AI in Banking: Amplified not Automated

The bank we’re choosing to build in the age of AI.

By Roman Balzan, Chief Marketing and Brand Officer and AI Adoption Lead, Alpian

Last verified: July 2026

Six years ago I joined a bank, which was strange, because banks had never particularly appealed to me. Not because of the people inside them, but because of what the economics had made them. Advice that deserved the name was expensive to deliver, so it was priced for a threshold most people never reach. Everyone below it got a product, a fee schedule and a hotline. I joined Alpian anyway, as the first brand and marketing hire, because the idea was compelling and the people building it even more so, and because it came with the chance to help write a bank’s story from a blank page rather than inherit it.

That is why the question of what AI does to banking is not an abstract one for me. It decides whether this industry finally becomes more human, or finishes the job of removing the human entirely.

Because right now, the industry’s loudest AI story is a story about fewer people. Bloomberg Intelligence surveyed bank technology chiefs who expect as many as 200’000 banking jobs to disappear globally over the next three to five years as AI absorbs tasks. Whether or not every bank actually runs that race, it is the race the industry keeps describing. And I think it is the wrong one.

Let us start with the honest numbers, because they are dramatic. Citi’s research places banking at the top of the industries most exposed to AI-driven automation: 54% of banking jobs have high potential to be automated, more than any other industry, with a further 12% that AI can augment. The same research estimates AI could add USD 170 billion to global banking profits by 2028. Numbers like these are why every board in the industry is asking the same question, and why so much of the industry’s AI business case has centred on the same promise: fewer people, more machines, lower costs.

But there is a detail in that research almost nobody quotes. Past waves of technology did not shrink banking employment. They changed what bankers do. The ATM automated one of the teller’s central tasks, but it did not eliminate banking employment. It changed branch economics and gradually changed what branch employees were there to do. The pattern of a hundred years of banking technology is not replacement. It is reallocation. The question was never whether the machine takes over tasks. It always does. The question is what the humans do with the capacity that gets freed, and that is a choice each bank makes, not a law of nature.

Banking is uniquely exposed for a simple reason: a bank is, at its core, an information business. Money is data. Risk is data. Advice is pattern recognition plus judgment plus trust. AI transforms the pattern recognition, and it can genuinely support the judgment. What it cannot do is carry the responsibility. The entire strategic question of AI in banking sits in that gap.

Here in Switzerland, we now have real evidence of how banks are handling this, and it does not match the conference-stage story. The Digital Advantage study by Information Factory and finews.ch assessed 23 Swiss universal, regional, digital and private banks through executive interviews, structured mystery shopping and client surveys, and found a consistent gap between what the industry says about AI and what is actually true inside the buildings.

Banks say AI adoption is visionary. In reality it is pragmatic: embedded in internal efficiency, invisible in the client experience. Banks say culture is the bottleneck. In reality the bottleneck is accountability: the industry still lacks a settled answer about who responds when the machine gets it wrong in front of a client. Banks say data availability is the constraint. In reality most banks have the data; what they lack is shared agreement on what it means.

I find one of these findings genuinely important, and I want to defend the industry on it for a moment. The hesitation to put AI in front of clients is not cowardice or lack of ambition. The study calls the restraint rational, and it is: in the absence of accepted reference cases, when it is your money, “the algorithm decided” is not an answer. A suitability assessment, a retirement plan, a difficult conversation in a falling market: these carry accountability that has to land on a person. Any bank that deploys client-facing AI before it has answered the accountability question is not innovating. It is gambling with the only asset a bank truly has, which is trust.

The same research also overturns an assumption much of the automation case rests on. Banks treat the choice between digital and human channels as generational: the young want apps, the old want people. The evidence says it is situational. When the moment is routine, everyone wants the app. When the moment matters, everyone wants the person. Build your bank around the first belief and you are building for a client who does not exist.

So the industry is stuck in an uncomfortable position: the most automatable industry on earth, sitting on technology it largely uses for internal plumbing, unable to point it at clients until someone solves the human question. That is the real state of AI in banking in 2026, and almost nobody says it out loud.

Three things, in my view, and they are choices, not technologies.

First, every bank needs to answer the accountability question before the deployment question. Our answer is old-fashioned and we are keeping it: where judgment affects a client, responsibility lands with a person. A human signs. AI can prepare, draft, research, calculate and flag. The decision, and the answering for it, stays with someone who has a name. That is not a limitation of our AI strategy. That is our AI strategy.

Second, the boundary between machine work and human work has to be drawn deliberately, task by task, rather than discovered by accident after the layoffs. We have been going through the bank department by department, asking one question of every piece of work: does this need a human, or does it just need to be done? The paperwork, the preparation, the first drafts, the searching: the machine carries more of that every month. The judgment, the advice, the relationship, the moment a client is worried and the market is red: those are where we deliberately keep a person responsible. One rule holds the whole exercise together: automate the work, never the relationship.

Third, and this is the part the industry underestimates most. Accountability is the obstacle when AI crosses into a client decision. Inside the building the obstacle is culture, and AI adoption is a human practice, not a software rollout. The research is clear that the technology is rarely the problem. What decides success is whether people change how they work, and that is culture, training, rhythm and trust, built patiently from the inside. You cannot install that. You have to grow it.

There is a reason a bank our size can hold this position credibly. Alpian is around 70 people running everyday banking and wealth advisory together, and our ambition is unapologetic: work that a traditional operating model spreads across many hundreds. Not so we can be smaller. So that real advice stops being a luxury good. At Alpian, wealth advisory starts at CHF 2’000, not at the threshold traditional private banking treats as an entry ticket. The only way a bank this size offers that seriously is if every one of its people is amplified by the technology, and every client still gets a human.

Two things describe what that produces. Advice that starts at CHF 2’000 rather than at a private banking threshold, and advisors who hold a 4.98 out of 5 rating from the clients they serve. The first is the reach the technology makes possible. The second is what we chose to preserve while extending it. Neither is an AI result, and I would not present either as one. Both are human results, produced by a team small enough that the technology has to carry everything else. That is the argument in practice. The machine does not produce the outcome. It produces the time in which the outcome becomes possible.

The rule is not only something we say about ourselves. In their assessment, the study’s researchers singled out how Alpian removes the notion of “support” altogether, positioning client interactions as advisory conversations rather than problem resolution. That is what the rule looks like from the outside: the machine carries more of the process, so the moments that call for judgment can be genuinely human.

The robo-advisor pursued the wrong version of scale. It used technology to remove the person. We are using the same technology to give that person greater reach.

Working with AI every day for years has left me with one conviction, and it is one Alpian can sign its name under: the point of this technology was never less humanity. It is more. Not less AI, more I. The machine amplifies whatever is brought to it, the judgment, the standards, the attention. Bring nothing and it automates you. Bring everything and it amplifies you. That holds for a person, and it holds for a bank.

So when people ask what Alpian is doing with AI, the honest answer is not a list of tools. It is a choice. The future of this industry will not belong to whoever replaces the most humans. It will belong to whoever uses the technology to become the most human.

Six years ago I joined an industry that had never particularly appealed to me. This is the bank that changed my mind, and we are choosing to build it.

Sources: Citi GPS, “AI in Finance: Bot, Bank & Beyond” (2024); Bloomberg Intelligence survey of bank technology chiefs (2025); Information Factory x finews.ch, “Digital Advantage: NextGen Operating Models” (2026).