Week 1: audit, touch nothing
The first days are read-only. We export every workflow with its trigger, actions, last-run date and error count into a register. We list pipelines and stage counts, custom fields and how many contacts actually use each, tags and their counts, calendars and their sync status, and the state of A2P registration and domain authentication. The output is a one-page findings summary the client can understand: what's leaking leads, what's harmless clutter, what's dangerous.
Week 1: stop the bleeding
Before anything structural, we fix the things costing money today: a broken or missing speed-to-lead, reminders that aren't firing, nurtures still messaging booked customers because they lack stop conditions, and any workflow throwing errors. These fixes are small, isolated and testable. The client sees results in the first week, which buys patience for the slower work.
Never edit a live workflow in place during a rescue. Duplicate it, fix the copy, test with real leads, then swap. If the fix is wrong, the original is one click away.
Weeks 2 to 3: rebuild the foundation
New pipelines designed with proper stage rules, new fields and tags following the naming convention, built alongside the old ones. Contacts are mapped from old stages and tags to new ones by workflow, with counts verified at each step. Channels get their A2P registration and domain authentication sorted if they never were. The team keeps working in the old structure while the new one fills up.
Week 3: migrate workflows
Each automation from the register is either rebuilt on the new structure, merged into another, or marked for retirement. New versions are tested with real leads across every channel. Only after a new workflow has run cleanly for a week does its old counterpart get switched off, not deleted, so it can be restored if something was missed. Twenty-three "follow up" workflows usually become four.
Week 4: retire and document
Old pipelines are archived once empty, old tags removed after mapping, orphaned fields deleted after a final check. Then the documentation: the register updated to the new state, a flow diagram, the naming convention, and recorded walkthroughs for the team. Without this step the account drifts back within a year; with it, the next person, or the next partner, can pick it up cold.
Signs an account needs rescuing
- More than ten workflows with "follow up" in the name, and nobody knows which are live.
- A pipeline stage holding hundreds of contacts from more than a year ago.
- Texts marked sent that customers say they never received.
- Booked customers still receiving "are you still interested?" messages.
- The team keeps a spreadsheet "because the CRM isn't reliable".
Two or more of these and the audit call will pay for itself in the first week.
Key takeaways
- Audit first, in read-only mode: list every workflow, its trigger, its last run and its owner.
- Stop the bleeding immediately: fix speed-to-lead and reminders before anything structural.
- Rebuild pipelines, fields and tags to the naming convention alongside the old ones, not instead of them.
- Migrate workflows one at a time, testing each with real leads, and switch the old one off only after the new one has run clean for a week.
- Finish with documentation and training, or the account will drift back within a year.