Ops Automation Console: $312K/Year Cut From Operating Cost

A 40-person B2B SaaS had every tool it needed. HubSpot, Stripe, Xero, Slack, Sheets, Notion, all in place and all paid for. What it did not have was anything connecting them, so every handoff between them was a person retyping data that already existed one screen away.

Six people in ops spent the better part of a week each month on that. Not thinking, not deciding. Retyping. Leads arrived and sat until someone routed them. Deals closed and waited for someone to raise the invoice. New customers signed and waited for someone to build their workspace.

We mapped seven processes end to end, then shipped them one at a time over twelve weeks. One at a time mattered. The team could absorb a single change a week, and we could prove each one before starting the next.

Every workflow ran in report-only mode for a full week before it was allowed to write anything. It logged what it would have done, we read the log together, and only then did it get permission. Nothing went live on our confidence alone.

The second rule was that automation never guesses. Anything the rules do not cover goes to an exception queue with the reason attached, and a person decides. A PO over tolerance. A missing billing contact. A classifier below its confidence floor. About four items a day reach that queue out of roughly 150 runs, and that ratio is the point. The system is not clever. It is bounded.

Where it landed: 4,458 runs a month happen with nobody watching. 491 hours a month came back to the team, and no headcount was removed. They moved onto work that needed judgement. Days sales outstanding fell from 41 to 22, because invoices now go out the hour a deal closes instead of the week after. All in, $312,529 a year of operating cost. The ops lead edits the rules now, not us.