More than 60 percent of London-based small and medium enterprises have now integrated some form of AI tooling into their daily operations, according to figures published last month by the London Chamber of Commerce and Industry — a jump of 22 percentage points from the same survey conducted in January 2025. The speed of that adoption is striking. So are the problems piling up behind it.
The timing matters. Europe is caught between crises — extreme heat, war on its eastern edge, economic pressure — and businesses are under relentless pressure to cut costs and move faster. AI is the obvious lever. But London's tech community is increasingly candid that the tools being deployed across the capital's offices and warehouses carry risks that the sales decks never mentioned.
The Promise Meets the Shop Floor
At the Level39 tech accelerator in Canary Wharf, where roughly 200 fintech and AI companies are based, the mood this summer is one of cautious confidence edged with anxiety. Startups there are building AI-powered credit scoring, fraud detection and customer service automation. The commercial logic is hard to argue with: one member firm reported cutting its customer query resolution time from four hours to eleven minutes after deploying a large language model in January. But that same firm discovered in April that the model was systematically misclassifying queries from users with non-standard written English — disproportionately affecting customers in East London boroughs including Tower Hamlets and Newham.
That kind of bias is not a fringe concern. The Ada Lovelace Institute, based in Bloomsbury, published research in March 2026 finding that AI hiring tools used by UK employers showed statistically significant disparities in screening outcomes across ethnic groups. The institute documented cases in which candidates with names common in British South Asian communities received lower automated scores for identical CVs. Several of the firms using these tools were London-headquartered. None have been named publicly.
For smaller businesses the stakes are different but no less sharp. On Brick Lane in Bethnal Green, independent retailers who adopted AI-driven stock management software over the past 18 months are discovering that systems trained on national retail data perform poorly in a hyperlocal environment where demand patterns shift week to week based on market days, community events and tourism. One software package marketed to SMEs at £299 a month reportedly over-ordered perishable goods for three consecutive months before the owner switched it off. That is money that a small independent cannot absorb.
Regulation Is Lagging, and Everyone Knows It
The UK government's AI Opportunities Action Plan, published in January 2026, committed £14 billion in public and private investment toward AI infrastructure over the next five years. It said relatively little about enforcement. The AI Safety Institute, based in Victoria, retains a narrow remit focused on frontier models and catastrophic risk — it has no formal mandate to investigate the kind of mid-market algorithmic bias that is causing harm in London businesses right now.
The EU's AI Act, which came into full effect for high-risk systems in August 2025, gives Brussels-based firms clearer legal obligations than their London counterparts. Post-Brexit, UK companies are in a legal gap: expected to comply with the EU Act if they serve European customers, yet facing no equivalent domestic statute. Law firms on Fetter Lane have been quietly busy advising clients on dual compliance obligations they were not prepared for.
For London businesses looking to get this right, the practical advice from organisations like TechUK and the Digital Catapult — which runs an applied AI programme out of its King's Cross site — is consistent: audit your tools before you scale them, document your training data, and test outputs against the specific demographics of your actual customer base, not a generic UK population sample. The firms that are doing this work now will be far better positioned when Westminster eventually catches up with legislation. The firms that are not doing it are accumulating liability they cannot yet see on their balance sheets.