Barclays expands Claude across its bank, targeting half of developers by end-2026

Anthropic has published a customer story saying Barclays is expanding its use of Claude across the bank. The post, dated 1 October 2026, covers software development, a staff knowledge assistant and the handling of client emails in Global Markets.[1]

It is a vendor's account, written with its customer. Every number below comes from that post. It gives no cost figure, no contract term and no independent measurement of results. Readers should treat it as a statement of intent and early usage rather than proof of savings.[1]

Barclays expects Claude Code, Anthropic's coding tool, to reach 50 per cent of its developer population by the end of 2026. It expects a majority of its software engineers to use it in 2027. These are targets, not achieved results, and the post does not say what share uses it today.[1]

Craig Bright, Barclays' Group Co-Chief Operating Officer, said skilled engineers, cyber specialists and trusted AI capabilities can improve how technology is built at scale. The post gives no productivity figure for coding, so the benefit remains a stated aim.[1]

The Colleague Knowledge Assistant has been live since 2025. Barclays says more than 16,000 colleagues have adopted it and it has handled over one million searches. It uses retrieval-augmented generation to help UK staff serve more than 20 million retail customers.[1]

In Global Markets, Claude models process about 120,000 emails each day. They classify, enrich and route client requests, and check that a request has the information operations teams need before operations teams act on it.[1]

Barclays says it applies robust governance, security controls and human oversight across its AI use cases. Paul Smith, Anthropic's Chief Commercial Officer, said Barclays is rolling Claude out with strong security and oversight. Both statements are assurances from the parties, not findings from an auditor or regulator.[1]

The post says nothing about error rates, incidents or how staff can challenge an output. Our reading is that the next useful disclosure would be measured: how often the assistant is wrong, and who catches it. Banks that publish such figures give regulators and customers more to check.[1]

The Barclays story is not about Malaysia, and the post names no Malaysian bank. Bank Negara Malaysia's August 2025 discussion paper shows the local picture. It reports that 71 per cent of banking institutions and development financial institutions had at least one AI application in 2024, up from 56 per cent a year earlier.[2]

The same paper says dependence on third-party providers was the biggest AI concern for banks, followed by cloud use. It is a non-binding discussion paper and may not become a policy document. Malaysian banks buying a model from a foreign vendor face exactly the oversight and vendor questions it raises, and the same applies to the finance teams of larger Malaysian firms.[2]