The Death of Static AI Assistants

What building a character taught me about the future of AI products.

For the past year, Rita has been the face of Renalto.

She helps contractors write quotes, understand projects, and navigate the product.

But until recently, Rita wasn't really alive.

She was a folder of PNGs.

One image for smiling.

One image for waving.

One image for working.

One image for thinking.

In 2024, this was enough.

Give an AI assistant a face and a few expressions and users would do the rest. Humans are surprisingly willing to anthropomorphize software.

By 2026, something has changed.

The best AI products are no longer trying to feel like tools.

They're trying to feel like characters.

AI Is Becoming Character-Driven

A few years ago, AI assistants were mostly text boxes.

Then they got names.

Now they're getting faces.

Duolingo has Lily.

Microsoft recently introduced Mico, a small blob-like companion for Copilot. Microsoft is careful not to call it a mascot, describing it instead as an optional visual identity that makes voice conversations feel more natural. The language itself is interesting. Not smarter. Not faster. More natural.

Apple has done something similar. During WWDC, a blue-and-white character with an oversized head began appearing throughout Apple Intelligence demos. The internet quickly gave it a nickname: Little Finder Guy.

Neither character exists because users need them.

They exist because AI is becoming increasingly relational.

The companies building these products understand something important: people don't just interact with intelligence. They interact with personalities.

The shift is subtle, but it feels inevitable.

We're moving from assistants to characters.

The most interesting AI products aren't just competing on intelligence anymore.

They're competing on presence.

Why I Chose Rive

Once I started paying attention, I noticed the same tool appearing everywhere.

Rive.

Duolingo uses it.

Spotify uses it.

Notion uses it.

Rive sits somewhere between an animation tool and a game engine. Instead of exporting videos, you build systems.

Characters can blink.

Look around.

React.

Change states.

Respond to user input.

The more I learned about it, the more it felt like the natural choice for Rita.

While researching, I came across a great article about Duolingo's Creative Technologists. The company created a role that sits somewhere between designer and engineer.

Reading it felt familiar.

Much of my work at Renalto already lives in that space between product design, software engineering, AI, and storytelling.

Bringing Rita to life felt like a natural extension of that.

Learning Rive

Thankfully, animation wasn't completely new to me.

Before software engineering, I spent years working in motion design and post-production.

The problem was that I hadn't animated seriously in a long time.

I was rusty.

The first few hours felt awkward.

Then things started coming back.

First Rita blinked.

Then her eyes moved.

Then she waved.

Then I built my first state machine.

What surprised me most was how quickly the tool clicked.

Rive feels less like traditional animation software and more like building behavior.

You're not animating a timeline.

You're teaching a character how to exist.

The Lily Effect

Around the same time, Duolingo published a breakdown of how they brought Lily to life for their AI Video Call feature.

One line stuck with me.

The goal wasn't to animate Lily.

The goal was to create presence.

That distinction matters.

Nobody cares whether a character is animated.

People care whether a character feels present.

Whether it acknowledges them.

Whether it feels aware.

Whether it feels alive.

Rita isn't having conversations yet.

She isn't reacting to voice input.

She isn't lip-syncing.

But the principle is the same.

Every blink.

Every glance.

Every small movement.

Each one makes the experience feel slightly less mechanical.

Falling Down the Rabbit Hole

Of course, once you start learning character animation, you discover there are levels to this.

While researching Duolingo's animation systems, I found an article about visemes.

A viseme is essentially the visual version of a sound.

When you say "O," your mouth forms one shape.

When you say "F," it forms another.

Simple enough.

Until you realize every character needs its own version.

Lily doesn't speak the same way Duo speaks.

Oscar doesn't speak the same way Zari speaks.

Suddenly making Rita blink felt easy.

Making Rita talk felt like an entirely different discipline.

Before and After

The funny thing is that the changes are subtle.

Rita doesn't do very much yet.

She blinks.

She looks around.

She waves.

That's about it.

But the difference feels larger than the feature list suggests.

The old Rita felt like artwork.

The new Rita feels like a character.

And I suspect users notice that difference immediately, even if they can't explain why.

The Future

The most interesting thing happening in AI isn't that models keep getting smarter.

Everyone is focused on intelligence.

I'm becoming increasingly interested in personality.

The history of AI interfaces feels surprisingly simple.

First we had text.

Then we got voices.

Now we're getting characters.

A few years ago an AI assistant was just a textbox.

Today it has a face.

Tomorrow it will have expressions, gestures, and personality.

Eventually, the line between software and character will become increasingly blurry.

Bringing Rita to life wasn't really an animation project.

It was a small step toward that future.

Today she blinks.

Tomorrow she might speak.

And someday she may feel less like a feature and more like a teammate.

At what point does a character stop being interface and start being something else?