Anthropic
Blogcast

Alex: A British banking giant just bet big on A I. Barclays is expanding its strategic collaboration with Anthropic, and this is not a pilot project or a press release demo. They are rolling Claude out across the bank, to the people who build the software, to the people who support customers, and to the people who move markets.

Jordan: This one caught my eye because banks are famously cautious. A universal bank with more than twenty million retail customers in the U K does not hand A I the keys lightly. So what does scaling actually look like here?

Alex: Three fronts. First, software development. Barclays says Claude Code adoption will reach fifty percent of its developer population by the end of twenty twenty-six, and a majority of software engineers by twenty twenty-seven. That is an explicit, dated, on-the-record commitment. For a bank running decades of legacy systems, that is a statement.

Jordan: And the framing from their leadership is interesting. Craig Bright, their Group Co-Chief Operating Officer, said software engineering and cybersecurity are both being reshaped by increasingly capable A I systems. They are moving toward A I as an increasingly agentic capability embedded in how they build, test, secure, and operate technology. The key word there is embedded, not bolted on.

Alex: Right. The stated goals for engineering are modernizing legacy platforms, improving software quality, and letting technical experts focus on the most complex challenges. That matches a pattern we keep seeing across these stories: A I takes the routine load, humans take the hard problems. But banks measure that differently than startups do. When Barclays says it, they mean at a scale of tens of thousands of engineers.

Jordan: Second front, and this one is already live. The Colleague Knowledge Assistant. This has been running since twenty twenty-five, powered by Claude through a retrieval augmented generation architecture, which is a fancy way of saying it searches Barclays' internal knowledge and drafts answers grounded in it.

Alex: And the numbers are real. More than sixteen thousand colleagues have adopted it, and it has handled over one million searches. It exists to help U K employees find information and answers more quickly when supporting those twenty million-plus retail customers. Faster answers for staff means quicker support for customers. Simple, measurable, and it shipped last year.

Jordan: I love that detail, by the way. One million searches is not a pilot. That is a tool people chose to keep using. Adoption is the metric that actually tells you whether the thing works.

Alex: Third front, Global Markets. This is where it gets operationally nerdy, in a good way. Teams are using Claude models to classify, enrich, and determine the optimal processing route for incoming client emails. The platform processes approximately one hundred twenty thousand emails each day.

Jordan: One hundred twenty thousand emails a day. Just sit with that for a second. If you have ever watched operations teams in a bank triage an inbox, you know how much manual handling that is. Claude sorts them, enriches them with context, and figures out where each one needs to go, so the humans spend their time on the requests that actually need judgment.

Alex: And that feeds the efficiency story the whole announcement is built on. Anne Marie Darling, their other Group Co-Chief Operating Officer, had a line I want to read closely. She said the true measure of any technology is the impact it has on customers, clients, and colleagues, and the outcomes it enables. Then the key part: expanding the use of Claude across Barclays is about more than adopting A I. It is about transforming how work gets done.

Jordan: That is the difference between a deployment story and a transformation story. Adopting A I means you bought licenses. Transforming how work gets done means you reimagined processes. She names it explicitly: reimagining processes, enhancing customer and client experiences, and building a simpler, more efficient organization, all within a secure and well-governed environment.

Alex: That last clause matters a lot for a bank. The announcement is clear that responsible deployment stays central. Barclays applies robust governance, security controls, and human oversight to A I use cases, making sure the technology is deployed safely and effectively. For anyone listening who builds in regulated industries, that is the permission structure that made this possible.

Jordan: Which brings us to the Anthropic side of the story. Paul Smith, their Chief Commercial Officer, called this a significant milestone for Anthropic in the U K. And he gave the one-line summary that I think will stick: Claude now helps sixteen thousand Barclays colleagues find answers for customers, sorts one hundred twenty thousand emails a day, and will be in the hands of most Barclays engineers by twenty twenty-seven.

Alex: Very few institutions hold A I to such a high standard, he said. And that is worth pausing on, because the standard is the story. Banks do not have the luxury of moving fast and breaking things. When a bank this size scales A I this publicly, it is a signal to every regulated enterprise watching: the governance playbook exists, and it works.

Jordan: Stepping back, what is the broader read here? I see a template forming. Knowledge assistant for the front line, email triage for operations, Claude Code for engineering. That is a complete stack: support, operations, build. And each layer has its own success metric. Searches handled. Emails routed. Developer adoption percentage.

Alex: It also rhymes with Anthropic's broader enterprise push. They have been rolling out Claude to financial advisers, partnering with Accenture on embedded evaluation, and talking about enterprise frontier safeguards. Barclays is the proof point at the other end of that pipeline. The enterprise motion is not theoretical anymore. It has a bank, a timeline, and one hundred twenty thousand emails a day.

Jordan: So if you are building with Claude and wondering what serious enterprise adoption looks like, this is your reference case. Start with something live and measurable, like the knowledge assistant was in twenty twenty-five. Earn trust with real usage numbers. Then scale to the harder stuff, like engineering and markets, under strong governance.

Alex: That is the Barclays story. One more episode proving that the interesting A I news is no longer the model launch. It is what happens when a careful, regulated, two-century-old institution decides to change how its people work. We will keep following it.