failure mode in Lena’s story was quiet by design. No single decision triggered it. No alarm went off when the senior specialist left with eight years of domain knowledge. The marketing engine looked healthy, even after the intuition that powered it was unplugged. The gap between what the team produced and what the team understood grew slowly, invisibly, until a buyer conversation and a stalled deal made it visible.

This pattern is not unique to Lena’s healthcare SaaS company. It’s common across every industry where AI efficiency models are applied to knowledge work. What made it difficult to catch was that the knowledge loss wasn’t visible in the metrics leaders tracked. It lived in the judgment calls that happened before content shipped, the ones no workflow captured and no dashboard measured.

Research published in the Human Resource Management Journal in February 2026 named this pattern precisely. A team at the University of Bath School of Management identified three forms of knowledge that AI is fundamentally incompatible with: embodied knowledge, developed through hands-on practice and real-world experience; encultured knowledge, the understanding of organizational culture and unwritten norms built through proximity and observation; and embrained knowledge, the analytical judgment and problem-solving capacity developed through years of applied expertise. “If people begin outsourcing thinking, decision-making, or interpretation to AI systems,” the researchers warned, “These critical forms of knowledge wither over time and create a dangerous dependency that could possibly compromise an organization or a company’s profitability.”

On the Human Strengths Protection Map, we identify the specific manifestation of this: Contextual Judgment. This is the hard-won wisdom required to decide when a situation is truly novel, and the existing data is dangerously incomplete.

This is what Lena lost — not output volume, but the contextual judgment her specialist carried, and the judgment that shaped every message before it went out.

THE BREAKDOWN

When claims oversight disappeared from the workflow

Australia’s Department of Employment and Workplace Relations commissioned Deloitte to conduct an independent audit of a government welfare compliance system, a contract valued at AU$440,000. When the 237-page report was published on the department’s website in July 2025, it did not disclose that Azure OpenAI GPT-4o was used to produce parts of it.

A University of Sydney health law researcher, Dr. Chris Rudge, reviewed the published report and flagged more than 20 errors. References pointed to academic papers that didn’t exist, and a quote attributed to a federal court judge had been fabricated. When Deloitte investigated, the firm confirmed that the footnotes and references were incorrect, issued a corrected version of the report, and refunded the final installment of the payment to the government. This consequence became a matter of public record.

The incident revealed that the technology failure was secondary to a more fundamental breakdown in claims oversight. The organization asserted claims in a document that would influence public policy, yet no human owner assumed accountability for their accuracy before the report was finalized. AI produced content that appeared superficially credible, but it was published without being vetted by anyone with the contextual judgment required to spot the fabrications.

For marketing leaders, this isn't just a cautionary tale about government audits; it’s a preview of the accountability gap. When AI-assisted volume outpaces human verification, the claims oversight muscle begins to atrophy. This capability—one of eight on the Human Strengths Protection Map—is the organizational habit of taking responsibility for every word published under the company’s name. In Deloitte’s case, a researcher caught the failure; in a marketing organization, the gap is usually discovered by a buyer or a competitor only after the damage is done.

THE AUGMENTATION MODEL

How Morgan Stanley kept humans as the judgment owners

Morgan Stanley’s deployment of AskResearchGPT offered a design model built around the opposite assumption: that the most valuable human strengths are worth protecting explicitly, not accidentally.

AskResearchGPT accelerated the retrieval and summarization of Morgan Stanley’s internal research library. Analysts surfaced relevant prior work faster, synthesized across larger bodies of research, and reduced the time spent searching for information that already existed inside the organization. The efficiency gain was real and documented.

This system reduced friction in the research process while humans retained the expertise required to apply that research with sound judgment. Human analysts remained accountable for the judgment calls: the interpretation, the client guidance, and the narrative that translated research into an actionable recommendation for a client.

The design decision Morgan Stanley made was the same one Lena’s team needed before the restructuring: a conscious determination about where AI would assist and where humans would own the outcome. It wasn’t a policy statement; it was a structural choice baked into the workflow long before deployment.

For marketing leaders, the equivalent scenario is a content operation where AI accelerates the production layer — first drafts, structural outlines, content variants — while humans retain explicit ownership of the judgment layer: positioning decisions, claims accountability, competitive framing, and the interpretation of what market intelligence means for a specific buyer in a specific moment. That separation requires the same deliberate design Morgan Stanley applied before the workflow was deployed, not negotiated after a mistake surfaced.

WHERE WE LEAVE LENA

The session Lena runs the following week

Lena brought the Human Strengths Protection Map to her team on a Thursday afternoon. Rather than presenting her results, she asked each team member to run through the assessment themselves.

The conversation that followed was the first honest one her team had about the real cost of the restructuring. Not the headcount. Not the output metrics. The human strength underneath both.

Two truths emerged in the room. For months, the team watched their contextual judgment erode in real-time, but they lacked two essential tools: a vocabulary to name the loss and the psychological safety to report it.

That is where the discipline of protecting human strengths begins: not in the assessment, but in the conversation the assessment makes possible.

That conversation is what Part 4 examines.

Here’s what’s coming in April:

Part 1: Lena’s situation and the human capability gap hiding inside her efficiency model. Published.

Part 2: The Human Strengths Protection Map — the assessment tool for identifying which marketing capabilities require active protection from AI displacement. Published.

Part 3 (this post): What the research reveals about AI and expertise erosion, and what the organizations that got the design right built instead.

Part 4 (next week): Lena revisits the restructuring decision with new information. What she’d change, what she can’t recover, and the one question still unresolved.

Sources

• University of Bath School of Management / Human Resource Management Journal: “On the Dangers of Large-Language Model Mediated Learning for Human Capital,” Professor Dirk Lindebaum et al. (February 2026)

• The Guardian / Fortune: Deloitte Australia government report — AI-generated errors, partial refund (October 2025)

• Morgan Stanley: AskResearchGPT press release

Most marketing leaders don’t deliberately eliminate human strengths. They optimize for efficiency, adopt AI tools, and make headcount decisions that look sound across all metrics. The human strengths erode quietly. The cost surfaces later: in a stalled deal, a positioning drift, a buyer conversation nobody on the team can hold.

This is the assessment Lena needed before she restructured her team. It works for any marketing leader navigating AI transformation, whether or not a restructuring is planned.

Each month, this series follows a fictional composite leader through a real professional challenge. The situations are composites drawn from patterns I observe across B2B marketing teams in AI transformation. The names and companies are invented. The failure modes are not.

THE FRAMEWORK

The map she needed before the audit

Lena’s channel audit measured the right things: reach, engagement, content volume, and channel distribution. It told her everything was working. What it couldn’t measure was the human capability underneath the workflow: the judgment, the fluency, the relational intelligence that made the content credible when a buyer pushed back.

That’s the gap the Human Strengths Protection Map is designed to close.

The map works as a pre-decision assessment. A marketing leader runs each of their team’s core capabilities through three columns: what the capability requires from a human, whether AI adoption has begun eroding it, and what a concrete protection action looks like. The output is not a score. It is a prioritized list of what to protect, reskill, or redesign before anything changes.

One thing this tool is not: a layoff planning guide. Most leaders who need it aren’t planning to cut anyone. They are adopting AI tools, accelerating workflows, and watching their teams produce more than ever. What is quietly eroding underneath stays invisible until a business consequence names it. The Human Strengths Protection Map is a reskilling and redesign tool. Use it before something changes, not after.

THE ASSESSMENT

Eight capabilities. Three questions each.

The map covers eight human strengths AI adoption puts at risk in any B2B marketing team. For each one, ask whether your team currently has the capability, whether AI has begun displacing it, and what you will do to protect it.

Here is a preview of the first three capabilities.

Buyer fluency

The empathy to understand how buyers think, feel, and decide — not only what they need. This is the capability Lena’s specialist carried and the junior hires hadn’t yet developed. It doesn’t live in a content brief or a persona document. It lives in the accumulated experience of sitting across from buyers and learning how they process risk, evaluate vendors, and decide who they trust.

Risk signal: Direct buyer conversations have decreased since AI tools entered the workflow.

Protection action: Conduct at least one unassisted buyer conversation per quarter per senior team member.

Competitive discernment

The discernment to position against competitors with confidence and without creating brand or legal exposure. AI can generate competitive comparisons faster than any human team. It cannot read the competitive landscape, assess partner sensitivities, or judge whether a positioning move will hold up when a buyer pushes back in a late-stage conversation.

Risk signal: Competitive positioning is generated by AI and enters workflows without a human review checkpoint.

Protection action: Require human sign-off on all competitive positioning before it moves downstream.

Claims oversight

Being accountable for what the organization asserts in the market. In a high-volume AI content operation, claims proliferate faster than anyone can track them. The human capability at risk here is the organizational habit of taking responsibility for what goes out under the company’s name.

Risk signal: AI-generated content publishes without a human verifying the underlying claim.

Protection action: Designate a named owner to review claims in every content type that touches product, legal, or compliance territory.

Read more

Each month, this series follows a fictional composite leader through a real professional challenge. The situations are composites drawn from patterns I observe across B2B marketing teams in AI transformation. The names and companies are invented. The failure modes are not.

The audit came back clean. The problem didn’t.

Lena inherited a marketing team built for a different era.

Nine months into her role as VP of Marketing at a mid-size, publicly traded healthcare SaaS company, she was still calibrating what she had taken on. The headcount model was designed before AI-assisted workflows existed. Two people were dedicated almost entirely to content volume, drafting, editing, and versioning messaging across a complex buyer landscape that included provider organizations, health systems, and value-based care networks.

The team delivered meticulous work. By every efficiency metric her CFO tracked, it was also expensive.

So Lena restructured.

She consolidated two roles into one, eliminated the senior healthcare content specialist position, and hired two junior marketers who could operate an AI-assisted workflow at pace. She’d run this math before, at a high-growth B2B tech company where AI-accelerated content delivered a genuine competitive advantage. She built a solid business case, and the CFO approved it in a week.

Output volume stabilized within six weeks. The dashboard looked fine for five months.

Then her sales leader put a single slide in front of her at the quarterly business review: three enterprise deals in the same provider vertical, all stalled at the solution overview stage. Buyers had gone quiet after the initial engagement. A smaller competitor consistently outperformed Lena’s team in late-stage conversations. Not on features. Not on pricing. On fluency.

Lena commissioned a channel audit. It came back clean.

THE SITUATION

The expertise that never made it into the workflow

The senior healthcare content specialist Lena eliminated spent eight years learning how provider organizations evaluate workflow change, where the language of clinical operations diverges from the language of enterprise software, and which claims create friction in a room full of compliance-aware buyers. None of that knowledge was documented. It lived in the judgment calls she made every time she shaped a message, softened a framing, or pushed back on positioning that would land wrong with a risk-averse healthcare buyer.

The junior marketers who replaced her were capable and fast. They learned the AI workflow quickly. No one remained to teach them the domain underneath the workflow.

That shortcoming doesn’t appear on a productivity dashboard — it appears in a sales call.

WHY THIS MATTERS NOW

The capability AI efficiency models quietly erase

The pattern Lena walked into is documented and accelerating. In October 2025, Gartner predicted that by 2030, half of enterprises will face irreversible skill shortages in critical roles due to GenAI skills erosion. Organizations are losing more than output quality. They are losing the conditions under which expertise gets built and transferred.

Gartner’s follow-on prediction sharpens that insight: through 2026, atrophy of critical-thinking skills due to GenAI use will push 50% of global organizations to require AI-free skills assessments. The concern is not that AI produces weak output. The concern is that sustained AI use, without deliberate protection of the human judgment underneath it, erodes the expertise that made the output credible in the first place.

Lena hadn’t reduced output — she restructured away the conditions that made it trustworthy to a healthcare buyer.

The AHA’s 2026 Health Care Workforce Scan identifies this dynamic explicitly in the healthcare context. As AI tools are embedded into healthcare workflows, organizations that fail to protect mentorship structures and redesigned education pathways remove the scaffolding where domain expertise becomes transferable. What holds true for clinical teams holds equally true for the marketing teams selling to them.

THE GAP

The competitive edge her dashboard couldn’t see

The channel audit told Lena nothing was wrong with her distribution. What it couldn’t measure was her team’s eroding domain fluency.

Forrester’s 2026 Buyer Insights research across 22 industries found that buyers in regulated sectors, including healthcare, rank expertise and trust above operational efficiency and price as purchase drivers — a pattern distinct from buyers in less regulated markets. Healthcare buyers don’t evaluate vendor content in isolation. They evaluate it against what they know about their own environment, and they notice when a vendor’s team doesn’t share that knowledge. That recognition rarely surfaces in a feedback form. It surfaces in a deal that quietly loses momentum.

Lena’s competitor ran a smaller content operation whose team could go deeper than the content when the room required it.

PwC’s 2026 health industries board survey found that nearly half of healthcare industry directors say management provides inadequate information on the risks associated with AI use inside their organizations. The board pressure pushing Lena toward efficiency was real. What the board wasn’t getting was the risk profile of an efficiency model that treated domain expertise as overhead. Lena hadn’t surfaced it.

WHERE WE LEAVE LENA

The question she still hasn’t asked herself

Lena has a clean audit, a stalled pipeline, and a structural question she hasn’t fully asked herself yet: not where the content is going, but what domain fluency her team lost when she restructured.

She restructured for efficiency, and she got it. What she optimized away was the institutional knowledge her buyers were looking for at exactly the moment the deal was in play.

She doesn’t have a framework for seeing that yet.

She will by the end of the month.

Here’s what’s coming in April:

Part 1 (this post): Lena’s situation and the human capability gap hiding inside her efficiency model.

Part 2: The tool Lena needs before she restructures again — the Human Strengths Protection Map, a framework for identifying which marketing capabilities require active protection from AI displacement. Paid subscribers receive the full downloadable one-pager.

Part 3: What other marketing leaders learned when they tried to protect human strengths — and what broke before they got it right.

Part 4: Lena revisits the restructuring decision with new information. What she’d change, what she can’t recover, and the one question still unresolved.

Sources

• Gartner: AI Revolution and Cost Pressures Drive Top Talent Acquisition Trends for 2026 (October 2025) — gartner.com/en/newsroom/press-releases/2025-10-07-gartner-says-ai-revolution-and-cost-pressures-are-two-forces-driving-the-top-four-trends-for-talent-acquisition-in-2026

• Gartner: Top Predictions for IT Organizations and Users in 2026 and Beyond (October 2025) — gartner.com/en/newsroom/press-releases/2025-10-21-gartner-unveils-top-predictions-for-it-organizations-and-users-in-2026-and-beyond

• Forrester: 2026 Buyer Insights: Industries (December 2025) — forrester.com/report/2026-buyer-insights-industries/RES186880

• AHA: 2026 Health Care Workforce Scan Executive Summary (December 2025) — aha.org/system/files/media/file/2025/12/2026-Health-Care-Workforce-Scan-Executive-Summary.pdf

• PwC: 2026 Corporate Governance Trends in Health Industries (February 2026) — pwc.com/us/en/services/governance-insights-center/library/annual-corporate-directors-survey/health-industries.html

Intelligently Human publishes every Tuesday, Wednesday, and Thursday. Subscribe to follow Lena's story through April — and get the Human Strengths Protection Map when it drops next week.

This is the final post in a four-part series following Maya, a composite VP of Marketing, navigating governance failures that arise when AI content workflows scale faster than judgment boundaries. Each post stands on its own. This week: what changed after she drew the line.

Three weeks after the competitor’s response to inaccurate competitive comparison content slowed a late-stage deal, Maya noticed a change in how work moved through her team.

The workflow itself remained largely intact; Campaign timelines stayed on track, and performance dashboards continued to indicate stability. The noticeable change appeared in the rhythm of decision-making. Questions surfaced earlier in the content production process. Legal partners reviewed claims before messaging reached final approval. Sales leaders began asking how marketing classified competitive positioning before using it in active buyer conversations.

The underlying system stayed the same, but visibility sharpened. Decision authority, once embedded quietly inside production steps, became an explicit leadership concern.

What Changed

After applying the Judgment Boundary Matrix, Maya introduced a classification checkpoint at the start of every externally facing content initiative.

Her team began evaluating messaging through two practical considerations: 1) the consequences of an inaccurate claim and 2) the degree of contextual judgment required to assess risk. They immediately moved competitive comparison content into a category requiring human approval. AI systems continued to support drafting content, but leaders assumed responsibility for publication judgment calls.

The team documented approval thresholds, clarified escalation ownership, and distinguished between permissions for drafting and permissions for distributing content externally. None of these adjustments required new technology. They required agreement about who would carry accountability as exposure increased.

This realignment altered how responsibility manifested in existing workflows.

What Became Harder

The first impact of establishing judgment boundaries appeared in production speed. Content that had previously cruised from draft to publication within hours now paused at each stage for deliberation. Teams now debated classification boundaries, occasionally escalating decisions that later proved routine. Managers recalibrated their tolerance for uncertainty while sales leaders continued to push for rapid responses when live deals were at stake. The team struggled with the slower pace.

Workflows that once felt efficient now felt heavier. Individual contributors questioned whether leadership overcorrected, and managers struggled to distinguish high-impact messaging from routine execution. Governance clarity introduced decision fatigue before it produced confidence.

These tensions reflected a transition from implicit judgment to explicit oversight, forcing the organization to confront trade-offs previously hidden.

What Became Easier

Over time, positive effects emerged. Public corrections became less frequent, and discussions about messaging gained depth. Cross-functional conversations pivoted from reactive problem-solving to earlier anticipation of potential consequences. Legal partners engaged more constructively, intervening before vulnerabilities reached the market rather than after.

Accountability also became easier to trace. When teams questioned a claim or positioning choice, they could quickly identify who had made the call and under what assumptions. Escalation processes felt more purposeful and less political. Exposure didn’t disappear, but leaders recognized it sooner and responded with greater coordination.

The organization began to treat governance less as a constraint and more as a mechanism to improve decision quality.

Operational Integration

As Maya continued refining governance in competitive messaging workflows, she noticed similar ambiguity in other areas of marketing execution. Customer segmentation models operated with limited human oversight. Campaign systems reallocated budgets and influenced buyer perception without a human in the loop. Meanwhile, automated partner outreach raised questions about authorization boundaries that had never been formally defined.

Addressing one area of vulnerability revealed others that had previously remained invisible.

Rather than launching a comprehensive governance initiative, Maya focused on targeted adjustments. Her leadership team began defining specific publication risk triggers, clarifying escalation ownership in customer-facing communication, and reviewing selected AI-enabled workflows with legal and compliance partners. These actions did not eliminate uncertainty, but they improved the organization’s ability to recognize emerging consequences before they escalated.

Governance maturity developed gradually through operational practice rather than policy declarations.

What Remains Unresolved

Consequences from the original incident continued to surface. Buyers raised credibility concerns in conversations with sales, and internal confidence remained low. Market perception shifted slowly, reminding leadership that reputational effects often outlast process improvements.

At the same time, governance clarity exposed new tensions. As decision authority became more explicit in competitive content workflows, leaders began questioning how much oversight other automated systems required. Each improvement revealed additional areas where judgment boundaries remained undefined.

For Maya, the experience reinforced a difficult but practical realization. Establishing judgment boundaries did not remove exposure. It reshaped how the organization recognized and managed it.

Greater clarity improved decision quality. It also increased leadership responsibility for the outcomes that followed.

This is the third post in a four-part series on judgment boundaries in AI-assisted marketing. Each post stands on its own. In Week 1, Maya’s competitive comparison content surfaced on LinkedIn before her team knew it was live. In Week 2, the Judgment Boundary Matrix identified the governance gap that enabled it. This week: four incidents from other industries that show the same structural failure at different scales.

The competitor’s response to the inaccurate content went viral — first publicly, then internally. Sales leaders began forwarding screenshots. A late-stage pipeline conversation halted. What had felt like routine execution was now a reputational conversation unfolding in real time.

Maya’s team had built an AI-assisted workflow for competitive comparison content, and it was working brilliantly. The workflow shortened drafting cycles, speeding publishing cadence. This automated process they deployed felt operational rather than strategic. Maya assumed the risk of using it was minimal.

She was wrong. The market quickly reacted to the published content. And not in Maya’s favor.

Working through the Judgment Boundary Matrix last week reframed the incident. The issue was not only content accuracy. It was also decision ownership. The workflow treated competitive comparison content as low-stakes production. In practice, it carried brand, legal, and revenue exposure. A high-impact decision had moved through a low-friction system. undetected.

This week, Maya isn’t trying to resolve the incident. She is trying to understand what allowed it to happen.

Across industries, similar failures are emerging in marketing-adjacent workflows. Contexts differ. Consequences vary. The structural gap across these failures is consistent. Automation increases output, but it can also make decision authority harder to see.

When drafting quietly becomes acting

In November 2025, Zoho CEO Sridhar Vembu received an acquisition pitch from an unnamed startup founder. The email included more than a pitch. It disclosed that another company was already in negotiations and revealed the competing price.

Moments later, a second email arrived, but not from the founder. It came from the founder’s browser AI agent, which had identified the error and transmitted an unsolicited apology to Vembu without the founder’s knowledge or approval.

The governance question this incident raised wasn’t about the original disclosure, which could be attributed to the founder’s judgment. The question was why an AI system held authority to transmit external communications in a negotiation context without a human review gate. The agent had drafting and sending access. No boundary distinguished the two.

Marketing leaders encounter this pressure point more often than they expect, through partner outreach, analyst briefings, and executive ghostwritten content. When AI systems execute inside these workflows, decision authority is easy to overlook. Once an organization transmits a message, it no longer shapes intent internally. It shifts to managing external perception.

That distinction belongs in the workflow before the message goes out, not after.

When a missing review step becomes a vulnerability

In December 2025, Unit 42, Palo Alto Networks’ threat intelligence team, documented a real-world attack designed to exploit an AI-based ad review system. Attackers embedded hidden instructions inside a deceptive advertorial page, tricking the AI reviewer into approving content it would otherwise have rejected.

The attack succeeded because of a governance gap, not a technology failure. The AI system held final decision authority. Human escalation protocols existed but had no defined triggers. There was no threshold for unfamiliar domains, unusual claim patterns, or new advertiser identities that would route a submission to human review.

A human in the approval chain, with a clear escalation trigger, would have caught it.

This is the “it can happen to you” case for marketing leaders managing AI-assisted media or content review. The attack vector was external. The governance gap was internal. When AI holds final authority in brand safety, content review, or media placement, accountability for the outcome must live somewhere. If leadership hasn’t explicitly assigned it to a person, it defaults to the model. And models can be manipulated.

Publication permission is a leadership decision. Wherever that permission lives in the workflow, authority lives there too.

When AI states policy it has no authority to state

In April 2025, a developer contacted Cursor’s customer support after being repeatedly logged out when switching between devices. The response came from an AI agent named Sam, which informed the user that Cursor allowed only one active device per subscription, a core security feature.

No such policy existed.

The fabricated limitation spread across developer communities within hours. Subscription cancellations followed before Cursor’s leadership could intervene. The co-founder eventually apologized on Hacker News, confirmed the user had been refunded, and acknowledged the error.

What the incident exposed was not a model performance failure alone. It was a boundary failure. Cursor’s support system held customer-facing authority over policy communication without a defined limit on what it could assert as fact. The model filled the gap the way models do: confidently, and incorrectly.

Customer-facing authority carries disproportionate consequences in marketing terms. Trust erosion that begins in support interactions surfaces later in retention metrics, campaign response rates, and brand advocacy signals. Recovery extends beyond corrective action into reputational rebuilding.

One boundary — “AI may surface verified policy, not interpret or state it” — would have changed the outcome.

When user permission is treated as platform authorization

In March 2026, a federal judge granted Amazon a preliminary injunction blocking Perplexity’s Comet browser from accessing password-protected sections of Amazon on behalf of users.

Perplexity designed Comet to extend user convenience; Amazon’s platform read it as an unauthorized access violation. The court found that user permission and platform authorization are distinct, and that operating inside a third-party system without platform consent is not a user-rights question; it’s an access question.

Marketing consequences followed quickly. Claims about ecosystem compatibility required revision. Growth narratives tied to distribution partnerships required recalibration. Leadership attention shifted from expansion to defensibility.

For marketing leaders operating in partner-dependent environments, the authorization boundary is worth examining before the workflow is built. It is not only what a user permits. It is what each platform in the distribution chain explicitly authorizes. When AI agents operate inside those systems without that clarity, the exposure is not a performance risk. It is a permission risk.

Where judgment boundaries actually break

Reviewing these incidents side by side, Maya noticed that the public consequences appeared too late for the organizations to correct course. Authority had moved gradually from human judgment to workflow assumption. The market response revealed what had already shifted internally.

Each scenario began with efficiency gains. Workflow friction decreased, output increased, and decision ownership became less visible. Leadership attention stayed anchored to performance indicators while governance assumptions went untested.

Maya’s situation was smaller in scale. It involved a competitive comparison asset, a reactive LinkedIn thread, and a weekend workflow audit. But the architecture of the failure was the same. A workflow had been designed for production efficiency. Decision authority had not been assigned. When the content went live, the process performed exactly as designed.

These incidents are not cautionary tales about AI going rogue. They document what happens when governance assumptions go untested at the point where automation meets external consequence.

What mature marketing teams do before the incident

Teams that avoid similar failures tend to define escalation triggers before automation expands: specific content types, claim categories, or relationship contexts that require human classification before entering the workflow.

They separate drafting from execution in external-facing communication. An AI system that can draft can only draft. One that can send requires a human approval gate for outbound action.

They restrict customer-facing AI to verified information retrieval. Interpretation, policy assertion, and judgment calls stay with the people who hold accountability for the answer.

They treat publication and transmission as leadership decisions at Judgment Boundary Q2 and above, not workflow outcomes.

None of these practices prevents every error. What they do is make accountability visible before errors reach the market.

For Maya, reviewing these incidents reframed her original decision. The workflow had not malfunctioned. It had performed exactly as designed. The design itself lacked a clear authority boundary. Recognizing that distinction is uncomfortable. It is also actionable.

Next week: Maya revisits the decision she made after reclassifying competitive comparison content as a human-authority call. Some risks became easier to manage. Others became more visible. Clarity moves leadership forward. It does not eliminate cost.

Sources

Zoho CEO Sridhar Vembu on X, November 28, 2025 — an unnamed startup founder’s browser AI agent disclosed confidential acquisition details and autonomously sent an apology without the founder’s knowledge. Reported by The Hans India and Business Today.

Unit 42, Palo Alto Networks (December 2025): Real-world indirect prompt injection attack designed to bypass an AI-based ad review system — unit42.paloaltonetworks.com/ai-agent-prompt-injection/

Cursor / Anysphere (April 2025): AI support agent fabricated a one-device subscription policy, triggering cancellations and a public apology from the co-founder. Reported by Fortune, The Register, and CX Today.

Amazon v. Perplexity, U.S. District Court, Northern District of California (March 9, 2026): Preliminary injunction blocking Perplexity’s Comet browser from accessing password-protected Amazon accounts — cnbc.com/2026/03/10/amazon-wins-court-order-to-block-perplexitys-ai-shopping-agent.html

The audit

Maya spent the weekend after the LinkedIn thread doing what most VPs of Marketing do after a public failure: auditing backwards. She pulled every AI-assisted workflow her team ran. Twelve of them.

At first glance, the competitive comparison workflow looked like the others. It had a review step and a quality check, which meant that on paper, the process worked. Several members of her team had already pointed this out in the initial conversations. The system did what it was designed to do.

That explanation held up until Maya looked more closely at what the review step was actually supposed to accomplish.

She realized the workflow included a review step, but it lacked a standard for what review meant for that content category. Her team treated it as a quality check when the situation called for a judgment call. The AI workflow never distinguished between those two things.

The question her workflow never asked

Most AI content workflows ask two questions before they publish: is the content accurate, and is it on brand?

For routine content, those are the right questions. For content that carries brand, legal, or competitive risk, they are incomplete.

The question missing from Maya’s workflow was: Does this decision belong to AI, or to a human?

Without that question, every piece of content moved through the same gate regardless of what was at stake if it was wrong. A blog outline and a competitive positioning claim require very different levels of human oversight. Maya’s process treated them identically.

What her team needed was a way to classify those decisions before the workflow began.

Classification has to happen before review. The review step existed, but the classification standard did not. That distinction turned out to be the difference between a functioning process and one that quietly allowed risk through the system.

A framework for drawing the line

Maya’s team had already built a review step into the workflow. What the process lacked was a classification standard that defined which decisions required human judgment before the content entered the review stage.

The Judgment Boundary Matrix is designed to solve that problem.

The framework gives marketing teams a repeatable way to classify AI-assisted content decisions before they enter the review process. It maps decisions across two factors: how much is at stake if the content is wrong, and how much contextual judgment the decision requires. The intersection of those two dimensions determines who owns the decision.

Use the matrix to classify the decision before the workflow reaches the review stage.

Each quadrant defines a different level of human authority in the workflow.

Read more

By the time Maya saw the LinkedIn thread, the post had 200 reactions and a tagged competitor.

Her CMO sent the screenshot with a single question:

“Did we approve this?”

The CMO’s question arrived with a LinkedIn screenshot attached: a competitive comparison Maya’s team had published through their AI content workflow, now live in a thread with 200 reactions and a tagged competitor.

The claims had been accurate when the workflow was built. They were not accurate when the content was published; No one flagged them for re-review because no one owned that decision.

What her CMO was really asking was whether Maya had a system for knowing which decisions needed a human. And she didn’t have a clear answer.

THE SITUATION

What was working and what it was hiding

Maya is VP of Marketing at a 200-person B2B SaaS company, 18 months into an AI transformation championed by her CMO. She assumed all was well. Her team delivered. Adoption metrics were strong. Workflow documentation was complete. AI usage across content, campaigns, and competitive intelligence became routine.

Maya’s team ran twelve AI-assisted workflows, and competitive-comparison content was among them. The asset had cleared the standard review process. No one questioned it because the AI workflow said it had cleared review, and clearing review had always been enough. That’s what made it a structural problem, not a human one.

WHY THIS MATTERS NOW

Adoption moved fast. Governance didn’t follow.

The pattern Maya walked into isn’t unusual. In a May–June 2025 Gartner survey of 360 IT leaders involved in generative AI rollouts, only 23% reported being very confident in their organization’s ability to manage security and governance when deploying GenAI tools. Over 70% cited regulatory compliance as a top-three challenge.

Adoption outpaced governance. That gap is where the exposure lives.

Gartner projects that by 2028, enterprises will spend more than $30 billion battling misinformation and disinformation, cannibalizing 10% of marketing and cybersecurity budgets combined. That figure reflects what happens when internal content governance doesn’t keep pace with the volume of content. Bad actors are part of the equation. Ungoverned internal workflows are too.

The accountability gap is documented: when AI-generated content causes legal or reputational damage, the liability belongs to the humans and organizations that published it. The tool doesn’t get sued, and the vendor doesn’t answer to the board. The marketing team does.

THE GAP

Designed to fail

Maya’s team didn’t make a careless mistake; They followed the process. What the process didn’t include was a defined point at which a human had to step in and own the judgment call. Research on human-AI decision-making published in Scientific Reports in early 2026 identifies this failure mode. When AI workflows lack explicit intervention triggers, humans shift from active control to passive monitoring and systematically fail to intervene when systems err. The team wasn’t negligent. They were operating exactly as humans do inside workflows that never told them when to stop and decide.

Competitive positioning. Claims about named competitors. Content that could attract legal scrutiny or go viral for the wrong reasons.

These are decisions, not tasks.

Maya’s workflow treated them as tasks.

The boundary was never defined, so no one crossed it. It simply didn’t exist.

WHERE WE LEAVE MAYA

The unanswered question

Maya knows the asset was wrong. She knows how it got published.

What she doesn’t have is a system for knowing where judgment belongs inside the twelve AI workflows her team runs.

Her CMO asked, “Did we approve this?”

Maya doesn’t have that answer yet.

She will by the end of the month.

This month I’m experimenting with a serialized format. One situation, explored over four weeks.

Here’s what you can expect:

Chapter 1 (this post): The protagonist’s situation (Maya) and the judgment boundary failure that created it.

Chapter 2: The tool Maya uses to drive change: the Judgment Boundary Matrix, a framework for mapping decisions by impact severity and context complexity, with the downloadable decision tool (for paid subscribers).

Chapter 3: What Maya learns from other marketing leaders who are drawing judgment boundaries and what they got wrong before they got it right.

Chapter 4: Maya revisits her solution. What shifted, what she’d do differently, and the one boundary she and her CMO still disagree on.

I’d love to hear your feedback on this new format. Share it in a comment!

Leave a comment

SOURCES

Gartner (May–June 2025 survey of 360 IT leaders): AI Regulatory Violations Will Result in a 30% Increase in Legal Disputes for Tech Companies by 2028 — gartner.com/en/newsroom/press-releases/2025-10-06-gartner-predicts-ai-regulatory-violations-will-result-in-a-30-percent-increase-in-legal-disputes-for-tech-companies-by-2028

Gartner: Enterprise Spending on Battling Misinformation Will Surpass $30 Billion by 2028 (October 2025) — gartner.com/en/newsroom/press-releases/2025-10-21-gartner-predicts-enterprise-spending-on-battling-misinformation-and-disinformation-will-surpass-30-billion-dollars-by-2028

Gartner: 50% of Enterprises Will Invest in Disinformation Security and TrustOps by 2027 (November 2025) — gartner.com/en/newsroom/press-releases/2025-11-20-gartner-predicts-50-percent-of-enterprises-will-invest-in-disinformation-security-and-trustops-by-2027

Cummings & Cummings Law: Legal Issues in Using AI-Generated Content for Business Marketing (January 2026) — cummings.law/legal-issues-in-using-ai-generated-content-for-business-marketing/

Scientific Reports / Nature (2026): Examining human reliance on artificial intelligence in decision making — nature.com/articles/s41598-026-34983-y