Why Users Want Humanized AI, But Still Need to Know It’s a Machine

There’s a very specific moment in digital products where people stop interacting with a system only as a tool and start emotionally responding to it. Conversation changes that relationship completely. The moment a product starts talking back, users naturally begin expecting something beyond functionality. They expect clarity, reassurance, patience, and guidance in a way they normally wouldn’t expect from a regular interface.

And that changes the responsibility of the experience itself.

A button can fail and still feel neutral. Language rarely does. The way a product responds during moments of uncertainty directly affects how reliable, competent, and trustworthy it feels. That’s why conversation design has always interested me much more as a behavioral challenge than a purely technical one.

I noticed this very clearly while working on BIA, Bradesco’s virtual assistant, back in 2019. At the time, most people still didn’t fully understand how these systems worked or what they were actually capable of. Many users approached BIA carrying years of frustration from interacting with traditional phone support systems. They expected rigid commands, robotic answers, confusing loops, and the feeling that they needed to “adapt” themselves to what the machine would understand.

But during the research process, something interesting started appearing repeatedly. Even when users clearly knew they were talking to a robot, they still expected warmth and empathy during the interaction. At the same time, they also wanted transparency. People appreciated when the experience felt lighter, more natural, and easier to follow, but they didn’t necessarily want the illusion of talking to a real person.

That contradiction became one of the most important insights of the project.

Because there’s a delicate line between making technology feel approachable and making it feel artificially human. And I honestly think many companies still struggle to understand where that line should exist.

A lot of conversational products today focus heavily on personality, trying to sound funny, casual, emotional, or extremely “human.” But in real usage, most people aren’t looking for artificial intimacy. They are looking for orientation. They want to feel understood, guided, and safe while solving a problem.

Those are very different things. During the BIA project, much of our work focused on understanding how users actually communicated their problems in real life. Not how the bank internally categorized services, but how people emotionally described confusion, urgency, insecurity, or frustration. Many financial terms that felt completely normal inside the company sounded distant and intimidating to users outside that environment. And when language creates distance, trust usually starts breaking down very quickly.

This is why I’ve always believed UX writing goes far beyond microcopy. The words inside a product shape emotional perception constantly. Every confirmation message, fallback response, explanation, or error state influences how supported people feel during the interaction.

That becomes even more important in financial products because people are often already anxious before the interaction even starts. They may be dealing with payments, debts, support requests, financial decisions, or uncertainty around money. In those moments, the tone of a system can either reduce tension or amplify it.

A confusing interaction is rarely interpreted as “just a technical issue.” Most people internalize it emotionally. They feel ignored, lost, or unsupported.

And I think this is where many conversational experiences fail today. Some products sound impressive in presentations but become exhausting in everyday use because they prioritize personality over comprehension. The interaction may feel innovative for five minutes, but over time users simply want efficiency, clarity, and predictability.

This discussion around conversational tone and trust isn’t new. Researchers and writers like Torrey Podmajersky have explored for years how consistency, clarity, and emotional predictability shape user trust far more effectively than exaggerated personality. Nielsen Norman Group has also repeatedly highlighted how users respond better to transparent systems than to interfaces attempting to simulate humanity too aggressively. More recently, guidelines from organizations like Google and Microsoft have reinforced the importance of expectation management, disclosure of limitations, and emotional safety within conversational systems.

And honestly, I think users already understand this intuitively.

People are generally comfortable interacting with technology. What they don’t want is feeling manipulated by it.

The most effective conversational experiences are usually the ones that know how to stay simple. They understand rhythm. They know when to explain something carefully, when to reduce information, and when silence is actually better than overexplaining.

Looking back, one of the biggest lessons from the BIA project was understanding that humanizing a product does not necessarily mean making it behave like a human. In many cases, users feel much safer when the system is transparent about what it is, what it can do, and where its limitations exist.

Warmth and empathy matters, but honesty matters too.

And maybe that balance between warmth and transparency is what makes conversational experiences truly trustworthy in the long run.