Table of Contents
For many large organisations, artificial intelligence remains stuck on the sidelines. Small innovation teams run pilots, experiment with tools, and showcase promising results—yet those efforts rarely make it into everyday work across the business. Citi has deliberately chosen a different route. Over the past two years, the bank has focused less on isolated experiments and more on embedding AI into how work actually gets done across the organisation.
Instead of limiting AI to specialists or innovation labs, Citi has worked to normalise its use across functions. The result is an internal AI-enabled workforce of roughly 4,000 employees, spanning technology, operations, risk, and customer support. This group did not emerge from a single mandate or centralised rollout, but from a people-first strategy designed to encourage participation rather than impose control.
Building participation, not bottlenecks
The scale of Citi’s AI adoption stands out. With around 182,000 employees worldwide, the bank now reports that more than 70% of its workforce uses firm-approved AI tools in some form. That level of adoption places Citi ahead of many peers that still restrict AI access to narrow technical groups.
A key reason is how the programme was designed. Rather than beginning with tools, Citi began with volunteers. Employees were invited to become “AI Champions” or join “AI Accelerators” programmes, gaining access to training, internal resources, and early versions of approved AI systems. These employees were not positioned as formal trainers. Instead, they acted as local guides within their teams—people colleagues could turn to with practical questions about when and how AI might help.
This peer-led model addressed one of the most common barriers to AI adoption: uncertainty. New tools often fail not because they are ineffective, but because employees are unsure how to use them in real workflows. By placing support inside teams, Citi narrowed the gap between experimentation and routine use.
Training that rewards visibility, not hierarchy
Training played a central role in sustaining momentum. Employees could earn internal badges by completing courses or demonstrating practical AI use in their own roles. These badges did not come with promotions or financial incentives, but they created visibility and credibility. They signalled who had hands-on experience, making it easier for others to seek help.
This approach proved more effective than traditional top-down rollouts. Instead of mandating adoption, Citi allowed AI use to spread through social proof and shared problem-solving. According to reporting by Business Insider, this grassroots dynamic helped AI gain traction faster than formal directives often do.
Everyday AI, with clear guardrails
Citi’s leadership has consistently framed the initiative around scale rather than novelty. With operations across retail banking, investment services, compliance, and customer support, even small efficiency gains can compound quickly. Employees are using AI to summarise documents, draft internal communications, analyse datasets, and assist with software development. None of these tasks are revolutionary on their own—but applying them consistently across thousands of employees makes a measurable difference.
At the same time, the bank has been cautious. Employees are limited to firm-approved tools, with clear rules around data usage and output handling. These guardrails may slow experimentation, but they also build trust. In highly regulated industries, confidence and compliance often matter more than speed.
What Citi’s model reveals about scaling AI
Citi’s experience highlights an important lesson for other enterprises: scaling AI does not require turning every employee into a data scientist. It requires enough people to understand the tools well enough to use them responsibly and explain them to others. By training thousands instead of dozens, Citi reduced dependence on a small group of specialists.
There is also a cultural shift embedded in the programme. Encouraging participation from non-technical roles sends a clear signal that AI is not reserved for engineers alone. It becomes part of everyday work—much like spreadsheets or presentation software became standard tools in earlier decades.
This approach aligns with broader industry trends. Research from firms such as McKinsey has repeatedly shown that many organisations struggle to move AI from pilots into production, often citing talent shortages or unclear ownership. Citi’s distributed model addresses both issues by spreading ownership across teams while keeping governance centralised.
Progress without the hype
The model is not without challenges. Peer-led adoption depends on sustained enthusiasm, and not every team moves at the same pace. Informal support networks can also become uneven, benefiting some groups more than others. Citi has tried to counter this by rotating Champions and refreshing training as tools evolve.
What ultimately sets the effort apart is how AI is framed. Citi has treated AI as infrastructure rather than a flashy innovation. Instead of asking whether AI could transform the business overnight, the focus has been on where it can quietly remove friction from existing work. That framing lowers expectations, makes progress easier to measure, and reduces pressure to deliver dramatic breakthroughs.
It also challenges the assumption that AI adoption must be driven entirely from the top. While senior leadership support was essential, much of the real momentum came from employees who volunteered their time to learn and share. In large organisations, that bottom-up energy often determines whether new technology sticks.
As more enterprises move from pilots to production, The behind-the-scenes effort powering Citi’s 4,000-employee AI rollout offers a practical case study. It shows that scaling AI is less about buying more tools and more about helping people feel confident using the ones already available—one team, one workflow, at a time.
