AI Transformation

From Traditional to AI-Native PDLC

Emerging tech, a new way to build across Product, Eng, and Design.

Code is the new shared language across Product, Engineering, & Design practitioners.

April 2025. A week inside the new generative tools — Replit, Lovable, Bolt, v0. The question shifted. We stopped asking whether AI could help and started asking what we could build with it.

AI Moderated Research

Research was the first loop we rebuilt. Userology let us drop a Figma prototype into a study, have AI moderate the session with an IT admin on the other side, and get back a synthesized read on themes and pain points by end of day. It joined the weekly cadence — fast enough to guide decisions with quant and qual data.

The PDLC pivot

Every stage — research, spec, prototype, build — assumed an answer took days or weeks. These compressed it to hours, sometimes minutes.

Slide titled 'We've Been Building the Right Things the Traditional Way' listing four pain points of the legacy loop: long handoff chains between PM, Design, and Eng; static prototypes that don't reveal real behavior; expensive feedback loops; and design and engineering speaking different languages late in the cycle. — The diagnosis we walked in with — the loop that had always worked, now visibly out of step.

When engineering caught up

For most of 2025, design ran ahead of engineering. Q4 changed that: Cursor licenses rolled out across the org and design picked up seats alongside. Our design system already lived in Vue 3 and React, so we wrote skills, rules, and page templates that taught Cursor how our product was actually built — our tokens, our components, our opinions. Designers built real interactive flows against the production system the same week they scoped the work.

A JC Skills node branching to Circuit DS, Tech Writing, AWS Deployment, and Eng Best Practices, shown alongside a dark code editor file — circuit-ds/usage-patterns.md — listing eleven numbered gotchas covering DataTableToolbar, ActionMenu, ActionsToolbar, AppNavigation, FormField, DataTable, FilterModal, ToastNotification, and PageHeader. — Skills, rules, and usage patterns — the vocabulary we handed Cursor so it could speak in our tokens.

Workflow diagram mapping Cursor Skills and the design system through five product lanes — Devices, Identity, Access, Workflows, Settings — into two collaboration zones: PM + Designer Collaboration (new branches, revise the design, live testing, shareable URLs) and ENG + Designer Collaboration (share for business logic, QA and refinement, push back to main). — From a handoff to a shared surface.

“The front-end stopped being a border between disciplines and started being a shared workspace.”

The Vision

A streamlined AI-assisted PDLC where everyone works in code and a shared context moves through the stages. That context also generates the help docs, blog posts, and GTM materials.

Three-card diagram under the heading 'One shared language: code.' 01 Prototype (Product Manager) — PRD to functioning prototype using JumpCloud design system components. 02 Refine & Validate (Design Engineer) — polished experience, deployed to AWS, real customer research from a live URL. 03 Ship (Engineering + Design Engineer) — production-ready, backend-complete, QA-validated. Footer: Code and context passes forward each phase. — Three phases, one shared language — the handoff becomes a baton, not a fence.

Role Evolution

Flexibility in tech has always been the game. The AI era is no different. Roles shift as capabilities emerge.

Table titled 'What This Changes — For Everyone' with three columns: Team, Before, After. Rows: PMs to Builders — writing specs, waiting for mocks to building prototypes, driving research. Product Designers to Design Engineers — late-stage polish to shaping the experience from day one. Manual Engineers to AI-Assisted Engineers — frequent scope changes, unclear intent to clear validated spec, faster execution. Manual QA/QE to Testing Automation — catching problems late to validating a pre-tested experience. Customers — seeing it at launch to shaping it before launch. — What changed, role by role — every seat at the table moved one step earlier in the loop.