Moving quickly without compromising quality using AI

We’ve all been asked to go faster and do more. The fear is that quality will slip. Recently, our team was tasked with bringing transparency to Meta's advertising enforcement system to unblock over $1B/year in revenue. We went from nothing to a Director Level-1 aligned MVP in just 5 days — and we didn't sacrifice craft. Instead, we used AI at every step to ensure that moving fast actually increased the quality of our output. Here is the 5-step AI first playbook we used to guarantee quality from brief to backlog.

Step 1: Instant, robust product briefs

• Tool — We generated our sprint proposal from scratch using a Claude skill, establishing constraints, risks, and requirements, which human experts then audited.
• The impact — We had a comprehensive, aligned starting point on day one without spending weeks debating scope.
• How it ensured quality — Traditionally, fast briefs are full of holes. By using AI, we ensured no edge cases or technical constraints were missed, creating a rock-solid foundation for the sprint.
• Duration — Minutes, instead of days or weeks.

Step 2: Grounded research synthesis

• Tool — We fed vast amounts of raw user research into Google’s NotebookLLM to create an instantly queryable knowledge base.
• The impact — We could ask direct questions about user pain points and get immediate, cited answers.
• How it ensured quality — Moving fast usually means skipping research synthesis and relying on assumptions. This allowed us to remain deeply user-centric and data-driven without slowing down to read hundreds of pages of PDFs.
• Duration — Instant answers instead of days of reading, trying to remember, or looking for decks.

Step 3: Parallel design exploration

• Tool — We used Manus and FigmaMake to rapidly explore, evaluate, and refine four distinct design revisions against our core user problems.
• The impact — We evaluated a wide breadth of solutions and deep edge cases simultaneously.
• How it ensured quality — In a traditional rush, designers are forced to pick one "good enough" concept early and stick with it. AI allowed us to stress-test multiple directions to ensure we were actually building the right solution, not just the fastest one.
• Duration — Hours instead of days.

Step 4: High fidelity prototyping

• Tool — We bypassed static Figma mockups entirely, using Manus to generate a fully interactive, database-backed prototype that simulated real state changes (like appealing an account restriction).
• Tool  — I also built a custom AI agent to act as a content design assistant — generating high-confidence UX copy for a zero-to-one product where every string needed to be written from scratch. I then reviewed and refined the output. This specialised AI tooling allowed me to scale myself across 5 teams; 40 engineers to support with self-serve content design assistance at scale.
• Tool  — To validate our designs, I created synthetic users using Manus to give feedback directly into Figma. These mimicked user queries, feedback and concerns eerily similar to real life, right down to user segment.
• The impact — The cross-functional team and leadership could click through a real experience with real content during the design review.
• How it ensured quality — Static screens with arrows leave too much to the imagination, and placeholder copy masks real usability issues. A live prototype with validated content forces you to solve interaction and communication problems at the same time. Without the AI content agent, we would have been 100% blocked — content design capacity is always tight, but always required.
• Duration — 1 day, instead of weeks of prototyping, user research and content iteration.

Step 5: Frictionless execution planning

• Tool — We utilized Claude Code to automatically translate the approved prototype into our execution backlog.
• The impact — We generated precise user stories, technical constraints, and engineering tasks instantly.
• How it ensured quality — The handoff from design to engineering is where quality often degrades due to missing documentation. AI ensured every detail of the validated prototype was captured perfectly for the engineers.
• Duration — Minutes instead of days of ticket writing.

The takeaway

Pushing back for quality doesn't mean delays. Today, in a world where everyone can execute at the click of a button, "quality" increasingly means stress-testing product thinking and strategy to reduce noise that can lead to XFN confusion and overwhelm. "Quality" doesn't just mean a good user experience. It means robust product briefs, grounded research, stress-tested prototypes, and bulletproof documentation.  We executed all five steps in 5 days. AI handles the heavy lifting of production and documentation, allowing you to focus entirely on judgment and craft. Quality and speed are no longer opposites.