Three Product Workflows to Contend With

AI and the New Reality of Product Workflows
‍AI has accelerated how products are built — but it’s also fractured the way teams work. What used to be a single, structured product development process has splintered into three parallel workstreams:‍

1. Traditional Workflow (Waterfall/Agile Hybrid)
‍The familiar process still exists: PM drafts a PRD or one-sheeter. Design asks clarifying questions, creates the feature or product experience. PM and Engineering review for feasibility, code is written, tested, and launched.This flow is deliberate, documented, and predictable — but also slower.

‍2. Engineering-Led AI Builds
‍Now, engineers often use design system components to quickly ship AI-driven features with little to no design input. Speed is prioritized over experience.New AI behaviors (like unpredictable outputs) create flow and user journey issues.Design is pulled in last-minute to “fix” things — but without dedicated time, most improvements are shelved.Features ship untested and often miss user expectations.

‍3. AI-Generated Designs & Code
‍PMs or designers use AI tools to spin up entire interfaces and code. Output is handed to engineering for refactoring. There’s no research, no user intent, no design review. Engineers inherit code they didn’t write and lack context for why decisions were made.The result: inconsistent, fragile experiences.

‍The Core Problem
‍We’ve introduced two additional, chaotic workstreams when we rarely had the first one running efficiently. Without strong leadership, this results in: Misaligned features that don’t serve the business. Burnout across teams constantly “fixing” each other’s work. Lost market opportunities as competitors ship with clarity and purpose.

‍How I Solve This as a Design Leader
Unify the Workstreams Under One Product VisionEstablish a single, shared source of truth for product direction — tying every build (AI-assisted or not) to clear business outcomes.

‍Integrate Design Earlier in All Workflows
Build processes where design has input before engineers code or AI tools generate output. Introduce lightweight, rapid design validation loops to maintain speed without sacrificing quality.

‍Create AI-Literate Teams & Guardrails
Train PMs, designers, and engineers to use AI effectively and responsibly.Define where AI accelerates — and where human judgment is non-negotiable (research, intent, brand).

‍Redefine "Done" in the AI Era
Shift the definition of "done" from “code shipped” to “solution validated against user needs and business goals.”