75% of executives admit their company's AI strategy is more for show than actual guidance. 54% say AI is tearing their company apart. The technology is working. The organisation underneath it isn't. Here's why - and what the 29% actually getting ROI are doing differently.
I've sat in enough transformation reviews to know what a healthy one looks like - and what a performance looks like. A healthy one has friction. People push back on timelines. Someone raises a concern about data readiness that wasn't in the deck. There's a real argument about sequencing. A performance is polished. The numbers are moving in the right direction on the slides. The executive sponsor is enthusiastic. Everyone in the room agrees that AI is the priority.
Six months later, the performance shows. Nothing has changed at the delivery layer. The tools are deployed, the licences are paid, the pilot metrics were technically achieved - and the frontline teams are working around the AI the same way they worked around the last three initiatives that were also declared the priority.
This is not a fringe experience. Writer's 2026 enterprise AI adoption survey, which covered 2,400 executives and employees globally, found that 'Enterprise AI adoption in 2026: Why 79% face challenges despite high investment' - and buried in that finding is a number that should stop every CEO mid-presentation: 75% of executives admit their company's AI strategy is "more for show" than actual internal guidance. Not competitors' strategies. Their own.
The technology is not the bottleneck. It never was.
The Gap Nobody Is Measuring
Here is the data point that should be in every board report but isn't: only 29% of organisations see significant ROI from generative AI, despite 97% of executives reporting their company deployed AI agents in the past year. That means three out of four organisations have committed budgets, announced AI transformation, deployed platforms - and produced negligible business impact.
The explanation most leaders reach for is adoption. "We need to get people using it." That framing is accurate but incomplete. The actual gap is structural. Individual AI super-users - the employees who have genuinely integrated AI into how they work - are 5X more productive than colleagues who haven't. That productivity is real, measurable, and happening inside the same organisations that report no significant ROI. The individual wins exist. They just aren't connecting to organisational outcomes.
What's in between the individual and the organisation? The layer that approves workflows, allocates resources, sets team priorities, manages performance, and decides what actually gets implemented. Middle management.
HBR's April 2026 analysis by researchers from the Wharton School, 'Managers and Executives Disagree on AI - and It's Costing Companies', puts it directly: AI initiatives stall not because leaders lack ambition or investment, but because middle managers - the people charged with making transformation work - see a fundamentally different reality from executives. The C-suite sees strategy. Middle managers see their teams, their timelines, their accountability structures, and an AI initiative that was handed to them without any of those things being redesigned.
Why Middle Managers Are Carrying a Weight Nobody Acknowledged
I want to be precise here, because the default narrative - middle managers are resistant, therefore they're the problem - misses what's actually happening.
Middle managers in 2026 are being asked to lead AI transformation for their teams while simultaneously worrying that AI is coming for their own roles. Gartner's research flagged that 20% of organisations plan to use AI to eliminate more than half of their middle management positions by 2026. That's not a secret. The people you're asking to drive adoption have read the same headlines. They know what the automation roadmap implies about their layer of the organisation.
Now add the execution reality on top of that. The AI initiative arrives as a mandate from above, often without redesigned processes, without clarity on how success is measured, without budget for the change management that implementation actually requires, and without answers to the questions teams will ask on day one: What do I do when the AI is wrong? Who approves what it outputs? Does using AI count against my headcount justification?
The middle manager who quietly deprioritises the AI rollout while hitting their other metrics is not being obstructionist. They're making a rational calculation about where their attention produces results that they will be held accountable for. That calculation is correct given the incentive structures most organisations haven't changed.
The Writer survey captures what this produces downstream: 29% of employees admit to actively sabotaging their company's AI strategy - and among Gen Z employees, that figure jumps to 44%. These aren't people opposed to technology. They're people who lost trust in a transformation that wasn't designed with them in mind.
The Performance Art Problem
The 75% figure - three quarters of executives admitting their AI strategy is more for show than guidance - is the most important data point in the Writer survey because of what it reveals about organisational psychology under pressure.
CEOs are under genuine stress about AI. The same survey found that 73% of C-suite leaders report stress or anxiety about their AI strategy, with 64% fearing they could lose their jobs if they fail to lead the AI transition. Under that kind of pressure, a strategy document, an AI governance framework, a transformation roadmap - these provide organisational cover. They signal that the company is moving, that leadership is committed, that the board question about AI has an answer.
What they often don't do is change how work gets done three levels below the presentation. The strategy exists at the level of slides and org charts. The actual operating model - how decisions get made, how outputs get approved, how accountability flows, how performance gets evaluated - hasn't moved.
This is the specific failure mode that separates the 71% reporting disappointment from the 29% generating real returns. The 29% didn't just deploy AI. They redesigned something: a workflow, a team structure, an accountability model, a governance process. The transformation touched the operating model, not just the technology stack.
What Genuine Transformation Actually Requires
I've watched enough enterprise technology rollouts - at Adobe managing large-scale platform deployments, at Zendesk working through CS tool integrations across enterprise accounts, and at Intelegencia delivering transformation for clients across industries - to recognise the pattern of what works and what doesn't.
What doesn't work: deploying AI tools, measuring adoption rates, and calling that transformation. Adoption is a leading indicator, not an outcome. The metric that matters is whether the AI is changing what people decide and how fast they decide it - not whether they logged into the platform this week.
What works looks less elegant in a board presentation. It requires identifying the two or three workflows where AI can change a business outcome - not just a productivity metric - and rebuilding those workflows around how the AI actually operates. It requires naming who owns the AI's outputs, who handles exceptions, and what the escalation path looks like when it gets something wrong. It requires middle managers who have been given the mandate, the resources, and the redesigned incentive structures to actually lead adoption rather than manage it alongside everything else.
It also requires accepting that the transformation will be slower and narrower than the announcement suggested - and that the narrow, deep version produces compounding results that the broad, shallow version never does.
The 5X productivity gap between AI super-users and non-adopters inside the same organisation is a measurement of this. Those super-users didn't emerge from a company-wide initiative. They emerged because something in their specific context - a supportive manager, a workflow that happened to be a good fit, a personal motivation to figure it out - gave them enough space to build genuine capability. The organisations that scale that aren't running bigger rollouts. They're systematically creating more of those specific contexts.
The Question Worth Asking Before the Next Strategy Deck
If 75% of executives know their AI strategy is performance art, the more useful question isn't how to build a better strategy. It's why the performance persists.
Part of the answer is that the pressure is real and the alternatives feel riskier. Moving more slowly on AI feels dangerous when every competitor announcement implies urgency. Admitting the transformation isn't working feels like a leadership failure. The performance continues because the cost of stopping it - even temporarily to redesign it - looks higher than the cost of continuing.
But the Writer data is clear about where this ends: 54% of C-suite leaders already say AI adoption is tearing their company apart. The sabotage is happening at 29% (and 44% among younger employees). The governance gaps are producing security incidents at 67% of organisations. The performance isn't holding. It's just not collapsing visibly enough yet to force the conversation.
The organisations that get ahead of this aren't the ones with the biggest AI budgets. They're the ones whose leadership is willing to ask - with genuine curiosity rather than defensiveness - what is actually changing at the operating layer, and whether the answer matches the presentation.
That conversation is harder than building a strategy deck. It's also the only one that produces the 29% outcome instead of the 71%.
