A nine-screen analytics product, a 25-slide deck and a clickable prototype — from a single meeting transcript to the conference floor in seven days. Figma Agents did the execution. The governed design system did the governing.
A national pharmacy benefit manager needed a demo-ready analytics product for its GLP-1 Benefit 360 program in time for the MedLab ’26 industry conference. The brief landed seven days before launch.
The product had to tell one coherent story to a live audience: population-level GLP-1 spend and utilisation, a clear path into two clinical programs, and a member-level view clinical teams could act on — all inside the client’s real brand system, backed by a defensible cost-and-outcomes narrative.
The starting material was a recorded internal working session. One transcript. No requirements document, no wireframes, no approved direction.
The same screen at two ends of the week. Structure was decided by hand before anything was generated; the final build came out of the governed component library with the brand audit already applied.
Most “AI in Figma” conversations are about prompt-to-webpage generation. That is a different tool for a different job, and knowing the difference is what made this timeline survive contact with an enterprise design system.
Directly in Figma, from the existing component library. Placeholder components, real structure, no visual polish — plus a parallel content-free zoning pass across all nine screens, restoring the wireframe-before-comp discipline the project had initially skipped.
Once the wireframe was validated, Figma Agents generated and iterated multiple full-fidelity interactive versions from the governed component library — real components and tokens, not freehand AI UI. Stakeholders reacted to several genuine options within hours instead of weeks.
Figma Make starts from a prompt and builds something new — typically a standalone page with no connection to an organisation’s real components, tokens or governance. For an enterprise product with a brand system already in place, Agents is the tool that keeps AI speed compatible with design governance.
Fourteen stages, one continuous working process. This is the deliverable list, not a narrative.
Before a single screen. One transcript, parsed into something traceable.
Two fidelities, every screen, built in Figma against the real library.
Three formal rounds, each one written up rather than left in a meeting.
Two separate sets, because the conference audience is not the eventual user.
The two things that actually went to the conference floor.
AI produced the work. Vetting it against reality was a person’s job, every time. Three catches made it into formal documentation because they had real downstream consequences.
The build used an amber tone for “Medium” risk, but the brand guide’s alert system defines only green, magenta and red. Rather than silently forcing it to match or shrugging past it, the mismatch was logged as a documented brand exception with the clinical reasoning reconstructed.
The brand guide PDF and a separate color reference listed different hex values for the same accent — a one-character difference easy to skim past. It was flagged as an open item requiring confirmation rather than guessed at.
Early Agents output showed small plausible drift — wrong typeface in a chart header, a color near the palette but not from it, an element built freehand when a governed component already existed. As prompts carried more design-system context, later versions needed noticeably less correction.
The deck and the prototype both shipped. Marketing vetted the final build against the brand system and signed it off before the floor opened.
The demo landed. It generated more qualified sales leads from prospective new customers than the team had forecast, and attendee feedback on both the product story and the walkthrough was strongly positive. The deck opened every booth conversation; the clickable prototype carried the hands-on half.
Why that matters more than the speed. A seven-day turnaround is only interesting if the thing that ships survives a brand review and holds up in front of buyers who do this for a living. This one did both.
The product spans nine screens. Two more of them, past the population view and into a clinical program.
The interactive prototype and the design file behind it.
The production-fidelity, navigable prototype used for the hands-on portion of the demo. Starts at the main dashboard and runs the full population → program → member path.
The underlying Figma file — the nine screens, the component usage, and the wireframe-to-final progression the prototype was assembled from.
Same method, your product. Three things worth knowing before a compressed timeline starts.
Agents accelerates work inside a governed component library. Without real components and tokens to build from, AI generates plausible-looking UI with nothing underneath it. If you do not have a system yet, that is the first engagement, not this one.
Speed comes from generation. Correctness comes from somebody who knows the system checking each output against it. We do not hand over anything that has not been audited against your own guide.
Requirements, findings rounds, documented exceptions. Compressed timelines usually lose the decision record first. Here it was a deliverable, which is why nothing in the final build was a last-minute guess.