Shaving the Yak with AI - Devs want specs, Product Owners want speed
summary
An AI pipeline records short screen videos with Windows+Shift+R or Teams, sends them to OneDrive then Whisper for transcription, and uses GPT‑4o plus Semantic Kernel to generate fully formed work items in Azure DevOps or GitHub. YakShaver creates title, repro steps, acceptance criteria, screenshot and a watch the video link, cutting average write time from ~30 minutes to ~1 minute and scaling to 419 shaves (top 30) in 30 days, deployed company‑wide and actively reducing meetings and rework.
AI-written summary · about
chapters (17)
- 0:00Shaving yaks with AI
- 5:24What is yak shaving
- 8:21Recording bugs and features
- 10:07Words into work items
- 14:42How people describe problems
- 18:54V0 for mockups and specs
- 23:13Time and cost savings
- 28:36Impact on development lifecycle
- 30:21Lessons from AI project
- 41:50Technical challenges
- 47:05JSON in AI workflows
- 49:15Vertical slice architecture
- 51:43Workflow vs agent mode
- 54:14Handling duplicate bugs
- 57:07Future of AI agents
- 59:14Summary and road ahead
- 1:03:10Q&A
related talks
01h 01mLessons Learned Building the Ultimate AI Bug Reporter
06mHow to shave a yak
27mHuman-first leadership, AI-powered
19mMore AI, More Human: Why the New Tools Demand More Care, Not Less
27mFrom AI Assistants to AI Coworkers: How Engineering Teams Ship Faster with Enterprise Context
01h 31mAccelerate Business Delivery With AI and DevOps
36mHalt and shave the yaks: How to add continuous delivery right now
54mGen AI & the SDLC: Changing the way we bring digital products to life