Every few weeks, a new survey lands claiming AI is either about to replace half the workforce or barely making a dent. Most of them are vendor-funded, thinly sampled, or several months out of date by the time they get quoted in a boardroom.

The New York Fed's regional business survey is neither. Published this month by Liberty Street Economics, it draws on the Federal Reserve Bank of New York's own district businesses and asks a straightforward question: are you using AI, and has it cost anyone their job.

The adoption numbers are striking. More than 60% of service firms in the district now use AI, up from 40% a year ago. Manufacturing has nearly doubled too, from 26% to around half. Whatever plateau AI adoption was supposed to hit by now, it hasn't.

The layoffs numbers are, by contrast, almost boring. Just 4% of service firms report laying anyone off because of AI in the past six months, up from 1% the year before. No manufacturers report any AI-related layoffs at all, in either year. Adoption is accelerating. Job losses attributed to it are not.

That gap between the headline fear and the ground-level reality is worth sitting with. It matches what other recent data has shown: Gusto's small business figures found AI-adopting firms growing headcount faster than non-adopters, not slower. For most ordinary businesses, the jobs story and the AI story are running in parallel rather than colliding.

The number underneath the number

The most useful figure in the Fed's survey, though, is not the layoffs rate. It is this: among businesses that count themselves as AI adopters, the median share of staff actually using it is 17% for service firms and just 7% for manufacturers.

Read that again. A business can tell a Fed surveyor "yes, we use AI" while five out of six of its people have never touched it.

This is the pattern the AI Accelerator Diagnostic is built to find. Vision and Opportunities are usually fine, most leadership teams can point to where AI could help. People and Culture is where the gap actually lives: who has the permission, training and confidence to use it, and who is quietly locked out.

The Turner Agency, a 36-person global events and film agency, is a working example of what closing that gap looks like in practice. The diagnostic found genuine appetite across the business rather than confined to one team, so the follow-on work was a six-session AI Accelerator training programme built around the whole organisation's actual work, not a generic course. Across 131 survey responses collected over the programme, the average satisfaction score was 4.6 out of 5, and not one response was negative.

That is not a coincidence of a well-run agency. It is what tends to happen when adoption is measured properly, planned for the whole team, and treated as a capability project rather than a tooling decision.

What this means for the next twelve months

The Fed's data gives agency and SME leaders two honest things to work with. First, the fear that AI is here to cut jobs is not, for now, matching the evidence: adoption is up, layoffs attributed to it are not. That is worth saying plainly to a workforce that is anxious about it, because the anxiety itself has a cost.

Second, and more usefully, the real opportunity sitting in most businesses right now is not deciding whether to adopt AI. Most already have, on paper. It is closing the gap between the one person or team who has actually adopted it and the rest of the organisation who has not.

That is a People and Culture problem, not a technology one, and it responds to planning rather than urgency.

Go wisely.