AI Adoption in Banking: Why Culture Is the Real Barrier
CCO Managing Partner
Banks taught people that change means cuts. Then they wondered why nobody would touch the AI.
For most of my years in banking, culture was not a values poster. It was what people believed would happen to them when the next programme landed.
After 2008, the industry rebuilt itself around risk, compliance and documentation. That was necessary. It also trained a generation to treat every new tool as something that would add a control, not remove a burden. Digital programmes after that added another layer: faster channels, new cores, fewer branches. Internally, the story employees heard was often simpler. Efficiency. Headcount. Another target.
I left the industry in 2016. The conversations I have now with bankers have a familiar sound. Frustration about slow digitisation. Quiet frustration about the status quo. Almost nobody claims the bank already delivers one coherent client journey. And almost nobody claims the organisation trusts the next transformation slide.
That is the culture AI is walking into.
How Years of Banking Transformation Changed Employee Behaviour
Three forces rewired how people behave at work.
Regulation made caution a career skill. After the crisis, conduct, KYC and documentation became the safe answer. Relationship managers learned to live with two mandates at once: bring money in, and do not scare the client with the file. Swiss private banking still sells trust. The week of an RM is increasingly consumed by systems, reconciliations and approvals that create no franchise value.
Cost programmes taught a second lesson. Profits can rise while headcount falls. Banks have been explicit that technology, including AI, is part of the efficiency equation. Morgan Stanley estimates that more than 200,000 European banking jobs could be at risk by 2030 as AI and further digitalisation reshape the industry. People can read.
Product silos taught a third. Onboarding, credit and advisory kept their own records. The client met departments. Staff learned that “end to end” was a slide, not an owner. When you later ask those same people to build with agents, they do not necessarily bring you the real process. They protect the workaround that still gets month-end done.
McKinsey’s own banking leaders now make a similar point about agentic AI: the shift is mostly about people and culture, not the model. That is not a soft finding. It is an admission that the last fifteen years of change trained many organisations to survive programmes, not to co-design them.
What the AI and Empathy Data Is Actually Measuring
New workplace research puts numbers around the same split.
Employees who see their leaders as empathetic experience AI very differently. Nearly nine in ten employees in empathetic workplaces say they received adequate AI training. Across all employees, it is around half. In toxic cultures, only a third say the same.
The pattern continues with control. Eighty percent of employees in empathetic cultures say AI gives them more agency and control over their work, compared with just over half of employees overall.
At the same time, nearly three in ten CEOs say reducing headcount is a primary motivation behind their AI investment. Sixty percent of C-suite respondents say AI adoption creates a much greater need to demonstrate empathy.
That is not a request for nicer town halls. It is a prediction of where the work will hide.
If the implied contract is “we buy tools to do this with fewer of you,” capable people will keep the clever version in a private chat. Shadow AI is not only a technology problem. It is what a low-trust culture does with a powerful assistant.
Culture Amp’s 2026 AI benchmark shows the same leadership gap in another language. Eighty-five percent of employees say their organisation encourages them to explore and experiment with AI. Only 58 percent say leaders have clearly explained how AI will help the company reach its goals.
Encouragement without a destination is how you get pilots and no operating model.
Why AI Adoption Is Especially Difficult in Banking
Banking already had a risk-averse innovation culture before the models arrived. In a survey of 150 UK banking innovation leaders, 71 percent said risk-averse culture and red tape kill experimentation. Testing delays and legacy cores add another layer.
AI does not dissolve that culture. It inherits it.
The client side makes it harder. Trust is the product. If staff believe the next wave is primarily a headcount exercise, they will not risk a client conversation on an output they cannot explain. Advisors will not put a fluent, unsigned number in front of a family. Operations will not retire the Excel that still matches the close.
I saw the practical version recently in Singen. We can give teams agents that plan, build and test, with a human at the gate. None of that scales if the room thinks the project is a polite way to shrink the team. They will not show you the real process. They will keep it.
The Swiss private bank that still wins on judgment and network has a choice. Use AI to give the relationship manager the hour back, and say so. Or use AI as another efficiency programme and watch adoption stall in the hallway.
What a Culture for AI Adoption Actually Looks Like
The banks that get this right will not need another values framework. They will make three things explicit.
Which work stays human on purpose. Advice, exceptions, the signature. Not as poetry. As decision rights.
How people will be measured after the speed arrives. If the only metric is FTE reduction, the culture has already answered the empathy survey.
Who is allowed to build, and what the company still owns after someone hits enter. Citizen development without a gate is shadow IT. A gate without psychological safety is a gate nobody uses.
Governance and empathy are not opposites. Governance without empathy produces policies nobody believes. Empathy without governance produces confident answers nobody can reconstruct.
The Real Question for Banking Leaders
Banks spent years teaching people that change arrives as control and cost. AI now asks those same people to put their knowledge into a system the organisation can reuse.
The data on empathy is not an HR footnote to that story. It is part of the adoption model. Where leaders can explain what happens to the human after the tool works, AI can feel like more control over the work. Where they cannot, it looks like another disruption you survive by hiding.
The question I would put to a bank leadership team is not whether you have an AI programme.
It is whether a capable person on the floor would show you what they have already built.
If you want to understand what is helping or blocking AI adoption inside your own organisation, book a 30-minute exploration call with us. We can look at where trust, governance and ownership currently meet in your AI initiatives, and where the operating model may be getting in the way of the technology.