The 5-week rollout pattern for automation projects
Most failed automation projects fail in the same way — too much scope, too late a launch. This is the lean rollout pattern that ships value in week two.
Most automation projects that fail share one cause: they tried to automate everything at once, didn't ship anything for 3 months, and lost stakeholder confidence before the system went live. The pattern that works is the opposite — narrow first, ship fast, expand from a working base.
Week 1 — Discovery & one workflow
Pick one workflow. Not three, not 'all of customer support'. One. Something like 'auto-reply to pricing questions on WhatsApp' or 'AI-screen incoming CVs for the senior engineer role'. Map the current manual process step by step. Identify the exact decision rules. Spec the integrations needed.
Week 2 — Build & test
Build the chosen workflow. End-to-end. Integrated. Tested with real (anonymized) production data. By end of week 2, you have a system that handles the chosen scenario correctly — but it's running in shadow mode, not live.
Week 3 — Live with safety net
Turn it on for real traffic, but with a human reviewing every output for the first week. The team validates the system is making correct decisions before the safety net comes off. Tune any edge cases that emerge.
Week 4 — Autonomy + dashboard
Remove the human review for the cases the system handles well. Keep humans on the edge cases. Ship the manager dashboard showing volume handled, escalations, error rate, and time saved. This is where stakeholders start asking 'what else can we automate?'.
Week 5 — Expand from the working base
With one workflow live, working, and measurable — you've earned the right to add the next. Pick the next workflow, repeat the cycle. After 5 such cycles, you have 5 automated workflows, each working, each measurable, each adding value.
What kills this pattern
- Scope creep in week 1 — 'let's add CRM integration too'. Don't. Ship one workflow first.
- Skipping week 3 — going straight to live without a human review week. The first edge case kills stakeholder trust.
- No dashboard in week 4 — if leadership can't see the value, they won't fund the next phase.
- Treating week 5 as 'project done'. The whole point is that week 5 is the start of the next cycle.