Onboarding with AI

Transcript

2. Event name or presentation Subtitle or speaker name M I C R O S O F T B U I L D 2 0 2 6 · D E M O Onboarding a Developer with AI. A better first-day experience.
4. 0 1 · T H E P R O B L E M The first day doesn't scale. 3–9 M O N T H S to full productivity in a new codebase 40% O F T I M E of week-one is spent reading, not writing 1 IN 3 Q U E S T I O N S are never asked out loud 02 / 11
5. 0 2 · T O D A Y Four acts. One new developer. 01 Understand the codebase. From zero to a mental map. 02 Surface hidden context. The decisions that shaped the code. 03 Identify what — and who — matters. Critical components, owners, fragile zones. 04 Ship the first contribution. A real PR. Real conventions. Day one. 03 / 11
6. 0 3 · T H E S C E N A R I O Meet Sarah. Day one. New team. Unfamiliar repo. T H E R E A D M E Last updated 14 months ago. T H E T E C H L E A D In meetings until 5pm. T H E C O D E B A S E 2370 files. 5 languages. T H E D E A D L I N E First PR by Friday. 04 / 11 $ git clone … $ cd the-codebase $ now what?
8. A C T 0 1 Explain this codebase to me. From zero to a mental map. > "I just joined this team. Give me a tour: entry points, key modules, data flow, and what looks like legacy vs. active code." 05 / 11
19. A C T 0 2 Surface the context I'm missing. The decisions that shaped the code — not the ones that made it to the README. > "Why does this file exist? Why this library? What unwritten rules would I get nitpicked on in a code review?" 06 / 11
27. s
33. s
37. A C T 0 3 What — and who — matters? Critical components. Owners. Fragile zones. > "What 5 files do I need to understand first?" > "Who should I ping about the auth module?" > "Which parts of this codebase are the most fragile?" 07 / 11
45. A C T 0 4 Ship the first PR. Day one. Real issue. Real conventions. Real impact. 01 Read the issue. 02 Plan the approach. 03 Implement, following the conventions. 04 Run the tests. Debug if needed. 05 Draft the PR description. 08 / 11
47. s
60. 0 4 · W H Y T H I S W O R K S Three principles to take home. 01 AI as a context engine, not a code generator. In onboarding, the value isn't in writing code — it's in understanding what already exists. 02 The codebase is the documentation. An agent reads what's actually there, not what we wish was there. 03 Senior time is the scarcest resource. Every question answered by AI is focus returned to the team. 09 / 11
61. 0 5 · M O N D A Y M O R N I N G Try it Monday. → Give your next hire an AI agent. Before day one. Not as a perk — as standard onboarding. → Measure time-to-first-PR. If you don't measure it, you can't improve it. → Invest one hour a week in ADRs and commit messages. Your future agent — and your future hires — will thank you. 10 / 11
62. T H A N K Y O U Onboarding used to be a test of patience. With AI, it can become a test of curiosity. That's a much better first day. 11 / 11
63. T H E U N L O C K One prompt. One plan. One PR. Day one.
64. T H A N K Y O U Q U E S T I O N S ? Questions? https://www.linkedin.com/in/hubertmichel