1. Lay the foundation

• One shared folder as the company operating system: strategy, pricing, and vision as .md files, a subfolder per function with its expected work and skills

• Design system first, so anything built anywhere calls the same components

• Zone the product: heavy construction off limits, stable areas open, low-stakes areas wide open

• Shared skills, not personal ones. A UX research skill that records, extracts takeaways, clips highlights, and posts publicly cuts a research cycle to under an hour

2. Explain the how

• Specific tactics, not vague mandates ("use AI for X, Y, Z" vs "be AI-first")

• Demo workflows live in meetings

• Embed with teams to find concrete 2x opportunities

3. Do it in public

• Route data and build questions through a public channel where AI answers first and tags the humans who can correct it

• File corrections back into shared context so the next answer is better

• Skeptics convert when they see the value and the failure modes at once

4. Reward the first ship

• Real incentive for the first non-engineer to ship to production

• Power users teach: workflows shared in Slack and one-on-ones

• If nothing ever ships slightly wrong, the guardrails are too tight

5. Prioritize high-impact work

• Target work that saves 10+ hours a week and needs judgment but not deep expertise

• Start with clear ROI wins to build momentum

• Skip vanity metrics like tokens burned or PRs shipped

The measure of success: product goes from 70 to 80% of the day in meetings to 70 to 80% of the day building.