CIO.com · Column
I Audited an Award-Winning AI Project. The Case Study Left Out the Cloud Bill
An AI workflow cut turnaround by 90% and cost about fifteen times more per document than the clerk it replaced.
- Published
- Reading time
- About 9 min
- Authors
- Richard Ewing
What is it about?
Ewing audited a document workflow that had won an industry innovation award. Turnaround on vendor agreements fell by 90% and the vendor published a case study. But during the pilot, compute and model fees were charged to a central innovation fund. Once finance moved those invoices to the business unit, the unit economics turned negative within 60 days.
Manual review cost roughly $0.80 per document; the automated pipeline cost $12–14. Messy real-world scans pushed about 40% of files into an exception queue, where a human needed ten minutes per file, twice the original manual time. The company paid for the cloud bill and the people at once.
Key numbers
- Cost per document: about $0.80 manually versus $12–14 automated.
- About 40% of files fell below the confidence threshold and went to human review.
- Each exception took 10 minutes, versus 4–5 minutes for the original manual review.
- Gartner, quoted in the piece, expects inference cost per agentic workflow to rise more than fivefold through 2028.
Why read it
- It is a concrete case for measuring cost per completed transaction rather than speed.
- It shows how pilot accounting hides the real cost of AI until the project code is closed.
- It ends with three questions you can add to every scale-up gate: full cost per transaction, human intervention rate and cost, and whether the economics hold at production volume.
Who should read it
- CIOs presenting AI business cases
- CFOs and FP&A teams reviewing AI spend
- Innovation teams running pilots on central budgets
About the author: Richard Ewing
Founder of Exogram and CareerWin; writes as “The AI Economist”
Ewing positions himself between engineering and finance. In his words, he doesn't manage a backlog, he manages a P&L. He created the Product Debt Index, a 0–100 score that translates technical debt into dollars.
He founded Exogram and CareerWin and, according to his website, is writing a book for O'Reilly.
Read it carefully: common misquotes
- It is a single anonymized case written as an opinion column, with identifying details changed.
- “90% faster” refers to turnaround on clean files, not total effort.
How it shapes the CIO's AI Playbook
- Step 5.2: Track fully loaded cost per completed transaction · Measure what the P&L sees
- Step 6.2: Pass a scale gate with three questions · Fund and scale
Notes by Pavel Veselov, last reviewed . Summaries are my own; quotes and figures belong to the original authors.
All 11 sources in the CIO's AI Library