Financial Times data journalists Joel Suss and Sarah O’Connor chart the uneven rollout of artificial intelligence across firms. Surveys diverge sharply: a Bank of England poll finds more than 70 % of UK companies using at least one AI tool, while the ONS reports only 35 %. In the United States, the Atlanta Fed cites 78 % versus the Census Bureau’s 18 %.

The gap narrows when analysts apply a stricter definition that limits AI to large‑language‑model generators such as ChatGPT or Claude. Under that lens, only 34% of UK firms claim adoption, according to the Bank of England, versus 16 % in the ONS data. Larger firms and ICT sectors lead, with Denmark at 42 % and Romania barely 5 % across 35 OECD economies.

Researchers attribute the lag to technical mismatch, costly system overhauls and regulatory uncertainty. Cambridge professor Diane Coyle warns that algorithmic pricing, digital‑twin employees and chatbot liability could trigger legal challenges. Cyber‑security fears and doubts over AI reliability further deter deployment, especially outside tech‑heavy industries. Without clearer rules and stronger management capability, AI adoption will remain well below the levels needed to lift productivity.