FROM PROTOTYPE TO CODE: LEADING MY TEAM INTO AI-LED DESIGN
How I set the foundation and coached my team through an AI-led prototyping workflow that shipped a new capability three weeks ahead of schedule.
The compressed path from client ask to shipped capability
Vision
An existing client needed monitoring alerts for their business and asked whether Total Plan Manager could absorb a capability a third-party vendor had been providing — consolidating more of their workflow onto our platform, on a tight timeline. I saw this as bigger than one client's request: built well, it was a capability every client on the platform could use, and a natural moment to move my team from AI-curious to AI-fluent. I sponsored the initiative, set the workflow, and coached the team through it directly alongside a product lead and engineering.
Prototype to shipped code in one week — delivered three weeks ahead of schedule.Business stakeholder, on the outcome
Problem
Our traditional design-to-development path was reliable but slow: a high-fidelity Figma prototype alone typically took two or more weeks, and even at that level of effort it couldn't capture the depth of interaction and micro-interaction detail engineering actually needed to build confidently. On a client-driven deadline, that gap between "the design looks right" and "engineering can build it" was the biggest risk to on-time delivery.
Goals
- Speed without guesswork. Compress the path from idea to working code without leaving engineering to interpret intent.
- Fidelity that means something. A prototype should be built from real product components and real user needs, not a generic approximation of them.
- A repeatable team capability. This couldn't be a one-off trick. I wanted my team to leave the project able to run this workflow again, on their own.
Our Approach
Facing a demanding timeline, I saw an opening to invite my team past our default, Figma-first habits and into something less familiar. Rather than hand down a mandate, I nudged the team to ideate confidently outside the traditional workflow, using Claude Cowork as the canvas. I wanted them to feel they had permission to try, to get it wrong in places, and to figure it out in the open — so I gave them the runway, the trust, and my own hands-on time to make that possible. Speed was the goal, but so was building a team that felt safe experimenting with a new way of working.
Enabling the Team
Before any project work began, I directed two pieces of groundwork so that moving fast wouldn't mean moving carelessly:
Design-system readiness. I sponsored reorganizing the Total Plan Manager design system in Figma specifically to make it AI-tool ready, then had the team build a custom Claude Cowork skill trained only on our real components and patterns — so AI output couldn't drift from the product.
User fit, by design. Once the team had the opportunity and trust to explore, they hit the ground running with what Claude could do. On their own initiative, they built a Persona skill that fed our user research into Claude as permanent context, so every prototype reasoned about our actual users by default, not just the interface. I stayed close through it all, coaching in working sessions rather than delegating and checking in later.
This allows UX/UI to get much closer to products and teams — you become a truly integral piece of the product development lifecycle.Product Lead, Total Plan Manager
From Prototype to Production
With the guardrails in place, the team built a fully clickable prototype of the entire initial user journey in three days — core flows like rule creation down to micro-interactions like detail reviews, status updates, and batch actions. Because it was built on the real design system, the hand-off carried real information: an engineer could look at it and know exactly which components and code to pull from, without a written spec standing between design intent and the build.
Working as One Team
What made this work was a small team building from one shared, living artifact. The prototype wasn't a deliverable we handed off and left behind — as feedback came in, the team kept tuning it directly in Claude Cowork, and the custom skill improved alongside the work. Every conversation that mattered, whether with program management, product, or engineering, pointed at the same concrete, interactive journey instead of a static mockup. An engineer who had joined the team only weeks earlier could see how the pieces fit and ask better questions, rather than discovering late that something needed to be reworked.
UX became literally just another team member on our product.Product Lead, Total Plan Manager
Outcome
| Prototype | Fully clickable end-to-end journey in 3 days, versus 2+ weeks the traditional way |
| Hand-off | One live walkthrough with product and engineering — no written spec required |
| Build | Working development build in approximately one week |
| Delivery | Shipped three weeks ahead of the original schedule |
Within the client's demanding timeline, we delivered a scalable capability that now serves every client on the platform, not just the one who asked for it — while giving my team firsthand, hands-on experience with a workflow they can now run without me in the room.
Conclusion
This was one experiment, not a finished playbook, and I said as much to the team: we shared it because it's early, not because it's done. The real win for me as a leader wasn't the three days or the three weeks — it was watching a designer go from cautiously testing a new tool to confidently leading a live design review with engineering, because the foundation I'd insisted on made the work trustworthy. That's the pattern I'm now investing in across the rest of the portfolio: build the guardrails first, mentor through the first sprint, then let the team run. Our next goal is at least 80% reusable code from design straight to production — closing the last gap between prototype and ship.