How to measure AI ROI — CFO AI Ledger

When it comes up

When approving AI business cases, reporting AI value to the board, renewing AI contracts, or when sponsors report hours saved and you need to know what it is worth.

Your options

Hours saved × loaded cost

Fits when: Early pilots in well-understood processes where time is the main benefit.

Watch out: Overstates value unless the time becomes cost out, avoided hires or measured new output.

Cash and P&L impact (cost out, avoided cost, revenue, working capital)

Fits when: The use case changes a budget line, a hiring plan, DSO or a revenue metric.

Watch out: Attribution: other changes happen at the same time; needs baselines and ideally a holdout.

Unit economics and quality (cost per item, error rate, cycle time, forecast accuracy)

Fits when: High-volume processes and decision-support use cases.

Watch out: Quality improvements must be tied to consequences (write-offs avoided, decisions changed) to count as value.

Portfolio return (total AI spend vs total measured benefit)

Fits when: Board and investor reporting across many use cases.

Watch out: Hides poor performers inside the total; needs use-case-level tracking underneath.

How to decide

Baseline first: volume, time, cost, error rate, cycle time.
Pick the value mechanism per benefit: cost out, avoided cost, redeployed capacity, quality, revenue.
Model it in the ROI builder: NPV, payback, sensitivity, break-even adoption.
Track monthly against the case: adoption, unit outcome, run cost.
Report two lines: P&L value and capacity/quality value, separately.
Aggregate to a portfolio view for the board, with use-case detail available.

Trade-offs

Hours saved are easy to measure but rarely cash; P&L measures are credible but slower and harder to attribute; unit economics are precise but narrow; portfolio returns tell the board the big picture but can hide failures.

Real examples

JPMorgan Chase CEO Jamie Dimon said in October 2025 that the bank spends about $2 billion a year on AI and saves at least about the same amount, a portfolio-level spend-versus-benefit comparison (company statement reported by Semafor). Source: Semafor — JPMorgan CEO Jamie Dimon says bank's AI investment has already paid off (Oct 8, 2025) ↗
Gartner's July 2026 CFO research advised setting AI success metrics tied to enterprise objectives rather than measuring pilots or hours saved, and found that 45% of CFOs say their AI investments lean toward productivity versus 20% toward decision quality. Source: Gartner — 45% of CFOs say their AI investments lean towards productivity, 20% towards decision quality (Jul 20, 2026) ↗
Esker's 2026 survey found 65% of CFOs report difficulty connecting AI usage to specific business outcomes, versus 40% of other finance leaders. Source: Esker — New research finds finance leaders want more proof and control before giving AI greater decision-making authority (Sep 24, 2026) ↗

What to check first

Whether a baseline exists for each use case.
How each benefit line becomes cash, capacity or quality.
Whether run costs (seats, tokens, people) are tracked against the case.

Learn and use

Sources

Last reviewed 2026-10-01.