Every year the Office for Budget Responsibility publishes long-term projections showing that Britain’s debt path is unsustainable. And every year, politicians respond by doing nothing. With no looming crisis from our ageing population mixing with a pay-as-you-go welfare state, procrastination proves irresistible. Why volunteer as the government that reforms the NHS, cuts the future state pension, or raises taxes aggressively when you can leave that pain to your successors? Besides, perhaps a burst of growth will turn up that means none of this is required.

In Westminster, the can-kickers now have one comforting bedtime story in regards artificial intelligence. AI, they hope, is the magic asterisk that will swell tax receipts, turbocharge NHS productivity, and overpower the headwinds of population greying and low fertility. Labour has already touted a “£45 billion jackpot” from AI-driven efficiency in government alone. So why sweat difficult reform when you can pin your hopes on a tech injection? Call it Ozempic for the debt. A collective jab of agents and robots that will help curb the red ink.

To be sure, an AI-growth surge is now feasible and would make our difficulties more palatable. Yet building a fiscal policy around AI-fuelled growth seems premature. Even beyond the uncertainties of the impact and roll-out speed of this technology across the economy and public sector, there are simply too many known unknowns to blithely assume AI will show up as a tidy Treasury windfall.

Consider three different angles. Let’s imagine, for example, that AI boosts labour productivity substantially, raising wages for those in work, but increasing unemployment. Tax revenues from wage earners will rise, but welfare payments to the jobless would, too. Including, of course, the biggest benefit of all: the state pension. The triple-lock means that the pension rises each year by the highest of earnings growth, inflation, or 2.5 per cent. So in the very scenario ministers would like most —an earnings boom — a chunk of the extra money is automatically earmarked for higher pension spending on an already growing retired population.

Now consider the NHS and social care. In the best case, AI automates admin, speeds up diagnosis, and makes each pound buy more care. But there’s another plausible “success” story. Earlier detection and better treatment keeps us alive far longer. That’s fantastic for humanity, but not necessarily for the Treasury. Longer lives improve budgets only if they translate into more work years. If instead they mean more years drawing a pension and requiring ongoing, expensive care—as the Brookings Institution has warned in the US context—then future deficits might actually increase.

Then comes the politics. If AI replaces a lot of workers and provides strong returns to capital that are taxed more lightly than wages, the revenue boost will be muted. Yet demands for more redistribution through relief, retraining grants, or perhaps even universal basic income would grow. Indeed, this past century the demand for government spending has tended to exceed the economy’s growth rate, as politicians always find new things on which to spend money. If government itself, without a profit motive, adopts AI more slowly, these problems compound.

To be clear, none of this is destiny. One could envisage sunnier scenarios for the public balance sheet, whereby AI-fuelled growth shows up in broad taxable incomes, longer healthy working lives, an end to the cost disease afflicting healthcare, and future governments banking the dividends rather than spending them.

But that’s a lot of “ifs” to hang a fiscal policy on. My fear from noises emanating from Westminster optimists is that any impending AI advances are providing a further excuse to duck the unglamorous work of fixing pensions and facing up to the NHS’s inefficiencies. Perhaps that sort of gamble will pay off. If not, delaying the repair just means postponing the reckoning, with a harsher, faster and nastier fix to come.