Helping Copilot Studio's makers build agents faster

Copilot Studio's homepage was trying to serve everyone at once. Returning makers needed to pick up where they left off. New makers needed to understand what they could even build. Both were being met with the same cluttered screen, a chat interface nobody was using, a wall of templates that led mostly to dead ends, and no clear sense of what Copilot Studio was actually for.

Previous research confirmed what was obvious from looking at it. Chat engagement was low. Template adoption was minimal, and the ones that did get used represented a tiny slice of maker types. Most users arrived, looked around, and didn't know where to start.

The original request was simple, add a toggle to the chat so makers could switch between creating an Agent or a Workflow. Our orchestrator at the time couldn't disambiguate between the two, so product needed a quick fix before a major conference. I pushed back. A toggle would have solved one small problem while leaving everything else broken. I made the case for more scope, got the capacity, and got to work. I was the only designer on it.

The first thing I did was separate the two audiences using the homepage. New makers don't know where they're going. Returning makers do. So I made a simple call — design the top of the page for the person who's never been there before, and let returning makers scroll. Recent agents moved below the fold. It sounds obvious in hindsight but the original page treated both audiences exactly the same, which meant it worked well for neither.

Templates were next. They were the most prominent thing on the page after the chat, and they were sending most makers into dead ends. I didn't try to fix them, I made the case to move the handful that actually worked to the Agents page and cut the rest entirely. This was an easier sell than it sounds. Product had already been signaling that templates might disappear in a future release, and I had built enough trust through earlier work that my recommendation landed. Sometimes the right design decision is knowing which battles you've already won.

The chat needed to actually communicate what it could do. The original version sat on the page with no context, no label, no prompt, no indication of what would happen if you typed something. I added clear framing around it and made the creation paths explicit so makers understood their options before committing to a direction.

One smaller detail worth calling out, I constrained the max-width of the entire page. On large monitors the original design stretched edge to edge, making everything feel sparse and disconnected. Pulling it in improved readability and made the page feel intentional rather than unfinished. It's the kind of change nobody notices when it's right, but everyone feels when it's wrong.

The conference deadline meant shipping in stages. The first improvement replaced the cluttered original with a single clear question, “What would you like to build?”, followed by the toggle product needed, and three quick start cards that finally communicated what Copilot Studio could actually create. Recent agents moved below the fold where returning makers could still find them without overwhelming new ones.

The toggle was a direct response to a technical constraint. Our orchestrator couldn't reliably determine whether a maker wanted to build an Agent or a Workflow from natural language alone. The toggle gave makers explicit control while we worked on improving the underlying technology. Once the orchestrator improved, the toggle was removed in the next iteration.

The response at the conference was immediate. Workshop attendees praised how much cleaner and simpler the page felt without losing any functionality. Our EVP noted the simplicity directly, which for a page that had been trying to do everything at once, felt like exactly the right outcome.

The conference ship was always meant to be a bridge, not a destination. With the immediate problems solved and stakeholder trust established, I designed the full homepage redesign that's now in development and will be in production by the time you read this.

The headline is dynamic, the verb cycles through words like “automate”, “create”, and “simplify” on a timed animation. It's not just visual polish. It's doing real work, showing makers the range of what Copilot Studio can do before they've typed a single character.

The chat helper text works the same way, sample prompts rotate through real agent and workflow examples, giving makers who don't know where to start an immediate foothold. The blank page problem is one of the biggest drop-off points in any creation tool. This addresses it directly.

Previous work moves into a collapsible section at the bottom. Returning makers can still find it, but it no longer competes with the primary job of the page, getting makers building something new.