What's inside
Every AI investment eventually meets a CFO asking what it returned, and most project teams can't answer — because nobody measured the process before automating it, and nobody agreed what "working" meant in numbers. This paper is the measurement framework we build into every engagement, written for the people who sign the checks.
Contents
- Baseline measurement during scoping — why ROI is decided before the build starts, and how to measure the manual process you're about to automate: task timing, volume counts, error rates, and the fully loaded cost per unit of work.
- The four metrics families — time (turnaround and hours returned), quality (accuracy and error cost), coverage (share of volume the system handles), and risk (auditability, permission integrity, exception handling). Every credible criterion belongs to one of the four; every credible business case uses at least two.
- Monthly measurement under operations — the reporting cadence that keeps a system honest after launch: what to measure continuously versus by sample, how to handle drift, and when a miss should trigger remediation versus renegotiation.
- Appendix: the criteria template — a fill-in acceptance criteria table with metric definitions, threshold guidance, and measurement methods, ready to attach to an RFP or a statement of work.
Who it's for
CFOs, program managers, and anyone buying AI systems or services who needs the spend to survive a finance review — including buyers evaluating vendors other than us. The template in the appendix works on any vendor's contract.