The 3-phase adoption playbook

The goal of Phase 1 is not adoption — it's an existence proof inside your own walls that leadership can't argue with and neighboring teams will envy.

• Pick one team with real deliverables, not an innovation lab. The proof must happen on work someone is already accountable for, or it proves nothing.
• Pair an executive sponsor with a working operator. The sponsor clears blockers; the operator uses the tool daily and ships with it. If the senior owner doesn't personally use it, the org notices.
• Instrument before/after on a small number of workflows — cycle time, review turnaround, output shipped. Two or three honest numbers beat a wall of vanity metrics.
• Ship something visible in the first two weeks. A real artifact the rest of the org can see and touch converts more people than any deck.
• Start the security review now, not at rollout. The security gate is the single most common place enterprise AI adoption dies — clearing it is Phase 1 work even though its payoff is Phase 2.

Expansion works when teams ask to join. Push-based rollouts create compliance theater; pull-based rollouts create users.

• Expand along the pull. Publicize the Phase 1 wins internally, then onboard the teams that raise their hands first — they become the next proof points.
• Name a champion per function. Adoption spreads peer-to-peer inside a discipline; a champion who ships with the tool beats a central enablement team broadcasting at everyone.
• Make onboarding frictionless: licenses pre-provisioned, access pre-approved, a 30-minute setup path, office hours, and working examples from their own function — not generic demos.
• Build function-specific playbooks. "How engineering uses it" and "how operations uses it" are different documents. Write them from real usage, not speculation.
• Clear the security and compliance review completely — data handling, retention, acceptable use. One published, approved policy unlocks every conservative team at once.
• Let the laggards lag. Chasing the resistant 20% in Phase 2 burns the energy you need for the willing 80%.

A rollout that depends on its champion is a pilot that hasn't admitted it yet. Phase 3 turns adoption into infrastructure.

• Build usage observability. You cannot govern, defend, or expand what you can't see. Instrument who uses the tool, how often, and on what — I built this layer myself (scraper → SQLite → FastAPI → React/D3) rather than wait for a BI project.
• Report ROI in operator terms: hours returned, cycle-time deltas, output per team — laddered to the business metrics leadership already tracks. Adoption percentage is an input, not the result.
• Set the policy and guardrails in writing: what's encouraged, what's reviewed, what's off-limits. Ambiguity re-freezes the conservative middle of the org.
• Govern cost per outcome, not cost alone. Spend that returns hours is a bargain; unexamined spend is how programs get cut in the next budget cycle.
• Institutionalize the enablement: training curriculum, onboarding for new hires, champions with named backfills — so the capability survives any single person leaving.