On 15 June the Government Office for Science published its updated set of AI scenarios for 2030. It is the most senior independent attempt the UK has made to describe how artificial intelligence could reshape work, the economy and security over the next five years. More than seventy experts fed into it, drawn from across Whitehall, the AI Security Institute, academia and the labs themselves.
The report does something unusual for a government document. It declines to tell you what will happen. It sketches five different worlds, ranging from a quiet “Slow Burn” where progress stalls, to a “Take-Off” in which systems outpace expert humans at nearly everything and slip beyond reliable control. And then, deliberately, it puts no odds on any of them. Estimating likelihood, the authors note, was outside the scope of the work.
That refusal is the interesting part. Most of us, when faced with uncertainty, want a forecast. We want to be told which way to lean so we can plan around it but these scenarios prompt us to resist doing that. They’re designed as a tool for policymakers to stress-test their thinking against, not a prediction to be proved right or wrong.
Which raises a fair question for anyone running an organisation rather than a government department. If even the people with the best available evidence will not say which future is most likely, what is a board supposed to do with five of them?
The point is not to pick the right one
Here is the trap to avoid. When we Read five vivid scenarios, the natural urge is to decide which feels most plausible, then prepare for that. But the real future will contain elements of several of these at once, and could slide from one into another. Betting the organisation on a single forecast is how you get caught out.
The more useful question is the opposite one. Not “which future is coming?” but “what holds true across all of them?” Because if something shows up whether progress stalls or accelerates, whether the labour market adapts gracefully or convulses, then it is no longer a bet but rather a certainty you can act on now.
What is true in every future
The report finds three things recur across all five worlds.
The first: the nature of cognitive work changes inevitably. Even in Slow Burn, the most muted scenario, where AI progress grinds to a near halt after 2028, the report still finds that routine execution gets automated and the human premium shifts to judgement, oversight and framing the problem in the first place. You do not have to believe in the dramatic futures for this to matter in your organisation. It is already happening, and it happens in every version of 2030 on offer.
The second: adoption is uneven, and the unevenness compounds. Across the scenarios, the report warns that differing speeds of adoption could let certain organisations, sectors and nations capture disproportionate gains while others fall behind. This is the gap I keep raising with clients. Not the gap between organisations that know about AI and those that do not, because by now everyone knows, but the gap between those that have built the capability to use it well and those still circulating articles about it. That gap does not stay constant. The organisations ahead get further ahead, because capability builds on capability.
The third: over-dependence on AI is a risk in its own right. Several scenarios describe workers who come to rely so completely on AI that they cannot function when it fails. This is not an argument against AI adoption but it reminds us that we need to deploy and govern it deliberately so that you keep capability rather than hollowing it out. Steering, in other words, rather than simply accelerating.
How can you use the five scenarios?
You don’t need a foresight unit or a workshop full of consultants to get value from them. You need a meeting and a willingness to take each world seriously for an hour, even the ones you instinctively dismiss.
Try holding your existing strategy up against each world in turn. Take your current AI plan, or your business plan if AI does not yet feature in it, and ask of each scenario: does this still work here? Where does it break? A strategy that only survives in the comfortable scenarios is a fragile one. The moves that hold up across all five are the ones worth starting now.
Work backwards from the future you want. Pick the 2030 that would be best for your organisation and your people. Then map the path back to today. What would you have had to start in 2026 for that world to be reachable? That is your near-term agenda, and it tends to look very different from the to-do list you would write looking only forwards from where you stand.
Rehearse the shock. The report is candid that the things we most confidently expect (that the current frontier labs stay ahead, that safeguards hold, that progress keeps climbing), are precisely the assumptions most worth poking. Pick one you are quietly relying on. Ask what you would do if it turned out to be wrong. The value is not the answer; it is discovering how exposed you are.
Notice what each of these does. None of them requires you to predict anything. They turn an uncertain future from something that happens to you into something you can interrogate, and that shift, from passenger to driver, is the entire point of governance.
The window is now, not later
The scenarios show very different futures on crucial points: how fast, how disruptive, who wins, who controls it? They agree on one thing with striking consistency. The favourable outcomes are not the default. The report is blunt that good trajectories are far from automatic and depend on the choices made in the near term.
That is true at the level of national policy, which is who the document was written for. It is just as true inside a single organisation. The difference between ending up in a version of 2030 you would have chosen and one that simply arrives is the work done in the eighteen months in front of us, while the future is still genuinely open.
You cannot choose which scenario the world delivers but you have a great deal of say over where your organisation sits within it. The leaders who understand that distinction, and act on it now while it still counts, will be the ones shaping what comes next rather than reacting to it.
The full report is worth an hour of any leadership team’s time. If you would like help running your own strategy through these five futures, that is exactly the kind of session we run.
Contact us at enquiries@aigovernance.co.uk
