Creative Agent: the future of Creatify

Creatify was moving from a set of separate tools to one agent that can do it all. I was asked to design what that future looks like.

Memory map grouping preferences, decisions and asset references

Role

Product design, end to end
UX, UI and everything in between

Design

Product vision
UX flows and UI
Visual direction
Frosted glass components

Motion

The agent orb and its states
Micro-interactions
Generation loading states

The brief

One agent, instead of a toolbox.

Creatify started as a platform of tools: one for video ads, one for avatars, one for images. The plan was to bring all of it into a single agent you could just talk to.

Below is one real run, start to finish: a single sentence asking for an Apple Watch ad, and the finished ad that comes back. Most clips on this page come from that run.

The challenge

Not another chatbot.

A glowing blue orb above the line How can I help you today

A light AI home screen with a soft yellow and pink glow behind the prompt box

A writing app on a pastel gradient: just start writing

A frosted glass description panel with a small icon set

A personal AI app set in a warm serif, with soft pink fades

The mood board: a glowing orb, soft gradients, frosted glass and warm, quiet type.

Most AI products end up looking the same: a chat window and a text box. I wanted the agent to feel like it had a personality, so I collected products that use light, glow and soft gradients to feel alive.

But this is a tool for making ads. The user's work has to stay the focus, so whatever I added couldn't compete with it. The personality had to live in small moments: the first screen, the waiting, and an orb.

Inside the agent

Where the personality lives.

Small moments from one real run, start to finish. Every clip was recorded in light and dark: flip the switch any time to see them all in the other theme.

Show people which mode they're in.

We were also building AdFlow, a node based canvas, and wanted it to live in the same place as the agent. So how do you show someone which mode they're in?

A small switch on the first screen moves between Chat and AdFlow. Pick AdFlow and the background changes: a dotted grid falls into place, the same grid the node editor uses. Before you've typed anything, you already know where you are.

A new chat: the glow rises, then Chat to AdFlow and back.

A glow that rises to meet you.

Open a new chat and gradient orbs float up from beneath the prompt box, drawn from Creatify's brand colors. Click into the box and the caret glows too: a small hint of how much you can do from here.

An orb that shows what the agent is doing.

Once you send a request, the orb becomes a kind of mascot. Researching, thinking, analyzing, generating: each state has its own color and motion, so you can always tell what's happening without reading a log.

In the chat, while it researches the product page.

A question before a storyboard.

Before it spends time generating, the agent pulls the product shots from the site and asks which creative direction to take, each option spelled out in a line. One click, and it's off.

Feedback, instead of waiting.

Generation takes time, so I designed the waiting too. While an image is being made, its frame fills with blurred orbs drifting around inside it, so you know something is being created.

When that image is turned into a video, it blurs and its own colors become the same orb loading effect, so you watch the video take shape from the picture you started with.

Images: drifting orbs, then the picture. Shown at double speed.

Video: each image dissolves into its own colors, then plays. Shown at double speed.

One sentence in, a finished ad out.

Five scenes, stitched into one video with titles, music and an end card, ready to post or to cut again in a different shape.

These small moments add delight, but they also do a job: they show people what the AI is doing instead of leaving them waiting with no feedback.

The launch video, produced together with a professional agency.