The 45-Point Gap That Sinks AI Rollouts
76% of executives believe their employees are enthusiastic about AI. Only 31% of employees actually are.1
That's not a rounding error. It's a structural misread of the room. And it shows up in the data on both sides: nearly half of CEOs separately report that most of their employees are resistant or openly hostile to AI-driven changes.2 So the gap goes both directions— leaders overestimate enthusiasm, and when they finally hear resistance, they treat it as the cause of failure rather than the symptom.
Most AI rollouts don't fail because employees are difficult. They fail because leaders are misreading the room and acting on bad assumptions about what the org has already heard, decided, and discounted. Silence isn't neutral. It's a signal— and the org reads it correctly.
The gap matters because what closes it isn't a better tool. It's leadership behavior. And the data on which behavior matters most is unambiguous.
CEO Oversight Is the #1 Lever on AI's EBIT Impact
CEO oversight is the single element most strongly correlated with EBIT impact from generative AI— outranking every other leadership, technology, or process variable McKinsey measured in 2025.3
That finding should change how a founder-CEO spends their next 90 days. It almost never does. Three McKinsey datapoints, stacked, explain why:
- Only 39% of organizations report any EBIT impact attributable to AI at all.4 The other 61% are spending on tools and getting nothing back.
- Only 33% of organizations say senior leadership even somewhat understands how AI creates value.5 The prerequisite for sponsorship is missing.
- CEO oversight outranks every other leadership variable McKinsey tested.3 The work cannot be delegated to the CIO.
This matters disproportionately for founder-led firms in the $20M–$100M range. In a 50,000-person enterprise, layers of management absorb and dilute the founder's signal. In a 200-person AEC or professional services firm, the founder's voice carries the room— for better or worse. A solid AI strategy doesn't begin with tool selection. It begins with the CEO deciding what they're going to say in the first all-hands and what they're going to keep saying after.
AI is a leadership problem wearing a technology costume. The costume is convincing, and that's why so many founders hand it off.
Knowing CEO oversight matters isn't the same as knowing what oversight looks like. The behavior that predicts success— or its absence— has a specific shape.
What "The Opening Act" Actually Is
The opening act— the term I use for the first all-hands or company-wide communication after the decision to deploy AI— is not a status update. It is a directional commitment, and it sets the entire trajectory of adoption.
Most CEOs think they've given an opening act because they mentioned AI in a town hall. They haven't. What they gave was ceremony. The org doesn't hear ceremony— it watches for commitment. And it has gotten very good at telling the difference.
The opening act is not "AI is the future." It is "here are the three workflows we're redesigning, here's the budget, here's how we'll measure it, and here's what I personally will be reviewing every two weeks." Ceremonial talk costs nothing. Operational commitment costs the CEO their next 90 days.
| Ceremonial Talk | Operational Opening Act | |
|---|---|---|
| Tone | "AI is the future of our industry." | "We are redesigning estimating, RFI response, and submittal review." |
| Workflows | Unnamed; abstract reference to "productivity." | Named, with owners and timelines. |
| Budget | "We're investing in AI." | "30% of the rollout budget is going to change management. Here's what that buys." |
| Cadence | One announcement, then silence. | Bi-weekly review owned personally by the CEO for the first 90 days. |
| Employee concern | Unaddressed or dismissed. | Named directly, with what is and isn't on the table. |
BCG's research backs the imperative: clearly communicating the AI vision is "absolutely vital" for maintaining culture, retaining talent, and easing organizational fear.6 But vision without operational commitment reads as posturing. Gartner's I&O survey reinforces this from the success side: among AI use cases that succeed, success is attributed to integrating AI into existing workflows and securing full support from business executives.7 Tools don't drive that. CEOs do.
This is why the work of AI implementation starts upstream of any pilot. If the opening act sets the trajectory, the silence after it sets the cost. And the cost shows up in three predictable places.
The Three Costs of CEO Silence
CEO silence after an AI announcement produces three predictable failures: workflows don't get redesigned, change management stays underfunded, and projects lose sponsorship before they ship value.
Cost 1: Workflows don't get redesigned
Only 21% of organizations using gen AI have redesigned a single workflow.8 The other 79% are paying for tools they've layered on top of broken processes— and that's the cohort that doesn't show EBIT impact. Workflow redesign requires CEO authority to override department-level resistance. Department heads protect existing processes because their headcount, budget, and identity are tied to them. Only the CEO can make redesign non-optional. When the CEO is silent, every department reverts to its prior shape and quietly absorbs the AI tool as a productivity layer rather than a structural change.
Cost 2: Change management stays underfunded
Successful AI transformations spend 30–40% of budget on change management. Most spend about 10%.9 That's a 3–4x underfunding pattern, and it is not an accident of finance. The CEO controls budget allocation at that scale. Underfunding is a leadership choice— a signal to the CFO and the board that this is a tooling line item, not an operating-model change. The change-management gap is the silence, denominated in dollars.
Cost 3: Projects fail at industry-typical rates
Between 70% and 85% of AI projects fail to deliver expected benefits— roughly twice the failure rate of traditional IT projects.10 Gartner's I&O data converges: only 28% of AI use cases fully succeed, while 20% fail outright.11 When success does happen, executive support is one of the two named drivers.7 Sponsorship isn't a tiebreaker. It's a prerequisite.
A brief acknowledgment is fair: 61% of CEOs report their boards are pushing for faster AI transformation than they think is prudent.12 Board pressure is real, and it's upstream context. But it's not an excuse. The CEO still owns the cadence inside the firm. Speeding up the announcement while staying silent on workflows, budget, and review cadence produces the failure pattern faster, not better.
Naming the costs is the easy part. The harder question is what to actually say in the opening act— and what to keep saying after.
What to Say in the Opening Act (And Every 30 Days After)
An effective opening act covers five elements: the why, the named workflows, the budget commitment, the review cadence, and the explicit acknowledgment of employee concern. Skip any one and the rollout reads as ceremony.
- The why. What business problem AI is solving in this firm specifically. Not "AI is the future." "We are losing 14 hours a week per project manager to RFI churn, and we are going to fix that." BCG's framing is right: the vision has to be communicated clearly or talent and culture pay the cost.6
- The named workflows. Two or three specific processes being redesigned, with owners. Not "operations." Estimating. Submittal review. Project closeout. Naming workflows is what authorizes department heads to actually change them.
- The budget commitment. Explicit dollars or percentage allocated to change management, named publicly. If it's not 30–40%, say what it is and why.9 The org needs to hear the number to believe it's real.
- The review cadence. The CEO's personal review interval, owned by them, not delegated. Bi-weekly for the first 90 days is the floor. What employees watch for isn't enthusiasm. It's whether the CEO returns to the topic in 30 days.7
- The acknowledgment. Direct address of employee concern about jobs, identity, and skill obsolescence. Leaders who skip this confirm the worst fears by omission. People are the answer here. AI amplifies what humans bring; it doesn't replace the bringer.
The cadence matters as much as the content. Single announcements decay. The 30-day return is the signal that this is real.
If your opening act would fit on a poster, it isn't an opening act. And if mapping the right workflows and the right opening-act content feels like a separate full-time job, that's exactly the kind of work an outside implementation partner does in a fraction of the time.
Frequently Asked Questions
Why do most AI projects fail?
Between 70% and 85% of AI projects fail to deliver expected benefits— roughly twice the failure rate of traditional IT projects.10 The pattern across McKinsey, BCG, and Gartner research is consistent: failures are leadership and operating-model failures wearing a technology costume. CEO sponsorship and workflow redesign predict EBIT impact more strongly than tool selection or data quality.3
What is the CEO's role in AI adoption?
Set direction publicly, name the specific workflows being redesigned, fund change management at 30–40% of budget, and personally review progress on a sustained cadence. McKinsey's 2025 research identifies CEO oversight as the single biggest predictor of EBIT impact from gen AI.3 In founder-led firms, delegating this role to a CIO is itself the failure pattern— the founder's voice is the lever, and silence from it is read as a verdict.
How much should companies spend on change management for AI?
Successful AI transformations typically spend 30–40% of total transformation budget on change management. Most companies spend about 10%— a 3–4x underfunding pattern that BCG identifies as one of the strongest predictors of failure.9 The CEO controls budget allocation at this scale. Underfunding is a leadership choice, not an oversight.
The Opening Act Is a Choice You're Already Making
Every CEO is currently giving an opening act about AI— including the ones who've said nothing. Silence is itself a message, and the org has already received it.
The cost of staying silent isn't measured in what you didn't say. It's measured in the workflows that weren't redesigned and the budget that wasn't allocated because no one believed it was real.
If this maps to your situation, naming the right opening act and the right workflow priorities is exactly the kind of work an outside partner can de-risk before the announcement. Dan Cumberland Labs helps founder-led AEC and professional services firms get the architecture right before the first all-hands— so the opening act is a commitment the org can actually feel.
References
- Harvard Business Review, "Leaders Assume Employees Are Excited About AI. They're Wrong." (2025) — https://hbr.org/2025/11/leaders-assume-employees-are-excited-about-ai-theyre-wrong
- HR Dive, "Nearly Half of CEOs Say Employees Are Resistant or Even Hostile to AI" (2025) — https://www.hrdive.com/news/employers-employees-resistant-hostile-to-AI/749730/
- McKinsey & Company, "The State of AI: How Organizations Are Rewiring to Capture Value" (2025) — https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-how-organizations-are-rewiring-to-capture-value
- McKinsey & Company, "The State of AI: How Organizations Are Rewiring to Capture Value" (2025) — https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-how-organizations-are-rewiring-to-capture-value
- McKinsey & Company, "The State of AI: How Organizations Are Rewiring to Capture Value" (2025) — https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-how-organizations-are-rewiring-to-capture-value
- Boston Consulting Group, "Why CEOs Must Lead Data Conversations for AI Success" (2025) — https://www.bcg.com/publications/2025/ceos-must-lead-data-conversations-for-ai-success
- Gartner, "Gartner Says AI Projects in I&O Stall Ahead of Meaningful ROI Returns" (2026) — https://www.gartner.com/en/newsroom/press-releases/2026-04-07-gartner-says-artificial-intelligence-projects-in-infrastructure-and-operations-stall-ahead-of-meaningful-roi-returns
- McKinsey & Company, "The State of AI: How Organizations Are Rewiring to Capture Value" (2025) — https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-how-organizations-are-rewiring-to-capture-value
- Boston Consulting Group, "Five Barriers CEOs Must Overcome for AI Impact" (2026) — https://www.bcg.com/publications/2026/five-barriers-ceos-must-overcome-for-ai-impact
- Boston Consulting Group, "Five Barriers CEOs Must Overcome for AI Impact" (2026) — https://www.bcg.com/publications/2026/five-barriers-ceos-must-overcome-for-ai-impact
- Gartner, "Gartner Says AI Projects in I&O Stall Ahead of Meaningful ROI Returns" (2026) — https://www.gartner.com/en/newsroom/press-releases/2026-04-07-gartner-says-artificial-intelligence-projects-in-infrastructure-and-operations-stall-ahead-of-meaningful-roi-returns
- Boston Consulting Group, "Sixty-One Percent of CEOs Say Their Boards Are Rushing AI Transformation" (2026) — https://www.bcg.com/press/4may2026-ceos-say-boards-rushing-ai-transformation
