The team rollout, without the chaos.
One person using ChatGPT well is a productivity story. Ten people using it ungoverned is a risk story; ten people not using it is a waste story. If you're the one who has to thread that needle, this is your playbook.
01 Plan choice, decided by three questions
Does company data need to stay out of model training by default? Do you need central billing and user management? Do you need admin controls over connectors and agents? Business-tier plans answer yes to all three and price per-seat; Enterprise adds serious admin (SSO, spend controls, group permissions, per-person overrides — the controls that came with ChatGPT Work matter here). Individual Plus accounts on personal cards answer none of them — fine for solo operators, a governance hole for teams. If people are pasting company information in, you want a company agreement around it.
02 The rules that prevent the horror stories
Write one page. Not eleven — one. It needs exactly four sections:
03 Adoption: surviving the week-two cliff
Every rollout has the same shape: novelty week, then quiet abandonment by everyone who didn't find a personal use. Beat the cliff:
- Seed each role with three recipes. Not "here's ChatGPT" — "here's how estimating uses it for bid letters, here's the ops weekly-report prompt." Specific beats general by a mile.
- Name a champion who isn't you — the person people already ask for help. Give them Pro+ depth (this track) and first-look on new features.
- Run a monthly 20-minute show-and-tell: two people demo one real win each. Peer proof converts skeptics; mandates don't.
- Measure something. Even crude: hours saved self-reported, drafts produced, turnaround time on proposals. What gets counted keeps getting budget.
A third of your team will run, a third will stroll with encouragement, a third will resist until the tools are simply how work happens. Plan for that curve instead of resenting it — and aim the training budget at the strolling middle, where it actually changes outcomes.
04 Revisit quarterly
The tools change monthly (it's why this site exists). Set a quarterly 30 minutes: what's new, what rule needs updating, what recipe should spread. Governance that never updates becomes either ignored or absurd — usually both.
Draft your one-page policy right now — ChatGPT knows the format from this lesson. "Draft a green/yellow/red AI use policy for a [your team] that handles [your data kinds]." Edit it into truth. Ship it this week.
Open ChatGPT →This week's challenge
This week: one-page policy shipped, three recipes seeded for one role, champion named, show-and-tell on the calendar. That's a real rollout — most companies never do all four. (Need help going bigger? That's literally our advisory practice.)