Decide
Given the economics and the evidence, what should be scaled, held, redesigned or stopped?
A record of cost and value is only useful if it changes what happens next. Decide applies four outcomes: scale where economics, evidence and operating conditions support it; hold where potential remains but proof is short; redesign where value is plausible but the workflow or contract weakens it; and stop where no credible route remains. The size of the commitment sets how much evidence is enough.
Do this
Templates, guides and methods for this phase, in a sensible order.
Browser tool, v0.5
Self-hosting break-even calculator
Compare paying per token with running an open-weight model on dedicated GPUs, including utilisation and people. Runs in your browser; nothing is sent anywhere.
Browser tool, v0.5
Cost per successful outcome calculator
Compare up to four model setups on what each successful outcome costs, including the people who fix failures. Runs in your browser; nothing is sent anywhere.
Browser tool, v0.5
AI budget stress test
Test whether an AI case still meets its hurdle if prices rise, demand grows, or both. Runs in your browser; nothing is sent anywhere.
Excel template, v0.5
AI Portfolio Review Pack
Put cost, claimed and realised value, evidence and risk side by side, and end the review in scale, hold, redesign or stop.
Word template, v0.5
AI Investment Decision Memorandum
A seven-part memorandum for a material funding request, built around a proof plan and exit criteria written before the money is spent.
Word template, v0.5
Stopped Initiative Review
A blameless record of a stopped initiative: what it cost, why it ended, what is reusable and what would need to be true to try again.
Excel template, v0.5
AI Supplier Economics Review
Review a supplier contract on pricing risk, lock-in, consolidation exposure and governance fit, with a twelve-term checklist.
Tool
AI Proof of Value Scorecard
Score an AI pilot on eight dimensions, from baseline and feasibility to economics, risk and measured results.
Framework
Scale, Hold, Redesign, Stop: Decision Rules for AI Investment
A practical set of decision rules for a CFO or CIO reviewing AI workflows, combining return on scarce capacity with the evidence level behind each claim.
Framework
How Much Proof for How Much Autonomy
Autonomy should be earned with evidence. Each rung of an autonomy ladder paired with the evidence needed to climb it, scaled to impact and reversibility.
Framework
The AI Value Lifecycle: Seven Gates from Funding to Decision
When each AI value decision is made, who owns it and what evidence it needs: seven stages from funding to a decision, with a separate path for bought AI.
See it worked
Illustrative cases, one for each decision. Every figure is invented.
Illustrative case, Scale
Contract review assistant
A holdout comparison showed the quality floor held and most of the claimed value was booked.
Illustrative case, Hold
Workforce assistant
Strong adoption and self-reported savings, but no outcome baseline and nothing booked.
Illustrative case, Redesign
Contact-centre summarisation
Helps on simple contacts, but correction time on complex ones cancels most of the saving.
Illustrative case, Stop
Customer-service agent pilot
No owner, no baseline and a vendor-demo benchmark, with an agent able to act on accounts.
Related pages
Practical pages that sit alongside the templates.
The thinking behind it
The models and evidence that explain why this phase works as it does, in the Understand section.
Hub proposal, v0.5
The AI Abundance Paradox
When generation is cheap, review and absorption set the ceiling on value.
Essay
When to Stop an AI Initiative
Organisations stop AI work often, but by attrition rather than rule. Why underperforming initiatives persist, and how to set exit criteria in advance.
Case file
When AI Usage Outruns the Budget: What the Uber Story Teaches About AI Value Management
Uber spent its 2026 AI coding budget in four months and its COO could not show the value. What went wrong, and what AI value management would have changed.
Essay
Rent, Reserve or Own Intelligence?
A decision framework for API, committed capacity, neocloud, private and owned AI infrastructure.
Next phase
Each phase feeds the one after it.