The impact of artificial intelligence on the economy takes many forms. A technology leader at one of the UK’s largest supermarkets, for example, tells the story of his company’s first experiments in generative AI (GenAI).
Like others, he began by dabbling in the daily drudgery, those laborious tasks ripe for automation. Could his company, he wondered, use an AI assistant to relieve someone of taking minutes at the weekly management meeting. This is a process – from transcription to formulation to distribution –that can be handed over to the machines. That’s the promise, at least.
Unfortunately, the first time the technology was used at the supermarket, the results were more than a little disappointing. The notes generated were disjointed, unfocused and unusable. The question the technologist asked next was crucial to his understanding of the possibility of AI. Were the underwhelming minutes a product of poor technology or did they reflect poor meeting technique? He had the self-awareness – and humility – to conclude that it was probably the latter.
The company sharpened up its meeting processes, making the gatherings shorter and more focused on outcomes and actions. The result? The minutes were shorter and more focused on outcomes and actions. The AI proved its value not only in speeding up an existing process but by improving the process. Not just faster, but better.
Artificial intelligence is exposing other areas of inefficiency that need addressing, too. Not least the quality of data that most firms possess. If AI experimentation forces organisations – not just in the private sector but in the public sector, too – to get their houses in order, then it will have served a valuable purpose even before the technology is deployed in earnest.
The tale of the supermarket minutes represents a small win within a small part of a single company. On its own, it is not going to transform productivity or kick-start growth, as the government hopes. But small wins have two big advantages. First, they add up to something larger, in aggregate, across the economy. Second, they demonstrate the potential of new technology. They challenge scepticism, validate further experimentation and adoption, and open the way to future funding.
When the Tony Blair Institute makes projections about annual £40bn productivity gains in the public sector, it is not thinking about better note-taking. Not in isolation, at least. It is talking about the fact that the NHS will use AI to screen 700,000 women for breast cancer later this year, and similarly significant projects. Nevertheless, the small gains matter too.
AI is not a productivity panacea, however. As we have discussed on these pages before, not all processes are easily transferrable from analogue to digital, or from pre-AI to AI. Moreover, sometimes the “friction” – to borrow a word from Imogen Parker of the Ada Lovelace Institute – that more manual processes impose on us elicits important discoveries, some light-bulb moments. Hard and laborious work can be an invaluable component of creativity that, in turn, leads to productivity and growth.
When the government launched its AI Opportunities Action Plan – written by tech entrepreneur Matt Clifford – in January, its response contained none of these caveats, no notes of caution. Instead, Keir Starmer described AI as the “defining opportunity of our generation”. The Prime Minister went on to insist that “in the coming years, there is barely an aspect of our society that will remain untouched by this force of change”. Perhaps, but will it deliver greater efficiency, transform public productivity, and boost future growth in every instance?
This is a boosterist take on the potential of AI and it needs to be met with critical inquiry. Projections are one thing but where is the evidence that AI will deliver the productivity dividend the likes of the Tony Blair Institute forecast? The technology industry wants us to believe in the transformative power of the products it creates and sells. It wants us to believe in the next big thing. But even when technology delivers a positive impact, progress is rarely linear. The analyst group Gartner famously talks of a hype cycle where a “trough of disillusionment” inevitably follows the initial excitement around a new technology – and is a precursor to a “plateau of productivity”. In other words, where technology succeeds it rarely does so first time.
Artificial intelligence is a tool like any other technology and its utility, naturally enough, comes down to how it is used. From cancer screening scans to better meeting minutes, application is everything. We should approach Starmer’s “defining opportunity” with guarded interest. Scepticism is a useful position when assessing AI’s efficacy and potential downsides. Useful until it atrophies into cynicism.
This article first appeared in our Spotlight Igniting Growth supplement of 14 March 2025.
Content from our partners
