Get Your Team Actually Using AI (Not Just Talking About It)

You bought the licenses. You sent the announcement. Three weeks later, two people use it daily and everyone else forgot their password. That's not a tooling problem — it's an enablement problem, and it's the most common way AI money gets wasted.

If you've seen usage cliff after launch week, read why AI adoption dies in week two for the pattern. This post is the playbook: who owns the install, what the next four weeks look like, and which one habit to measure so you're not managing vibes.

Train on real work, not toy demos

The fastest way to lose a room is to demo something that has nothing to do with their job. Bring the actual artifacts: the messy spreadsheet, the support ticket backlog, the proposal nobody wants to write. When people watch AI handle their work, adoption stops being theoretical.

Champion job description (name one person)

Every team that sticks with AI has one person who answers questions, shares wins, and keeps a running list of prompts that work. A committee diffuses ownership; a champion concentrates it. Put this on a real JD — even if it's 20% of someone's week.

Owns:

• The shared prompt / project library (canonical versions, not twelve forks)
• First-response help in Slack/Teams when someone gets stuck
• One named win per week posted where the team already looks
• The weekly habit metric (definition below) — pull it, don't "feel" it
• Escalating workflow bugs (bad outputs) as system fixes, not user failures

Does not own:

• Being the only person allowed to use AI
• Building custom software (that's a different lane)
• Security policy by vibes — they enforce the written one-pager, they don't invent it alone
• A steering committee's slide deck

Time: Protect a few hours a week. Unfunded champions burn out and the rollout dies with them.

Pick for: Credibility with peers, bias to documentation, willingness to say "that output isn't good enough — here's the fix." Not "most excited about AI." Enthusiasm without stewardship creates demos, not habits.

Make security boring and clear

People hesitate when the rules are vague. Spell out what data can go where, which tools are approved, and what's off-limits. A one-page, plain-language policy beats a 40-slide deck nobody reads — and it removes the biggest excuse not to start.

Four-week rollout calendar

One workflow at a time. Protect the calendar like the habit matters — because if it isn't scheduled, it isn't real.

Week 1 — Install one workflow

• Champion + manager agree on the single first workflow (e.g., call prep, ticket draft, meeting → actions)
• Live session on real artifacts; correct the instructions in the room
• Everyone produces one real output before they leave
• Boundaries one-pager shared the same day

Exit: Each person has one saved win in the shared library.

Week 2 — Manager loop on

• Manager inspects outputs in the existing meeting (pipeline, standup, ticket review) — not a new "AI meeting"
• Champion posts one win; logs top two failure modes and fixes the prompt
• Habit metric baseline recorded

Exit: People see that quality is coached, not policed after a mishap.

Week 3 — Standardize and widen slightly

• Freeze v1 of the prompt/project instructions
• Optional: add a second closely related workflow only if week-1 habit is sticking
• Office hours (30 minutes) for edge cases; everything else stays out of scope

Exit: New hires or late adopters can run the workflow from the doc without a private walkthrough.

Week 4 — Keep, fix, or kill

• Honest review against the habit metric and "would we complain if this disappeared?"
• Keep and schedule ongoing coaching — or fix the instructions — or kill the workflow so dead process doesn't teach cynicism
• Only then queue the next workflow

Exit: A written decision, not "let's give it another month of hoping."

One-habit metrics per function

Don't measure "productivity." Measure one concrete behavior per team. Examples you can actually count:

Function	One habit to count weekly
Sales	# of reps who ran call prep in Claude before a live meeting
CS / Support	# of agents who drafted first reply with the shared prompt before sending
RevOps / Ops	# of pipeline or ops reviews that started from an AI-generated readout
Product	# of feedback clusters produced with evidence quotes (not vibes) before roadmap debate
Leadership / EA	# of internal updates that started from a bullet brief → AI draft → human edit

Track the habit, watch it climb, and let visible wins recruit the holdouts. If the number is flat by week 4, you don't need another tool — you need a harder look at install quality, manager inspection, or scope.

Standardize the prompts that work

Ad-hoc prompting means everyone reinvents the wheel and most do it badly. Capture the handful of prompts that reliably produce good output for your common tasks, put them somewhere shared, and treat them like any other operating doc. The champion's job is curation, not a museum of every experiment.

The throughline

Adoption is about changing how work happens, not adding a tool to the stack. Get one real workflow into daily use — with a named champion, a four-week calendar, and one habit metric — and the rest follows.

If you want the workflows pre-built for GTM teams, the GTM AI Operating System packages them with the rollout system. If you'd rather have it installed with your team directly, that's the GTM AI Sprint. And if you want a quick read on where you stand before week 1, use the readiness scorecard.