The setup
The engagement looked normal at the start: a fixed-scope statement of work, a kickoff call, weekly check-ins for the first month. The consultancy built out lead routing, task assignment, approval chains, and a handful of email automations tied to deal stage changes. The client had no in-house CRM admin — that was the whole reason for hiring outside help — so nobody on staff was reviewing the work at the level of "will I be able to maintain this."
Documentation was promised as a deliverable in the SOW: a runbook, a diagram of the automation logic, an offboarding call. As the project wound down, weekly check-ins became biweekly, then monthly. The final invoice went out. It got paid. The offboarding call got rescheduled twice, then stopped getting scheduled at all.
The collapse
Nothing broke immediately, which is part of why it took months to become an emergency. Automations kept running. But roughly ten weeks after the consultancy went quiet, a routing rule started misfiring — leads from one region were being assigned to a rep who'd left the company two months earlier, because the automation referenced that rep's user record directly rather than a role or queue. No one currently on staff knew the automation existed, let alone how to open it, since none of the 41 rules had names more descriptive than "Rule 12" and "New Flow 3."
The ops manager, who had no formal CRM admin training, ended up reverse-engineering the automation layer by opening every rule one at a time, tracing trigger conditions and actions by hand, and building the documentation the consultancy was contracted to deliver and never did. Two of the 41 rules turned out to be doing nothing at all — leftover test automations that had never been cleaned up. One was quietly duplicating a set of records every time a deal closed, which had been slowly inflating a reporting number for months before anyone noticed the pattern.
The autopsy
Root causes on record
- Documentation was a deliverable on paper, not a gate on payment. The final invoice was tied to build completion, not to a verified handoff.
- No internal owner reviewed the work as it was built. A client with zero in-house CRM literacy has no way to sanity-check what they're being handed.
- Automations referenced specific users instead of roles or queues. This is a common shortcut that turns any staffing change into a silent failure.
- No naming or labeling convention was enforced. "Rule 12" tells the next person nothing about what it does or why it exists.
- No offboarding gate in the contract. There was no consequence for the partner walking away before the handoff call happened.
Recommendation pending
Editor's note: this slot will point to a resource for vetting implementation partners and structuring SOWs so documentation and handoff are payment gates, not good-faith promises.
What the post-mortem actually changed
The ops manager pushed to bring automation ownership in-house going forward, with any future partner contract structured so a portion of payment is withheld until a documented, recorded handoff session happens — not just a deliverable checkbox. Every automation now carries a plain-language name and an owner. It's more overhead per automation. It's also the only way anyone would know what "Rule 12" used to be.