Finance Engine: finance function rebuilt around agents
Invoice intake through to the close, with exceptions routed to the person who can clear them.
Modelled inside NetSuite, Ramp, Concur and the shared drives a 24-person finance team actually works in.
- Audit · 4 weeks
We map how the work actually moves
14 sessions with 11 people, 38 documents ingested and 9 systems mapped. Produces the system map, the exception taxonomy in your own language, a volume baseline and a build plan. $15–25k, credited in full against a build signed within 90 days. - Build · 12–14 weeks
Agents shipped inside your stack
6 agents against your connectors, your thresholds and the exceptions the audit found. Human gates sit where the confidence floor or the value threshold puts them. Fixed fee. - Operate · monthly
Somebody keeps it correct
Monitoring, model swaps, retuning against the corrections your team makes, and extending into the next workflow. The fee tracks that workload rather than a share of the build.
From the first interview to agents running in production
Against $180M of revenue and a 24-person finance function
One department, end to end. 12 workflows analysed
- Cost reduction$412k58%
- Revenue growth$133k19%
- Risk mitigation$161k23%
Cost reduction is 12,784 agent hours a year at $43 fully loaded, discounted 25% for ramp. Revenue and risk are derived separately and listed in the basis. Value is captured across all three, not from cost alone.
The Engine already exists. What the audit adds is your systems, your thresholds and the exceptions nobody has written down, which is what an agent needs before it can be pointed at real data.
One department carried end to end beats five departments each given a pilot. The second Engine at the same client reuses the connectors and the cores, which is why it costs less than the first.
Modelled against Northbridge Industrial Group, a modelled composite: $180M of revenue, 640 employees, 4 legal entities across United States and Canada, 6 sites, private equity owned, year two of a five-year hold. Every figure in this chapter derives from that profile. 18,800 in-scope manual hours a year sit behind it.