AI AdoptedRebuild your business processes as self-improving loops
Y Combinator partner Tom Blomfield argues company hierarchies become obsolete once domain knowledge is legible to AI.
Instead of using copilots in existing workflows, companies should design nightly loops that diagnose code failures, update context, and commit fixes overnight. This turns ephemeral software into a compounding asset, scaling revenue per employee without adding headcount.
13:28● Watch on YouTubeHow to Build a Self-Improving Company with AIY CombinatorSell outcomes instead of software seats to align incentives
OpenAI research leader Bob McGrew says pricing outcomes instead of software seats forces builders to solve real customer problems first.
50:42● Watch on YouTubeThe FDE Playbook for AI Startups with Bob McGrewY CombinatorRun AI agents in public channels to teach your whole team
Shopify CEO Tobi Lutke runs internal agents in public Slack channels so employees can observe and copy successful patterns.
Centralise company context to let staff build their own agents
Y Combinator partner Pete Koomen says unifying database access and tool registries lets employees build their own automated workflows.
46:30● Watch on YouTubeInside YC's AI PlaybookY CombinatorYour business value survives in work that cannot be benchmarked
Venture capitalist Sarah Guo says AI value resides in proprietary workflows where correctness relies on private, internal human judgment.
Fix your data and permissions before you deploy agents
Box CEO Aaron Levie warns that messy data permissions block agents long before they hit model limits.
1:13:01● Watch on YouTubeState of Enterprise AI 2026: Aaron Levie on Tokenmaxxing, Rise of Headless, and AI-Proofing Your JobThe MAD Podcast with Matt Turck