Workplace AI Adoption Stalls Where Manager Influence Runs Thin

The accountability gap at the center of the AI boom

Companies have spent the past two years buying AI. Many of the people now expected to show results from it had no say in what was bought. That is the argument Forbes contributor Tony Gambill makes in a new piece. He describes a growing group of leaders who are judged on the returns from AI investments but have little control over which technology is chosen, when it rolls out, or how their teams' work is reorganized around it11. He calls this "accountability without authority" and says the gap between adopting AI and seeing results from it is a leadership influence problem as much as a technical one11.

His example is a functional leader whose team must use an AI platform picked by enterprise technology executives. That leader did not choose the tool, set the timeline or decide how workflows would change. The same leader still takes the blame if employees resist, service quality drops or the promised productivity gains never show up12. His proposed fix has three parts: build trust, read situations accurately and adapt to them, and handle high-stakes conversations through a framework he calls ACES11.

On its own, this could be read as standard management advice. The workplace data published over the past several months suggests otherwise. Taken together, it shows that whether AI spending pays off depends heavily on the human layer between executives and the front line. That layer is under strain.

Broad usage, thin returns

The starting point is a gap almost every major survey now records. Gambill cites McKinsey's 2026 State of AI research. Nearly nine in ten respondents report regular AI use in at least one business function, yet only 37% link any positive effect on earnings before interest and taxes to AI, and just 6% count as high performers12. A PwC survey of 4,454 CEOs released in January found that only 30% saw revenue rise from AI over the prior year, and 56% said it had brought no cost or revenue improvement at all21. Gartner estimates that only one in five AI investments produces any measurable return21.

The widely quoted claim from MIT's Project NANDA, that 95% of generative AI pilots had no measurable effect on profit and loss, has become shorthand for the problem23. It should be read as a direction rather than a precise figure. Critics note that the report was preliminary, was not peer-reviewed, relied on self-reported data and measured success over roughly six months26. Independent data from S&P Global, IDC and IBM still points to a real gap between pilots and production26.

The more important point is how analysts explain the gap. Coverage consistently says the cause is not the models themselves. They point to workflow fit, data readiness and governance28. McKinsey's research found that redesigning workflows, not the tool chosen, was the strongest predictor of financial impact among high performers28. Workflows are redesigned by managers and teams, which is why the influence argument matters.

The manager effect

Gallup's data puts a number on the case. Among employees at organizations adopting AI, 78% of those who strongly agree their manager supports AI use frequently, compared with 44% of everyone else5. Gallup counts manager support as one of the two conditions most closely tied to frequent use, along with building AI into workflows. Yet only 36% of employees at AI-integrating organizations strongly agree that their manager actively supports their team's use of the technology5.

In effect, the most important adoption lever in the data is being pulled by about a third of managers. Separate research cited by Worklytics puts the effect at employees being 2.5 times more likely to use AI when their leaders support and encourage it14.

The gap between executives and the front line is equally consistent. In one study, 83% of executives believed they had communicated a clear AI vision, while only 37% of frontline employees said the message had reached them14. Boston Consulting Group found that only 25% of frontline employees get enough guidance from leadership on using AI well2. Microsoft's 2026 survey of 20,000 knowledge workers who use AI found that only 26% describe their leadership as consistently aligned on AI4. Pigment's survey of finance chiefs found the same pattern inside companies: 34.8% of CFOs rated their organization as "leading" on AI maturity, compared with 16.3% of managers22.

The Society for Human Resource Management (SHRM) makes the trust problem explicit. Its survey of 5,875 U.S. workers found that 74% of directors and above had been told about AI plans before rollout, compared with only a third of individual contributors3. Trust in senior leaders when they talk about AI follows the same pattern: 80% among directors and above, 47% among individual contributors. Another 38% of individual contributors said they were unsure3. Heavy reliance on AI by leaders also cut trust for 26% of individual contributors3.

These figures support Gambill's central claim. Messages from the top are not getting through, and the trust needed to carry them is weakest among the people asked to change the most.

Fear complicates the message

Gambill notes that what looks like resistance may really be worry about job security rather than reluctance to learn11. Managers seem to know this. In Beautiful.ai's survey of 3,000 American managers, 72% said their employees fear AI will make them less valuable, and 70% said employees fear it could eventually cost them their jobs8. The same survey found that 35% of managers now think replacing employees with AI would benefit their company, up from 23% a year earlier8.

Executive behavior is adding to the strain. In WRITER's survey of 1,200 C-suite executives, 75% admitted their AI strategy is "more for show" than real guidance. Sixty percent plan layoffs for employees who do not adopt AI, and 77% said those who refuse to become proficient will not be considered for promotion6. A strategy built mainly on threats is hard to square with the trust-first approach Gambill describes. It also leaves middle managers explaining a policy they may not believe in to teams that have good reason to be wary.

Where the coverage diverges

The sources agree that leadership matters. They disagree on which leaders matter most.

Gambill's focus is the leader in the middle, who works upward and across teams without formal authority11. Others look higher up. One AI advisory firm argues that the strongest predictor of success is having an executive who owns AI outcomes full-time15. BotsCrew points to Prosci data showing that 43% of respondents blamed failed adoption on weak executive sponsorship17. A Consultancy.eu commentary argues that leadership development should start in the boardroom and work downward, because organizations need direction before adoption20.

Deloitte takes a third view. Its modeling suggests results are strongest when decision rights are shared among complementary executives, such as the CTO, CFO and chief strategy officer, rather than held by one person18.

The positions are less opposed than they look. Clear direction from the top explains why the AI programs exist. Middle managers turn that direction into changed behavior. A business with one strong layer and a weak other layer will probably stall either way.

One real tension remains. Leaders' own reports are much rosier than independent measures. In GoodFirms' survey of 343 business leaders, 93.7% said they were confident their AI investments are worth the money9. That sits awkwardly beside PwC's CEO figures and Gallup's finding that only 12% of workers globally think AI has meaningfully changed how their organization works2. The more reliable reading is the more cautious one.

The capacity problem nobody budgets for

Talk of influence can make this sound like a question of soft skills alone. Writing in Inside Public Accounting, Stacy Parker raises a structural objection. Managers are already carrying hybrid teams, compliance demands, growing lists of performance targets and now AI rollouts, and every new corporate priority lands on the same layer16. Asking that layer to also build trust across functions and lead difficult conversations assumes time it may not have.

Parker argues that AI could free managers from reporting and coordination work so they can coach and lead. She warns that the freed time will be wasted if it fills up with more meetings16.

Budgets point the same way. Deloitte research cited by one consultancy found AI spending split roughly 93% on technology and 7% on people29. Another analysis says only about a third of organizations invest seriously in change management alongside their AI projects24. Gallup reports that frequent AI use among leaders has risen to 51%, while individual contributors sit at 26%5. The people who use AI most are also the ones setting expectations for everyone else.

The reading

Gambill's argument holds up, but it understates the problem. Managers' influence is clearly the strongest lever in the current adoption data, and the Gallup gap between 78% and 44% is too large to put down to individual temperament5. Influence cannot fully make up for missing authority, though.

Companies that hold managers responsible for AI results should also give them a say in choosing tools, set aside budget for workflow redesign and make time for the trust-building that frameworks like ACES require. Otherwise, "leadership influence" becomes a polite term for shifting blame onto the people with the least control over whether AI succeeds.