Citi makes AI training mandatory for 175,000 staff as Jane Fraser starts her next phase

You have probably watched a restructuring CEO declare victory, then discover that cost discipline alone does not win clients. Fraser, the first woman to run a major Wall Street bank and number one on Fortune's 2026 Most Powerful Women list, told the summit last week that Citi is past the heavy lifting. "If we are going to be the best in the businesses we're in, we have to beat the best," she said. That mindset of excellence, she explained, defines the next phase.

The vehicle is mandatory AI training for 175,000 employees, with a focus on prompt writing. Her rationale is three-fold: AI "can help you earn money. It can help you save money. It can help you secure the bank." Asked about jobs, she described "puts and takes" as some roles are created and others automated, and admitted the net headcount effect remains unknown. "Honestly, we don't know yet," she said.

What happened

Citi posted its best third-quarter revenue in a decade the day before Fraser appeared virtually at the Summit. She traced her approach to a mortgage-business restructuring she led earlier in her career: employees want to hear the hard truth, along with what the company will do to help them. "You do not have to be tough and unpleasant about it, but you do have to be straight," she said.

The AI training programme sits inside that philosophy. "I never think people perform well if they're fearful," Fraser said, and the training is pitched as prompt craft rather than a threat of replacement. One example already in production is an AI avatar agent for Citi's wealth business, developed with Google DeepMind's team, that talks with clients at any hour "when our bankers hopefully are getting some beauty sleep." Fraser added that it is not a replacement team member.

The technology also surfaces a trust problem. "We have to move fast and retain trust," she said, describing collaboration among big banks to find and patch vulnerabilities. When you innovate at speed, she argued, you must manage risk at the same speed.

What this means for L&D

Fraser's framing splits capability building from cost theatre. Mandatory training for 175,000 people signals that prompt literacy is now table stakes, not a specialist skill reserved for quants or product teams. If a global bank treats generative AI as a core competency for relationship managers and operations staff alike, learning functions in other sectors will struggle to justify pilot-only programmes that leave most employees untrained.

The wealth avatar offers a second lesson: production use cases clarify what training must cover. Once an agent is talking to clients at night, the people around it need to know how to supervise it and explain it. That requirement shapes curriculum in ways a generic "AI awareness" workshop cannot. L&D teams that start with a deployed tool and work backwards to the skills it demands will design faster, more relevant programmes than those that begin with vendor content libraries.

Fraser's honesty about unknown headcount effects also matters. Executives who promise that AI will only augment jobs lose credibility when the first automation wave arrives. Her puts-and-takes language gives employees a mental model: some roles grow, others shrink and the organisation will not pretend to know the final number today. Learning leaders can borrow that frame when they explain why upskilling is urgent even when the future org chart remains unclear.

The risk-at-speed principle applies directly to capability programmes. If you train people to use frontier models without teaching them to spot hallucinations, data leaks or bias, you have simply distributed liability faster. Citi's inter-bank collaboration on vulnerabilities suggests that risk training cannot be an afterthought or a compliance tick-box. It must move in lockstep with prompt craft, tool access and production deployment.

Finally, Fraser's mortgage-restructuring story offers a template for change communication. Employees tolerate hard decisions when leaders explain what is happening, why it is happening and what support the organisation will provide. L&D can apply the same structure to AI capability programmes: name the skills gap, explain the business case and describe the training, access and coaching that will help people close it. Straight talk without fear-mongering builds the trust Fraser says her bank must retain while it moves fast.

Sources:

• Fortune, 14 October 2026