Customers do not want a digital friend. They want an AI assistant they can trust

Giving an AI assistant a name, a face and human characteristics may seem like a logical way to make AI more approachable in financial services. But the fact that people trust human advisers does not mean they will automatically trust an AI that looks and behaves like one.

The research points to something else. Humanising AI assistants does not increase banking customers’ trust or their intention to use them. Some elements can even have the opposite effect. The more an AI assistant tries to appear human, the more apparent its artificial nature becomes. Customers are not looking for a digital friend. They mainly want to know what to expect and retain control themselves.

Control and security matter most

This is what emerges from a quantitative survey of 1,053 Dutch banking customers across generations, including customers of ING, Rabobank, ABN AMRO, ASN, Knab, Revolut and Bunq. We also conducted in-depth qualitative interviews with retail banking customers from Generation Z, who tested two AI prototypes: one warm and adviser-like, and the other neutral and task-oriented.

Respondents could select up to three conditions that would make them more willing to use an AI assistant from their bank. The ability to switch to a human employee at any time came out on top, with 52% selecting it. This was followed by a guarantee that their data would not be shared with third parties (47%), oversight by an independent regulator such as the Dutch Authority for the Financial Markets (AFM) or Dutch Central Bank (DNB) (36%), and a clear explanation of why the assistant makes a particular recommendation (31%).

At the bottom of the list is the very factor that receives considerable attention in AI design: human characteristics such as friendliness and empathy. Only 3% of respondents say these characteristics would make them more likely to use an AI assistant. Among customers aged 61 to 79, this figure was just 1.8%.

The priorities are therefore clear. Customers do not build trust because an AI behaves like a human. They build trust when the bank makes control and security visible and when they can easily switch to a human employee.

Friendly AI is still AI

The difference in design had little effect on how participants perceived the assistant. Even participants who preferred the warmer design continued to see the AI assistant primarily as a tool.

One participant put it this way: ‘I wouldn’t see her fully as an adviser, because I don’t trust it enough for that.’ Another said: ‘It’s still a tool, because you know it’s an agent.’

Customers are therefore not rejecting a warmer design. But a more human appearance does not increase their willingness to use the AI assistant.

Trust is already embedded in the bank

So where does trust come from? The interviews provide an important clue: customers bring their existing trust in the bank with them when they interact with AI.

One participant said: ‘I trust it as long as it is an institution I can trust... I transfer the trust I have in the banking sector to that artificial intelligence.’

Banks do not need to create a new, human personality for their AI. They can build on the trust customers already place in the institution. The condition is that it remains clear that the AI assistant is part of the bank and operates under its responsibility.

Not every decision is suitable for AI

There is also a limit to what customers are willing to delegate to an AI assistant. That limit is largely determined by the impact of the decision.

For routine tasks, such as changing an address or checking spending patterns, AI is an obvious fit. For major financial decisions, such as taking out a mortgage or loan, the need for human interaction remains strong.

This is consistent with earlier PwC research on mortgage advice. It found that AI-only advice scores around ten percentage points lower on customer satisfaction than advice provided by a human adviser or a hybrid model in which humans and AI work together.

Customers are not asking for more human AI. They are asking for AI that knows when technology is enough and when a human is needed. For banks, that is the most important design choice.

What this means for banks investing in AI

The research gives banks a clear direction. The personality of the AI assistant should not be the starting point. Instead, banks should focus on the conditions that enable customers to trust the technology.

The answer is clear: control, transparency, security and a human employee within easy reach. That is where the opportunity lies to make AI not only smarter, but above all more trustworthy.

Authors: Josien van den Hoek, Sonja Pieters, Rachel van der Plas.