If AI’s Economic Promise Is to Manifest, We Must Fix the Data Dysfunction

Further to the theme of some of my recent missives there is continued, governments and economists are pinning their hopes on Artificial Intelligence (AI) to revive sluggish productivity, trim runaway public spending and give overworked public services a long-overdue upgrade. AI is being positioned as the economic superhero of our age, faster, cheaper, tireless. However, before we hand it the cape, we need to talk about its diet.

AI does not run on magic. It runs on data. Right now, much of that data is the equivalent of reheated leftovers from a mystery fridge, often badly labelled, sometimes expired, often not resembling anything like its original state (ultra-processed) and increasingly generated by AI itself. Yes, we are entering the era of AI feeding on AI-generated content, a kind of digital inbreeding that risks diluting accuracy, originality and trust. Call it the AI Risotto Problem (a dish that reuses what is available), once AI starts cooking with its own ingredients, flavour (and truth) can degrade rapidly. Compound this with the ‘Digital Dementia’ I coined back in 2017 – Digital Dementia born of Artificial Intelligence and things get interesting.

Then there is the corporate data legacy inject, as if the patient was not going a bit Frankenstein already! Despite decades of digital transformation slogans, most organisations still do not truly understand the data they hold, where it resides, who owns it or whether it is even correct. Classification is inconsistent, context is missing, inconsistencies are rife and governance is often reactive if at all. Without meaningful and disciplined data lifecycle management, AI risks becomes less an economic engine and more an expensive autocomplete machine.

Despite reports that 75% of businesses are not seeing ROI from AI yet, AI is understood to be delivering real gains in some areas. UK government trials report civil servants saving 26 minutes a day using AI tools; NHS pilots forecast hundreds of thousands of staff hours freed every month. International forecasts predict a 1% productivity lift over five years, helpful, if not heroic.

So what is the truth? I suspect AI will not save our economies unless we clean up its food supply. Better data lifecycle management, robust governance and clear classification are not optional, they are the fuel for economic impact if AI is to be that catalyst.

AI is still in an incubation phase in real terms. It is maybe the talented trainee, but feed it junk and like any adolescence it will hallucinate. Feed it context-rich, well-governed data and it really might help rescue the economy.

The question is not just what AI can do, it is whether we are ready to serve it something better than yesterday’s leftovers.