8 October 2026 · Breaking
Martin Lewis's 'The Small Ideas Initiative'. Three Small Fixes for the AI Gap.
Martin Lewis launched The Small Ideas Initiative today1. Until 12 November, anyone can send in a fix for one of life's glitches. Charities will check the ideas, a cross-party panel will pick about 50, and the initiative will push them at government, councils and regulators. The bar: practical, cheap, decent impact, and "hard to disagree with", the sort of thing 80% of people would back whatever their politics.
WE's usual argument fails that test, and fairly. Who owns the machines is a political question. But it has small versions, and they pass.
Here is the gap. Firms already use AI to set your price, approve your loan, read your claim and answer your call. You meet all of that on your own. Yet the same tools are now on everyone's phone. Claude, ChatGPT and Gemini can read a contract, compare tariffs or draft a complaint in a minute. What stops you using it back is small and fixable.
1. A reason you can read. When a machine says no, to a loan, a claim, a job application, the refusal should say why in plain English, and what would change the answer. The law already goes some way. Under UK GDPR2, a firm making a significant decision solely by machine must tell you about it, hear your side, give you a human and let you challenge it. In practice "information about the decision" often means a form letter. The fix: the reason and the thing that would change it go in the refusal itself, not on request. It costs a firm a sentence it already has.
2. Your bills in a form your AI can read. Open Banking lets you share your bank data safely with apps you choose. The same law that allows this for banks now allows it for other sectors3. The government's Smart Data strategy4 names energy and telecoms among ten priority sectors and aims for twenty schemes by 2035. Insurance isn't on the list. The fix: start with energy, phone, broadband and insurance renewals, and set a date sooner than 2035. Then anyone can ask their own assistant, every month, whether they're overpaying. Martin Lewis has spent two decades telling people to check. This lets the checking happen while you sleep.
3. Always a route to a person. If you're talking to a bot, it should say so, and there should always be a way through to a human who can change the outcome. Europe's AI Act now requires chatbots to say they're bots5. Britain has no general rule. The fix: a bot says what it is, and "speak to a person" works in every automated service a regulated firm runs.
None of this needs a new regulator or much money. Each fix can be tried, measured and dropped if it fails. And each one moves a little of AI's power from the firm's side of the counter to yours.