Skip to content

Orientation

Start Here

Start with the task in front of you, or with your role. The map below shows how the rest of the site fits together.

Start Here at the top, leading to three sections. Understand: the argument, the models, the evidence, and the debates. Implement: the system in four phases, templates and tools, decision cases, and the operating model. Reference: glossary, key numbers, vendor map and standards map.
Pick a door below, or start from a task or a role further down the page.
Start Here at the top, leading to three sections. Understand: the argument, the models, the evidence, and the debates. Implement: the system in four phases, templates and tools, decision cases, and the operating model. Reference: glossary, key numbers, vendor map and standards map.

Pick a door below, or start from a task or a role further down the page.

Two ways to use this site

UnderstandThe argument, the evidence and the models. Read this to know what is happening and why.ImplementThe system, templates, worked cases and the operating model. Use this to do it in the real world.

You can switch at any point; every page links to its counterpart.

Start with what you need to do

Grouped by the phase of the AI Economics System each task belongs to. Each path is three or four pieces, in order.

Phase 1

Measure

Full cost, consumption and ownership.

Start from scratch

  1. 1The first 100 days
  2. 2The portfolio and cost register
  3. 3Attribution coverage calculator

Work out what AI really costs

  1. 1AI TCO framework
  2. 2The real cost of AI inference
  3. 3Reference cost ranges

Compare models and hosting options

  1. 1The economics of model selection
  2. 2Rent, reserve or own intelligence
  3. 3Token economics explained

Set up showback or chargeback

  1. 1FinOps and AI
  2. 2The instrumentation blueprint
  3. 3Programmatic access to AI costs

Phase 2

Prove

Baseline, outcome, attribution and evidence level.

Build an AI business case

  1. 1What CFOs should ask of AI ROI claims
  2. 2AI TCO framework
  3. 3The seven-sheet TCO worksheet
  4. 4How to measure AI value

Prove whether a use case created value

  1. 1How to measure AI value
  2. 2The instrumentation blueprint
  3. 3The AI value ledger
  4. 4Proof of concept versus proof of value

Phase 3

Decide

Scale, hold, redesign or stop.

Run a portfolio review for the board

  1. 1The portfolio review pack
  2. 2The benefits register
  3. 3The board reporting pack
  4. 4Decision cases

Review an AI supplier contract

  1. 1The supplier economics review
  2. 2Embedded AI, hidden tokens
  3. 3Per-seat pricing under pressure
  4. 4The vendor map

Decide whether to scale or stop a pilot

  1. 1Scale, hold, redesign, stop
  2. 2When to stop an AI initiative
  3. 3Proof of value scorecard

Phase 4

Operate

Ownership, cadence and reporting.

Forecast next year's AI budget

  1. 1How to budget for AI
  2. 2Why cheaper AI will cost more
  3. 3ITFM and AI

Control agentic spend

  1. 1Agentic AI economics
  2. 2When every manager can create AI spend
  3. 3Proof for autonomy
  4. 4The cost of verification

Design the operating model

  1. 1Who owns AI economics?
  2. 2Operating model overview
  3. 3The CIO can orchestrate AI value, but cannot own it

Or start with your role

Finance lens

CFO and finance transformation

AI spend is variable, often buried inside automated loops and SaaS renewals, and harder to trace than cloud. Much of the instrumentation built for cloud does not reach it.

Lead metric

Cost per successful outcome, and attribution coverage as the portfolio headline

Technology lens

CIO, CTO, CDO, Head of AI / CoE and Product

You are accountable for AI value across functions you do not control, with tooling that measures cost well and value barely. The fix is one accountable view, not another dashboard.

Lead metric

Attribution coverage, plus the portfolio value-versus-cost view that names winners and laggards

FinOps and practitioner lens

FinOps Leader and Procurement

FinOps for AI is FinOps plus a value layer, not a new world. Your visibility, allocation and culture playbook transfers; what is new is outcome attribution, agent governance and the meter going dark.

Lead metric

Attribution coverage as the extension of allocation; cost per successful outcome as the new unit economics

Governance and board lens

Risk/Legal, HR/People and Board

Three questions cover the whole estate: what share of our AI spend connects to a measured outcome; are we automating, redesigning or reinventing; and can we govern what we are scaling?

Lead metric

Attribution coverage, the redesign/reinvent mix, and the governance register

A path for CFOs

Six reads, in order, for a finance leader who needs to decide what to fund, hold or question.

  1. The AI Value Gap for why spend is visible and value is not.
  2. What AI actually costs for planning ranges and what sits behind the model invoice.
  3. What CFOs should ask of AI ROI claims for the questions that test a claim.
  4. The AI Value Ledger for one record per workflow, with an evidence level and a decision.
  5. Scale, hold, redesign, stop for the decision rules.
  6. When to stop an AI initiative for exit criteria written before the money is spent.

More specific roles

CTO / Engineering

Engineering gets measured on AI vanity metrics first and blamed for the bill second. The defence is measuring outcomes, not activity, with frameworks that resist gaming.

Lead metric

Cost per successful outcome, DORA and SPACE for the productivity layer

CDO / Data

Value telemetry is a data problem, and the enabling foundation, pipelines, lineage, readiness, is the largest unfunded line in most AI business cases. The cost taxonomy finally gives it a place to be costed.

Lead metric

Outcome telemetry coverage, and the Prepare-stage cost made visible

Head of AI / CoE

You have to scale past pilots, govern agents, and justify the budget, simultaneously. The value gap is your enemy and your argument: close it visibly and the budget defends itself.

Lead metric

Attribution coverage rising quarter on quarter, and the laggards you retired

Procurement

AI breaks the SaaS pricing model you negotiate against. Per-seat assumes people do the work; agents do not occupy seats. And AI cost is increasingly hidden inside renewals you already signed.

Lead metric

Cost per outcome in contracts; consumption exposure in renewals

Risk / Legal

The risk shifts from wrong answer to wrong action as AI gains autonomy, and liability needs an owner before the agent acts, not after. Value reported without netted-off risk is optimism, not measurement.

Lead metric

Share of AI features with an owner, a risk tier and evaluation coverage

HR / People

The binding constraint on AI value is human: habit, fear, incentives, redesigned work. The blockers are human, and so is the fix.

Lead metric

Genuine adoption depth (repeat use, not log-ins), and reskilling delivered against roles that change

Product

Every AI feature now has unit economics, and pricing is migrating from seats toward outcomes. Feature decisions are economic decisions.

Lead metric

Cost per session and value per outcome, per feature

Continue exploring

Most readers move between value, cost, governance and the operating model. These are the main sections.

FrameworksOperating ModelOpen questionsLibraryGlossaryAbout the publicationMethodology