The 30/60/90 plan for AI rollout
A practical timeline for going from "we should try AI" to "we shipped three plays" without breaking ops.
Most "AI strategy" docs read like a wish list. This is a real 30/60/90 we've run with 40+ small businesses — built around shipping three plays in 90 days without breaking ops.
Days 1–30: pick one painful workflow
Don't start with strategy. Start with the workflow that wakes you up at 2am.
- Week 1. List your top 10 recurring pain points. Pick one with high volume + clear inputs + clear outputs. Good candidates: missed calls, quote requests, no-show recovery, insurance verification.
- Week 2. Time the human cost. How many minutes per instance? How many instances per week? This is your ROI denominator.
- Week 3. Pick a tool. Don't custom-build. Use what exists (Voiceflow, Synthflow, Rasa, an off-the-shelf AI receptionist).
- Week 4. Ship the first version to a single channel (one phone line, one inbox). Not the whole company.
By day 30 you have one workflow live and a baseline metric.
Days 31–60: measure, fix, expand
Now you have data. Spend this month making it real.
- Pull the logs daily for the first two weeks. Tag every interaction: clean win, handoff needed, failure. Don't trust the dashboard summary — read the actual conversations.
- Fix the top 3 failure modes. Usually these are prompt issues or missing context (calendar access, customer history). Cheap to fix.
- Expand to a second channel only after fail rate is under 10%.
End of day 60: one workflow running cleanly, second one ramping.
Days 61–90: institutionalize and pick play #3
This is where most rollouts die. Two things to lock in:
- Write the SOP. Who owns the AI? Who reads the failure logs? Who has authority to update prompts? If the answer is "the founder," you don't have a process — you have a hobby.
- Train the affected team. Show them where AI helps, where it hands off, and how to give feedback. The people closest to the work catch the failures fastest.
Pick play #3 from the original top-10 list. Repeat the 30-day loop.
The mistake to avoid
Don't try to roll out three plays in parallel. Sequential wins compound. Parallel pilots stall. Ship one, measure, ship the next.
By day 90 you'll have three working plays, a real ROI number, and a team that trusts the system.