Anthropic announced on 1 October 2026 that Barclays is extending Claude across more of the bank. Claude Code should reach half of its developer population by the end of 2026 and a majority of software engineers in 2027. It is among the largest disclosed rollouts of an agentic coding tool inside a systemically important bank.
Key takeaways
- Claude Code to reach 50% of Barclays developers by end of 2026, a majority of engineers in 2027
- Colleague Knowledge Assistant live since 2025, used by more than 16,000 staff
- Over one million searches handled, supporting staff who serve 20 million-plus UK retail customers
- In Global Markets the platform classifies and routes roughly 120,000 emails a day
- Remaining areas: legacy platform modernisation and cyber security
Three deployments, three different scales
| Area | What it does | Scale |
|---|---|---|
| Colleague Knowledge Assistant | answers questions from the bank's internal documents | 16,000 staff, over one million searches |
| Global Markets | classifies, enriches and routes inbound client mail | ≈ 120,000 emails a day |
| Software development | Claude Code plus legacy platform modernisation | 50% of developers by end of 2026, most engineers in 2027 |
| Cyber security | support for security teams | scale not disclosed |
A knowledge assistant instead of a search box
The longest-running piece is the Colleague Knowledge Assistant, live since 2025 and built on a retrieval-augmented generation architecture. Staff ask in plain language and the system answers from internal documents. It has handled more than one million searches for over 16,000 people serving more than 20 million UK retail customers.
Email that sorts itself
The second deployment handles inbound client mail in Global Markets. Barclays reports less manual handling and faster identification of priority requests.
Code and cyber security
The third area covers software development, legacy modernisation and cyber security.
Software engineering and cyber security are both being reshaped by increasingly capable AI systems. We're moving towards AI as an increasingly agentic capability embedded within how we build, test, secure, and operate technology.
Craig Bright, Group Co-Chief Operating Officer at Barclays.
The rollout fits Anthropic's wider push into financial services. In September 2026 it announced Enterprise Frontier Safeguards, a zero data retention?Zero data retention: A mode where the model vendor keeps no copy of conversation content — data stays in the customer's own infrastructure, under the customer's keys. mode where conversations land in the customer's own cloud rather than the vendor's servers. It was built with more than 100 customers and every US global systemically important bank?Systemically important bank: An institution whose failure would destabilise the whole financial system. Such banks face stricter capital and supervisory requirements..
Why it matters
The bank is disclosing a timetable with numbers rather than a pilot, which is rare in a sector that usually discusses AI in generalities. Committing to half the developer base within months moves agentic tooling out of the experiment column and into infrastructure. The risk sits with code oversight, because the more a machine writes, the heavier the load on review and testing. Other banks now have a reference point.
What's next?
- The checkpoint is the end of 2026, when 50% of developers should have Claude Code access
- Enterprise Frontier Safeguards is due to begin rolling out in autumn 2026 according to Anthropic, setting data residency terms for financial-sector customers
