YOU CAN'T AI YOUR WAY OUT OF A BROKEN PLATFORM
summary
Claim: layering AI atop a fragmented platform multiplies failures instead of fixing them. Example: asking AI to deploy a service still needs repo discovery, pipeline understanding and infra configuration — teams often have five different pipeline patterns, missing docs and manual handoffs, so AI churns out changes that hit prod, skip lower environments and increase incidents. AI behaves like AI as a bulldozer user: fast, literal and context-free. Fixes that work: enforced guard rails, reduce variation with golden paths, and make ownership explicit so outcomes are traceable.
AI-written summary · about
chapters (15)
- 0:00AI everywhere today
- 1:54AI as productivity myth
- 3:41Broken systems reality
- 5:52AI amplifies platform flaws
- 8:39DevOps lessons revisited
- 11:25AI pressure on assumptions
- 12:07Treat AI as platform user
- 13:53Everyone as platform users
- 16:02Platforms must enforce safety
- 17:26Make right thing easy wrong hard
- 18:08Guardrails over documentation
- 18:29Consistency in pipelines
- 19:12Golden paths matter
- 19:55Explicit ownership needed
- 21:25Build solid platforms
related talks
17mAI-Driven Platform Engineering
29mHow to use AI to fight your DevOps transformation!
28mWhy AI Works for Individuals but Fails in Enterprises, and How to Fix It.
39mFrom Demo to Production: Why Agentic AI Systems Fail - and How to Fix Them
28mAI for Everything After Code
25mAI for Platform Engineering
08mBuilding an AI Platform Your Engineers Actually Trust
05mThe Evolving Role of Platform Teams in the AI Era