01 · Why
I started building these tools because I kept running into the same problem: we had more ideas than time to make them.
Product demos, web graphics, motion, product screens. The creative was different, but a lot of the production behind it was surprisingly repetitive.
So I started figuring out what I could systemize. Some of it uses AI, some of it uses code, and some of it is just a better workflow. The goal is the same: spend less time setting up the work and more time designing it.
02 · Product demos
From product UI to a finished demo
We often need to tell a product story before the product screens are ready to tell it.
The UI might be outdated, unfinished or missing the states we need. Rebuilding everything manually can take as much time as the motion itself.
I built a workflow that takes the screens we have and brings them into the current BitGo brand visual language. From there, I can build the missing states, use real content and data, and turn the flow into something ready for motion.
I’ve used the same process across demos for AI Search, trading, settlement, security and other product stories.
From rough product screens to polished, editable, on-brand UI without rebuilding every frame by hand.
03 · Brand graphics
Turning one feature into a brand image
The demos show the product being used, close to one to one. These are the opposite. A creative takes a single feature, magnifies it well past life size and stages it as a brand image, so it can sit next to a line of copy on the website and carry it.
Same source material, different job. A demo has to be faithful. A creative has to read from across the page, hold up beside a headline, and still be recognisably our product.
So there's a second tool for that. I give it the cleaned-up UI, the approved copy it's meant to carry and a short brief, and it runs the same order I would.
It starts from what the feature actually does rather than what the copy says about it. One product moved money by mirroring a balance instead of sending it, and that single fact became the picture: the asset, its reflection, and the line between the two. The copy on its own would never have gotten me there.
Then it puts up three directions, and at least one of them is not a screenshot. Left alone, everything becomes a dark frame with a card in the middle, and a page of those reads as one graphic repeated. I pick the direction.
03 · Brand graphics
It builds the chosen one in Figma. The UI gets cropped down to the single control the copy is about, then scaled past life size. Everything behind it drops back so there's a real foreground. Brand material comes from measured values in our own files rather than an approximation by eye.
The type stays live and every layer stays editable.
I stop it there on purpose, before anything moves. Composition is cheap to change while it's still.
Then motion. A demo plays a flow. A creative shouldn't. This one reveals once and then settles into something ambient that never restarts, written as code on a single clock so every duration divides the loop and the final frame is the first frame. A graphic that keeps replaying its own entrance is one you stop looking at.
The last pass is verification, and it's why any of this ships. Type checked against the ladder, numbers checked against each other, fonts confirmed by asking the browser instead of trusting my eye, the loop seam measured on the exported frames. A loop can match at both ends and still tear in the middle.
Sometimes the first pass is close. Sometimes it isn't. Either way I start with the product staged, the composition standing and the motion already looping, instead of an empty frame.
04 · Internal tools
Building the brand into the tools
The more I built these tools, the more obvious one problem became. Telling an AI to “make this on-brand” doesn’t mean much unless it actually knows the brand.
So I’ve been taking the same rules and references our designers use and building them into the tools. Type, color, effects, materials, examples and the decisions behind them.
That gives the system a much better starting point before I ever ask it to make something.
There are also clear limits. BitGo is a regulated company, so the tools don’t invent product claims or marketing copy. Those come from approved inputs. The tools can help make the work. They don’t decide what BitGo says.
05 · Motion
Using AI where it actually helps
AI has changed how I approach motion too, mostly at the beginning of the process.
I use it to explore style frames, environments and visual directions quickly. Once I know what I want, I move into the tools that give me the control I need for the final animation, including code when it makes sense.
I don’t need AI in every part of the process. I use it where it helps, then switch to the tools that give me the precision the final work needs.
06 · What changed
The result isn’t more AI-generated work. It’s less time spent rebuilding the same things.
Product screens don’t always start from scratch. Web graphics don’t always start with an empty Figma frame. I can explore motion before committing to full production. And a lot of the setup work that used to take hours now happens before I touch the file.
The tools give me a better starting point. The design still comes from me.
