By Emma Weber
Nearly every organisation I speak with now has an AI strategy. Very few have a people strategy inside it.
Strategies typically cover - tool selection, use cases, governance, a roadmap with quarters on it. And then near the back, a line about training and a line about communications, and that's the people part handled.
But look at what we're actually asking of our organisations. We're asking people to change how they work, and not slightly. To hand over parts of a job they've spent years getting good at, to a system they didn't choose and mostly don't understand, on a timeline someone else set. That's a behaviour change ask, and a big one. And after 23 years in behaviour change it's a red flag! You don't get a shift of that size from a comms plan and a lunch-and-learn. You get it when people understand why it matters and what it means for them.
BCG's fourth annual AI at Work study, published in June, surveyed nearly 12,000 workers across more than a dozen markets.
Points to note - only 33% say their leaders communicate clearly about AI. And only 28% see any alignment between what leaders say about AI and what the organisation actually does. So around seven in ten people are watching a gap open up between the message and the behaviour, and are then faced with a decision as to which of the two to believe.
And of more concern - 66% get limited or no guidance on what to do with the time AI gives back to them. We've handed people hours and said nothing about what those hours are for. If you're on the receiving end of that silence, you'll fill in the blank yourself, and the answer you land on probably isn't a generous one.
This is where a lot of the anti-AI feeling inside organisations is coming from, and it is rising. Writer and Workplace Intelligence surveyed 2,400 knowledge workers across the US, UK and Europe (including 1,200 C-Suite executives) in April.
Fortune reported on the survey in April and July 2026. To get into the detail, 29% admitted to actively undermining their own organisation's AI strategy: refusing the mandated tools, faking usage, doing the work by hand and saying otherwise. Among Gen Z it was 44%. Around a third named fear of losing their job as the reason. Now, let's note that it's a vendor-commissioned survey, and "sabotage" was defined broadly to include passive non-use. Regardless, the stats are concerning.
Clearly, that isn't a training problem.
People operating from fear don't do their best thinking. They protect, they hedge, and they stop volunteering the awkward observation that would have saved you six months. And the version that looks healthier on a dashboard is often worse: people complying because they feel they have no say. You get the usage figures. Logins, prompts, a nice adoption curve.
What you don't get is the judgement, the curiosity and the human effort that make an AI transformation work.
Which brings me to mavericks. On the McKinsey Talks Talent podcast recently, one of their talent experts put a question to CEOs and CHROs that intrigued me: do you have a good sense of who your mavericks are, and are you strategically allocating them to drive this transformation? A maverick, in their framing, is someone bold enough to challenge and not defensive about their own patch. Erik Brynjolfsson, from Stanford, in the same conversation, replied "I like that, this is a time for mavericks."
I agree with both of them – they are the type of people that will make a difference. What I notice is that the same organisations saying they want mavericks have performance systems, approval chains and risk registers built to smooth exactly those people out. Management is mostly selected to protect what already works, which is usually right and isn't what gets you through this. So the person experimenting with something unapproved is either your most valuable early adopter or a governance breach, and which one depends almost entirely on whether anybody told them what good looks like.
So what does a people strategy for AI actually contain? Five questions, and they're harder than they look. What does this mean for our people, said plainly, including the parts that are uncomfortable to say. How does AI serve our purpose, so the reason for all this connects to something people already care about. What will we use AI for. What won't we use AI for. And how are the people who deliver for this organisation at the heart of the change rather than downstream of it.
That fourth one deserves real consideration. Naming what you won't automate is a signal that takes fear out of the room, and it's often missed. Silence gets read as everything being on the table.
There's a business case underneath this, not only a human one. David Schatsky's June paper for the Harvard Kennedy School's Belfer Center cites Deloitte finding that 40% of companies achieved cost savings from AI while only 20% achieved revenue improvement. The organisations aiming at growth rather than headcount reduction are the ones seeing skills deepen and roles become more valuable.
We keep treating the people part as the soft bit that follows the real work. It was always the real work. Brynjolfsson and his colleagues found this pattern with earlier general purpose technologies: measured productivity dips before it climbs, because the complementary investment - new processes, new roles, new skills - takes years to build and shows up as cost long before it shows up as output.
Which brings me back to what sits underneath all of it. When we ask someone to work alongside a system that can do parts of their job, we're asking them a question about what they are for. That deserves a better answer than a training module and a change curve on a slide. It goes to the heart of what it means to be human at work, and there's no shortcut through it.
If you're writing an AI strategy at the moment, give those five questions more time than the tool list. The technology will keep arriving whether we are ready or not. Whether our people arrive with it is the part we actually decide.
About the Author: Emma Weber spent 23 years building Lever - Transfer of Learning, the global behaviour change business she sold in 2025, and created Coach M, one of the first AI coaching chatbots, back in 2017. She now works with organisations on the human side of AI transformation through Being Human in the Age of AI, and is the author or co-author of three books including Turning Learning into Action.