A cluster of surveys landed in the middle of this year, and read together they tell a story that no single one quite spells out. Enterprise AI is spreading fast, and confidence in it is going backwards. Kyndryl's second annual People Readiness Report, drawn from 1,100 senior leaders across eight countries, found that 57 per cent of organisations now have AI embedded in core processes or deployed broadly, up from just 35 per cent a year ago. Over the same year, the share of leaders who believe their workforce is fully prepared for AI fell to 23 per cent, down six points. Deployment is climbing; readiness is sliding under it.
The gap shows up most starkly in what all this activity actually delivers. Only 32 per cent of those same organisations say they have achieved even one of their top two AI objectives, and a mere 11 per cent have hit both. Nearly four in five leaders agree that the pace of AI development will simply outrun their governance, their operating models and their people. This is not a story about models being too weak. It is a story about companies buying capability faster than they can absorb it.
Down at the level of the actual employee, the numbers get bleaker. The Achievers Workforce Institute found that just 19 per cent of workers feel confident using AI tools and only 18 per cent feel supported in adapting to them. A separate UK study from TrustedTech warns of an emerging "AI underclass": 74 per cent of decision-makers say they feel confident using AI at work, but only 44 per cent of junior staff say the same. Roughly 38 per cent of employees describe themselves as self-taught, while fewer than a quarter have had any formal training from their employer. The tools are being handed out; the instructions are not.
Notion's data sharpens the same point into a hierarchy. It finds 88 per cent of workers sit at the bottom two rungs of AI readiness, treating the technology as a brainstorming aid or a glorified assistant, while only 12 per cent have reached the levels where AI acts as a genuine teammate or runs real workflows. Leaders and their staff are not even looking at the same reality: 60 per cent of decision-makers believe their organisation is ready for the next wave of autonomous, agentic AI, against just 36 per cent of the employees who would have to work with it.
That disagreement matters because the agents are coming regardless. In Kyndryl's survey, 81 per cent of organisations expect AI agents to make consequential business decisions within the next year, yet only 25 per cent say they fully trust such systems to operate without a human watching, and barely a quarter keep proper registries or monitoring across all their AI. Deploying software that acts on its own into a workforce that neither trusts it nor understands it is not a recipe for productivity. It is a recipe for quiet risk.
The more useful thread running through the research is that a small group is getting this right, and doing something specific. Kyndryl labels them Pacesetters, about 9 per cent of respondents, and they are roughly twice as likely to have fully implemented AI governance, 1.5 times more likely to report AI-driven revenue growth and 1.6 times more likely to see genuine product innovation. What sets them apart is unglamorous: they redesign roles around AI rather than bolting it on, they run structured change management, and they invest deliberately in training. A parallel study from Info-Tech found the same shape from a different angle, with firms that have a formal, governed AI strategy three times more likely to report measurable impact than those winging it. The lesson, arriving from several directions at once, is that the hard part of enterprise AI was never the model. It is everything, and everyone, around it.