AI Training for Banking: What Leading Institutions Are Doing

Citi just mandated AI training for 175,000 employees across 80 countries. JPMorgan made AI training part of onboarding for every new hire. Wells Fargo put 4,000 people through Stanford's AI program. The five largest banks have all launched formal AI upskilling programs.

This isn't experimentation. It's a signal that AI training in banking has shifted from "nice to have" to operational priority.

The Banking AI Skills Gap in Numbers

Banks are adopting AI tools faster than they're preparing their people to use them. The result is a growing gap between what organisations deploy and what employees can actually do with it.

The core issue, according to Caspian One's 2025 report: financial institutions are "hiring the wrong people" rather than lacking AI technology itself. Domain-specialised AI talent — people who understand both the tech and the regulatory environment — implement nearly 80% faster than generalists.

That's why the smartest banks aren't just hiring data scientists. They're training the people they already have.

What the Largest Banks Are Actually Doing

Here's what's happening inside the institutions leading this shift.

JPMorgan: AI Training as Default

JPMorgan has made AI training mandatory for all new employees, including prompt engineering and AI tool literacy. Their internal "AI Made Easy" program has reached tens of thousands of staff. The bank also rolled out its LLM Suite — an internal AI assistant — and onboarded 200,000 users in eight months.

With $18 billion in technology investment for 2025, AI training sits at the centre of JPMorgan's workforce strategy. President Daniel Pinto has said the impact will be most significant for the bank's 60,000 developers and 80,000 operations and call-centre staff — close to half the company.

Citi: Mandatory Prompting for Everyone

Citi's approach is the most aggressive. 175,000 employees — across 80 locations — must complete AI prompt training. The program uses adaptive learning technology that adjusts to each person's level. Experts finish in under 10 minutes. Beginners need about 30.

The reasoning is practical: Citi employees had already submitted 6.5 million prompts using internal AI tools. The training isn't about whether people should use AI. It's about the gap between "great prompting versus basic prompting to generate impactful results," as Citi's head of learning Peter Fox put it.

Wells Fargo: Deep AI Partnerships

Wells Fargo took a different route — sending 4,000 employees through Stanford's Human-Centered AI program. They also adopted Google's Agentspace platform and rolled out tools like Deep Research and NotebookLM to about 2,000 staff, with plans to expand.

The Regulatory Push: EU AI Act Makes Training Mandatory

It's not just competitive pressure driving bank AI training. Regulation is making it compulsory.

The EU AI Act — the world's first comprehensive AI law — classifies many banking AI systems as high-risk. Credit scoring, loan approval, fraud detection, and AML risk profiling all fall under strict requirements.

Since February 2025, organisations must ensure appropriate AI literacy among staff operating these systems. By August 2026, high-risk AI systems in finance must comply with full requirements around risk management, human oversight, transparency, and auditability.

For banks, this means:

• Role-specific training — a compliance officer needs different AI knowledge than a relationship manager
• Documented competence — you can't just run a workshop; you need to prove staff understanding
• Regular updates — AI literacy training must keep pace with both the technology and the regulation

Where Banks Need AI Skills Most

AI in banking isn't one use case — it's dozens. Different teams need different skills, and the training gap varies by function.

Fraud detection remains the primary use case. AI systems now process millions of transactions per second, with detection accuracy exceeding 90%. But fraud analysts need training on interpreting AI predictions, understanding confidence scores, and knowing when to override automated decisions. This is where machine learning literacy becomes a daily skill, not a one-off workshop.

Risk and compliance teams need to validate that AI models meet fairness requirements, understand algorithmic bias, and maintain audit trails. With 57% of Chief Risk Officers identifying talent shortages as a top long-term risk, the compliance function can't wait for perfect hires — it needs to upskill now.

Client-facing roles are changing too. PwC estimates a 50% productivity boost from human-AI collaboration in banking, with 30% increases in lead conversion through AI-driven personalisation. Relationship managers who understand prompt engineering and AI agents will outperform those who don't.

Operations staff — the 80,000 JPMorgan employees Daniel Pinto flagged — face the most immediate transformation. AI-driven onboarding already cuts client verification costs by 40%. Teams that understand how to work alongside AI agents, rather than being replaced by them, will keep their roles and expand them.

Why One-Off Workshops Don't Work in Banking

The banking sector has a specific version of a universal training problem: people forget what they learn.

Research shows employees lose up to 90% of new information within a week without reinforcement. Banks that run annual AI workshops — even good ones — are pouring knowledge into a leaky bucket. And in a regulated industry where competence must be documented, "we ran a workshop" doesn't prove ongoing capability.

Citi's adaptive approach is a step in the right direction. But even 30 minutes of training still follows the traditional model of a single learning event. What the research supports is something different: short daily practice with spaced repetition.

Six minutes a day, every day, builds the kind of durable knowledge that a full-day workshop can't match. It's why kju exists — to make AI fluency a daily habit, not an annual event. Our financial services learning path covers fraud detection, risk assessment, regulatory compliance, and client advisory — all contextualised for banking professionals.

The Bottom Line

Banking is further along in AI adoption than most industries. But adoption without training is like buying a trading floor full of Bloomberg terminals and not teaching anyone how to use them.

The banks getting it right — JPMorgan, Citi, Wells Fargo — share three things: mandatory participation, role-specific content, and continuous learning structures. They're not running workshops. They're building habits.

The regulatory pressure from the EU AI Act will only accelerate this. By August 2026, "we didn't train our people" won't just be a competitive disadvantage. It'll be a compliance violation.