A Series B technology company's finance team was exhausted. Every month-end was a 14-day sprint — the controller and two senior accountants working nights and weekends just to get books closed. Annual overtime cost was running $180,000. The CFO knew it was unsustainable. Headcount was growing, complexity was increasing, and the close wasn't getting shorter.
The company had been quoted $320,000 by a consulting firm to "redesign and automate" the close. The project timeline was nine months. They called us instead.
We started by mapping every single close task — not at a high level, but at the granular level that actually drives elapsed time. We identified 87 distinct tasks in the close cycle, from the first journal entry on Day 1 to the final CFO sign-off on Day 14.
We then classified each task into three categories: fully automatable (no human judgment required), AI-assisted (judgment required but AI can draft), and human-only (requires professional judgment and cannot be automated). The breakdown was striking: 34 tasks were fully automatable, 28 were AI-assisted, and 25 required human judgment.
We designed and built automation for the 34 fully-automatable tasks first, targeting the highest time-consumers:
For the 28 AI-assisted tasks, we built AI drafting workflows that produce first-draft journal entries, reconciliations, and flux analysis — the accountant reviews and approves rather than creating from scratch.
"The controller used to manage close with a spreadsheet and prayer. Now she manages it with a dashboard and a coffee."
First close after deployment: 5 days. By the third close: 3 days. Annual overtime eliminated. The controller now describes the close as "almost boring" — which is exactly how a close should feel. The CFO reclaimed her weekends and the finance team stopped dreading the last week of every month.
Total engagement cost: $18,500. Annual savings: $180,000 in overtime plus immeasurable improvement in team morale and retention risk.