A services firm, 140 people. 31 senior hours per month, returned.

Context

A Northern European professional services firm of ~140 people. Senior partners were spending 20+ hours a week on review and reconciliation work that was neither billable nor delegable. Growth had paused because the people doing revenue work were the same people doing administrative work.

The board asked for the AI plan. "We're using ChatGPT" was not an answer that survived the next question.

What we found

The 10-day Scan surfaced four candidate cases. All four had CFO-defensible ROI numbers. The firm chose to build one first — month-end reconciliation and client-report drafting, rebuilt around structured AI assistance with senior review gates. The other three sit in their backlog, pre-scoped.

What we changed

• First-draft reconciliation reports generated from accounting-system exports, flagged by exception rather than re-keyed.
• Client report narratives drafted from a structured knowledge base, senior review on every outgoing document.
• Review gates preserved: nothing client-facing leaves without a named senior's name on it.

Outcome

31 hours per senior, per month — returned to revenue-generating work.

~€9,300 per month in margin per senior, at the firm's blended rate.

Scan: 10 business days. Build: 6 weeks. In production: week 8.

What it cost them

Scan: €4,000. Sprint: €22,000. Payback: month 2.

A note on the guarantee

The Scan surfaced four candidate cases. The firm built one. The refund clause was never invoked — but it was the reason the procurement conversation took ten minutes instead of ten weeks.

What's next

The three cases the firm didn't build first are pre-scoped and costed. Two are on the 2026 H2 roadmap.

---

Identifying details withheld. NDA on request. For reference calls, email sales@cbrx.ai.