Skip to content
AI Economics Hub

Independent analysis of AI cost and value

Make AI investment economically legible

AI Economics Hub connects AI cost, evidence and business outcomes so leaders can decide what to fund, redesign, stop or scale.

Prepare an investment decisionThe 5-minute version

Prefer to watch? The narrated explainer, with captions

What do you need to work out?

Or browse by understand or implement, or pick a route by task or role.

Map of the site

The AI Economics Map. Framework, 8 minFrameworksThe AI Value GapThe AI Abundance Paradox. Framework, 10 minProof for autonomy. Framework, 4 minLibrary (64)Open questionsStart HereThe first 100 days. Guide, 9 minThe portfolio and cost register. Template, 4 minAI TCO frameworkAttribution coverage calculatorHow to measure AI value. Guide, 10 minThe instrumentation blueprint. Guide, 5 minThe AI value ledger. Framework, 8 minScale, hold, redesign, stop. Framework, 3 minDecision casesThe AI Portfolio Review Pack. Template, 6 minOperating modelBoard reporting pack. Template, 4 minGlossary (95)Key numbersVendor mapStandards mapPosition ledger

The AI Economics System

One chain from AI cost to an investment decision

  1. 1 MeasureWhat does AI fully cost, where is it consumed, and who owns it?
  2. Start Here, interchangeChange for The AI Economics Map
  3. The first 100 days
  4. The portfolio and cost register
  5. AI TCO framework
  6. Attribution coverage calculator, interchangeChange for The AI Value Gap
  7. 2 ProveWhat changed because AI was used, and how strong is the evidence?
  8. How to measure AI value
  9. The instrumentation blueprint
  10. The AI value ledger, interchange
  11. 3 DecideGiven the economics and the evidence, what should be scaled, held, redesigned or stopped?
  12. Scale, hold, redesign, stop
  13. Decision cases
  14. The AI Portfolio Review Pack, interchange
  15. 4 OperateWho owns each part, how often is it reviewed, and what reaches the board?
  16. Operating model
  17. Board reporting pack

Understand

The argument, the evidence and the models

  1. The AI Economics MapChange for Start Here
  2. Frameworks
  3. The AI Value Gap, interchangeChange for Attribution coverage calculator
  4. The AI Abundance Paradox
  5. Proof for autonomy
  6. Library (64)
  7. Open questions

Reference shelf

Look something up

  1. Glossary (95)
  2. Key numbers
  3. Vendor map
  4. Standards map
  5. Position ledger

Every station is a page. Ringed stations are interchanges, where one line of thinking meets another.

Same arithmetic, different decision

A positive return is not enough on its own. The evidence behind it decides how much to commit next.

See the decision rulesExplore the cases
Decide

Document processing assistant

Annual cost
£60,000
Forecast annual benefit
£96,000
Base-case net benefit
£36,000
Evidence
Before and after

Hold expansion and test against a comparison group.

Illustrative example. Invented figures.

Key numbers

PwC, 29th Global CEO Survey
56%

of CEOs report no significant financial benefit from AI so far; 12% report both cost and revenue gains

PwC, 29th Global CEO Survey, 2026
4,454 CEOs in 95 countries and territories, surveyed 30 September to 10 November 2025

State of FinOps 2026 (FinOps Foundation)
98%

of FinOps practitioners now manage AI spend, up from 31% two years earlier

State of FinOps 2026 (FinOps Foundation), 2026
1,192 respondents, over 83bn USD annual cloud spend, self-selected practitioners

Gartner, reported by CFO.com
57%

of finance AI use cases had unclear returns, even though 55% of finance leaders reported positive returns overall

Gartner, reported by CFO.com, 2026
160 senior finance-function leaders; fieldwork dates not stated in coverage

Surveys describe their respondents, not the whole market. All verified figures and caveats.

Built to be challenged

By Ed Epstein, who works at IBM. The postings on this site are my own and don't necessarily represent IBM's positions, strategies or opinions. AI tools are used in research, drafting and production under human editorial direction, and material claims are checked against the cited source before publication.

Used the method on a real decision? Contribute a case.