WhoBuilt

Why Prototypes Beat PRDs: How to Ship 10x Faster with AI

Aug 3, 2025· 43 min· 5.0K views·Watch on YouTube

Speaker

Jeremy Epling, CPO of Vanta and former VP of Product at GitHub and Microsoft, talks with Peter Yang about prototyping, AI workflows, product careers, and building reliable AI products.

Use prototypes to make product ideas concrete

Epling says writing helps teams reason through customer problems and business impact, while prototypes make the experience easier to judge. At Vanta, PMs and designers use tools such as v0, Figma Make, and Lovable; a shared v0 template helped nontechnical PMs get started. Prototypes let teams discuss customer value and flow before getting lost in implementation details.

Run product walkthroughs around the customer experience

Epling calls these sessions product walkthroughs and asks presenters to clarify whether they want early concept feedback or polish before launch. His advice is to get to the prototype quickly, click through the experience, and set expectations for feedback. Reviews can include detailed attention to wording, especially because Vanta serves customers who may find security compliance intimidating.

Use AI to learn from work and customer conversations

Epling uses LLMs to learn unfamiliar compliance topics, synthesize performance feedback, and turn spoken thoughts into structured documents. Vanta also built internal tooling on Gong and Slack to surface customer conversations about competitors, features, and product areas. He values access to the original call so product teams can hear what customers actually said.

Grow through ownership, technical skills, and business context

Epling sees a strong path for senior individual contributors, with progression measured by autonomy and the ability to take ambiguous problems from questions through customer research and cross-functional decisions. He expects PM roles to involve broader business and customer scope, and says technical expertise will remain valuable. He advises PMs to build relationships across disciplines and understand areas such as sales, marketing, and pricing.

Evaluate AI products for quality and trust

Vanta's agent surfaces policy inconsistencies, shows supporting sources, and proposes edits for customer approval rather than leaving users with a blank chat prompt. Epling says the team uses shared evaluation tools, subject matter experts, and more than 50,000 evaluations to improve answer coverage and accuracy. He also describes testing emerging capabilities over time and choosing chat, cards, or other controls based on the task.

Takeaways

  • Pair written product thinking with a prototype so teams can judge the customer experience early.
  • Give prototype reviewers a clear feedback goal, then get to the experience quickly.
  • Use AI to find and inspect customer evidence, not just to summarize it.
  • Senior PM growth depends on taking ownership of ambiguous problems and building cross-functional context.
  • Set explicit quality measures and involve domain experts when evaluating AI features.

Mentioned in this talk

Search inside this talk

What was claimed 16

Related talks