Barclays is expanding Anthropic’s Claude across software development, operations, and customer support as the bank moves enterprise AI deeper into its global technology estate.
The bank expects Claude Code to reach 50% of its developer population by the end of 2026, with adoption increasing to a majority of software engineers during 2027.
Deployment extends beyond coding. Barclays is using Claude to support legacy-system modernisation, internal knowledge retrieval, operational processes, and client-service workflows inside an organisation subject to extensive security, resilience, and regulatory controls.
Anne Marie Darling, group co-chief operating officer at Barclays, said: “The true measure of any technology is the impact it has on customers, clients and colleagues, and the outcomes it enables.”
One of the more established deployments is the bank’s Colleague Knowledge Assistant, which has been live since 2025. More than 16,000 employees have adopted the tool and it has handled more than one million searches while supporting staff serving more than 20 million UK retail customers.
Claude is also being used within Global Markets to process and route client enquiries. Anthropic says the workflow handles around 120,000 emails a day, creating a high-volume operating environment in which response speed needs to be balanced against permissions, accuracy, and auditability.
Banks are increasingly moving beyond general-purpose AI chat tools and embedding models inside software engineering, customer service, compliance, operations, fraud prevention, and data workflows. That changes the governance question from whether employees can use AI to which systems the technology can access and what actions it can take.
Financial institutions hold large quantities of sensitive personal and commercial information while running technology estates built up over decades. New AI systems therefore have to work within established requirements covering resilience, cyber security, outsourcing, record keeping, data protection, and model risk.
Software development has emerged as one of the larger areas of adoption because output can be reviewed directly by engineers. Coding assistants can generate or explain code, support testing, help document older applications, and reduce some of the manual work involved in modernising legacy systems.
The economics become more significant when deployment reaches thousands of developers. Even modest reductions in development time can accumulate across a large engineering organisation, although any productivity gain has to be weighed against reviewing generated code, security testing, and the possibility of errors entering production.
Barclays is also exploring more agentic technology, where systems can perform defined actions rather than only produce responses. The operational controls required increase as software becomes more autonomous, particularly when workflows touch customer information or regulated processes.
Permissions, escalation routes, human approval, monitoring, and the ability to reverse actions become part of the technology architecture rather than separate governance paperwork.
The deployment also reflects the broader problem of modernising bank infrastructure without replacing every underlying system. Mainframes, older applications, and complex databases remain embedded in many large financial institutions because they support high-volume processes that cannot easily be interrupted.
AI-assisted engineering can reduce some of the work involved in understanding and changing those estates, but it does not remove the need for sound architecture, accurate data, testing, and controlled migration.
Barclays’ year-end adoption target puts the programme beyond a limited proof of concept. The next phase will be judged on whether deployment produces sustained improvements in development and operations while preserving the reliability and governance expected of a major bank.
You must be logged in to post a comment.