Most companies approach AI the same way: a workshop, a 'transformation roadmap', six months of committees, while a competitor quietly automates support and starts winning on response time. You don't need a transformation. You need one AI employee, doing one real job, in 30 days.

The 30-day plan we actually use

Week 1: Pick the job, define 'done'

  • One role only: support tickets, lead follow-up, invoice chasing, weekly reporting, pick your biggest leak
  • Write the success metrics before anything else: response time, % auto-resolved, meetings booked, hours saved
  • Gather the knowledge it needs: FAQs, price lists, policies, past ticket/conversation samples. You have 80% of this already.

Week 2: Build and connect

  • Train the AI on your actual data, not generic answers
  • Connect it to your tools: WhatsApp Business, CRM, ticketing, calendars
  • Set escalation rules: exactly when it hands over to a human

Week 3: Supervised pilot (the honest week)

  • It works real cases; a human reviews everything it does
  • This is where week-2 assumptions break, we fix them. Every deployment has this week. Vendors who promise otherwise haven't done one.
  • Tune tone, edge cases, escalation thresholds daily

Week 4: Go live, measure, expand

  • Human review drops to spot-checks
  • Metrics dashboard goes to whoever owns the P&L
  • Decision point: expand to the next role, or adjust scope. Either is fine, you're deciding with data, not opinions.
The companies winning with AI didn't pick the perfect use case first. They picked a real one, shipped in a month, and let results choose the next step.
Free strategy call

Want Your Own AI Employee?

Book a free 30-minute session. No commitment, no pitch, we'll map exactly where AI can save you hours every week and grow your business. If AI isn't the right answer for you, we'll say that too.

Book a Free Strategy Call →

What you need on your side

  • One decision-maker who can say yes (committees are how 30 days become 3 quarters)
  • One domain person for 4–5 hours a week (the one who knows the process)
  • Access to the tools it will connect to
  • Honesty about what's broken, the pilot exists to find it, not to hide it

Common Questions

What if the pilot doesn't hit the metrics?

Then you've spent 30 days and a pilot budget to learn that this role isn't ready, with full documentation of why. Compare that to the alternative: guessing for a year. But in practice, starting with genuinely repetitive roles, most pilots clear their bar and expand.

Which role should we start with?

The one with the most repetition and the clearest metric. For most companies that's customer support (tickets auto-resolved) or sales follow-up (leads responded to in minutes). Reporting is the third most common.