AI ROI research, analysis, and glossary
How enterprises measure the return on AI investment.
MEASUREDVALUE.AI is an index of primary research, sourced analysis, and defined terminology for measuring the return on enterprise AI spend. Entries are tagged by what they measure: adoption, cost, or outcome. Written for finance, FinOps, and AI governance teams.
AI ROI statistics, with sources
56%
In PwC's 29th Global CEO Survey, published January 2026, 56% of CEOs reported neither a cost benefit nor a revenue benefit from AI.
$3.70
An IDC study commissioned by Microsoft and widely cited in 2026 reports a return of $3.70 for every $1.00 of AI spend.
95%
MIT research reported by Fortune in August 2025 found that 95% of enterprise generative AI pilots do not reach measurable P&L impact.
Research
View all →Ramp · Sep 2026
Ramp Economics Lab
A monthly Ramp AI Index update built from card and bill-pay data across 70,000+ US businesses, now splitting AI spend across subscriptions, coding agents, and tokens/APIs. Built entirely from Ramp's own customer base — sample-flag it the way this index flags any single-platform data source.
Cost
Ray Rike and Peter Buchanan · Sep 2026
AI to ROI
A weekly issue with an AI story, report, metric, and use case of the week, plus a Saturday news roundup. The authors also publish the Big Book of AI Metrics, a standing collection of 81 metrics. Also on Substack.
Cost
Quesma · Sep 2026
Inside the First Tokenomicon
A firsthand recap of the inaugural Tokenomicon in Amsterdam, covering the Big-T Notation, the Tokenomics Mind Map, and talks from UNICC and Infracost. Quesma sits on Tokenomics Foundation working groups — noted here as an affiliation, not a disqualifier.
Adoption
Analysis
View all →Aug 20, 2026
Why the same ROI number means opposite things
IDC's $3.70 and MIT's 95% failure rate are not contradictory. They are measuring different cohorts, on different timelines, answering different questions.
Aug 5, 2026
Why most pilots stall before P&L impact
Why AI ROI is so hard to prove to the board: Deloitte's lengthening payback periods and MIT's genAI divide describe the same gap from two directions — the distance between a working pilot and a number the P&L actually notices.
Jul 22, 2026
What finance teams should count as AI return
IBM's KPI framing and Wharton's productivity metrics point at the same fix: define the unit of return before the pilot starts, not after.
Adoption, cost, and outcome are three different measurements. A tool that reports AI adoption is not reporting AI cost, and a tool that reports cost is not reporting outcome. Every entry on this site records which of the three its source measures.
Glossary term
AI business case
A structured estimate of expected AI costs, benefits, risks, implementation requirements, and strategic alignment.
Full glossary →Frequently asked
What is AI ROI?
AI ROI is the measured financial return on money spent on artificial intelligence, expressed as a ratio or net figure against the total cost of that AI. In practice the term is used for three different measurements: how much AI is being used (adoption), what AI consumption costs (cost), and what business result the AI produced (outcome). Published AI ROI figures are frequently not comparable with one another because they answer different ones of the three.
Why do AI ROI estimates vary so widely?
Published estimates range from a claimed $3.70 return per dollar spent (IDC, commissioned by Microsoft, 2026) to findings that 95% of enterprise generative AI pilots reach no measurable P&L impact (MIT, reported in Fortune, August 2025). The estimates differ mainly in four ways: what they count as return, whether costs are taken from invoices or estimated from token prices, whether a counterfactual baseline exists, and who commissioned the study.
What is the difference between AI cost visibility and outcome attribution?
AI cost visibility reports what was spent, broken down by model, team, application, or user. Outcome attribution connects that spend to a business result and establishes that the AI caused the result. Cost visibility is largely a solved engineering problem. Outcome attribution requires a baseline or counterfactual, which most published methods do not establish.
How should a CFO measure return on AI spend?
There is no single accepted method as of 2026. Current approaches fall into four groups: reconciling metered cost against the provider invoice, measuring cost per unit of output or per outcome, comparing an AI-assisted cohort against a matched non-AI baseline, and self-reported time savings converted at a loaded labor rate. The four differ substantially in defensibility. The research index catalogs published methods and the assumptions each one requires.
How are entries in this index verified?
Every entry records the source, publication date, sample, and method of the study it covers, along with the date the entry was last checked. Research entries are tagged by what they measure: adoption, cost, or outcome. Tags are assigned editorially and are never self-assigned by a submitter. The site carries no advertising and accepts no payment for placement, ranking, or inclusion.