• Key insight: McNamara broke down three types of agentic deployment she sees in the financial industry — internal productivity, faster workflows and net-new products — and argued the real ROI gap lies in the second and third areas. 
• Expert quote: "The biggest mantra I want you to take back: stop prompting, start delegating," Anthropic Head of Banking Katie McNamara said, adding that the shift in mindset is what separates banks that are merely experimenting from those actually accelerating.
• Forward look: McNamara cautioned against centralizing AI development in a single innovation team, instead urging banks to decentralize experimentation to the employees closest to the work while keeping governance, safety and security controlled.

Anthropic's Head of Banking Katie McNamara urged banks to adopt agentic AI on Wednesday. In a June 17 presentation at the 2026 New York Banking Summit — which was cohosted by Fitch Ratings and KPMG — McNamara argued that banks must move toward an agentic workforce operating alongside humans. She outlined where Anthropic has already observed that shift taking hold, and offered practical advice on how financial organizations should govern and lead through the technological transition.

According to McNamara, AI agents are currently able to run hours-long tasks autonomously without any human intervention. The crucial question for banks — as a heavily regulated industry — is how to prepare for the rise of an autonomous agentic workforce, she continued. 

"Where the math says we're going next is an entirely AI workforce alongside humans, so humans and AI working together," McNamara said. "We're already starting to see this."

At an early panel during the conference, Daragh Murphy, founder and CEO of Imprint — a fintech that builds and manages co-branded credit cards for major consumer brands — said his team has sought out opportunities where instead of hiring "15 humans and creating a back office somewhere," they can instead build the AI system. Murphy described the technology as an "equalizer" as his company grows, with aspirations to obtain a banking charter and go public. 

"You end up running into the same problem all the companies have, which is you just have massive back offices. It's really hard to move quickly when you end up building a bureaucracy inside of your organization," Murphy said. "The way AI can help is, you start from first principles now, and … that's an opportunity for you to rebuild the architecture."

McNamara listed three areas where she sees financial service companies utilizing agentic services to conduct "targeted experimentation": employee productivity, faster processes and transformative products. 

While digital native businesses are more familiar with AI-integrated products, McNamara said that traditional financial services are also working on them. She added that increasing the speed of work processes both in-house and with consumer-facing products is where Anthropic sees "tremendous ROI" for banking institutions. 

McNamara highlighted the productivity benefits of delegating the "vast majority" of coding to agents — adding that 82% of Anthropic's own code that ships to production has been written by Claude. While AI is currently at the level of a junior analyst for financial firms, she posited that — "if the exponential holds true" — the technology will soon be able to operate at the senior analyst or VP level. 

Given the rapid development of AI, McNamara said the challenge facing the finance industry is the pace and ability to "absorb the models and the changes."

"In order for us to be able to close this gap, we need to pick the really powerful workflows that are going to drive meaningful change and meaningful ROI within an organization," she said.

She advised against centralizing AI development into "one central hub," and instead encouraged banks to foster a "culture of experimentation" where workflow innovation can be driven by the "people actually doing the work." What should remain centralized, according to McNamara, is governance, safety and security mechanisms. 

McNamara told C-suite executives to "drive the change from the top" by showcasing their personal uses of AI to their teams. She added the importance of open conversations about "what the new workforce is going to look like," and hiring for adaptability to evolving roles. 

"The biggest mantra I want you to take back: stop prompting, start delegating," McNamara concluded. "When we start to change the unit of work from 'Did it give me a good answer?' to, 'Did it get the job done?' That's when you know you've accelerated."

Anthropic has been entangled in an escalating battle with the federal government the past week over the public release of its Fable 5 model. Three days after its launch, the administration issued Anthropic an export control directive ordering it to suspend all access to its newly released Fable 5 and Mythos 5 models for any foreign national — including noncitizen Anthropic employees — whether located inside or outside the U.S. 

Senior Anthropic staffers met with Trump administration officials in Washington on Monday to try to resolve the dispute, and the company sent technical staff to Washington over the weekend to help work through the issue. The negotiations coincided with the G7 summit this week, where President Donald Trump and other world leaders were joined by AI company executives including Anthropic founder and CEO Dario Amodei.