Skip to content
Understand

Decision cases

Decision stories, not success stories

Most AI case studies describe wins. These end in a decision. Each follows the same chain: cost, consumption, outcome, evidence, decision. Together they cover all four outcomes: scale, hold, redesign and stop.

All four cases are illustrative. Every organisation and figure is invented, to show the mechanics. They are rows of the sample portfolio in the portfolio review pack, so the numbers reconcile across the site.

Portfolio review

Claimed is not realised.

  • Contract review assistant

    BoardingDecision: Scale

    Figures: Contract review assistant
    Annual cost
    £75k
    Claimed
    £160k
    Realised
    £120k
    Evidence
    E3
    Next review
    31 March 2027
  • Workforce assistant

    DelayedDecision: Hold

    Figures: Workforce assistant
    Annual cost
    £420k
    Claimed
    £300k
    Realised
    £0k
    Evidence
    E1
    Next review
    10 December 2026
  • Coding assistant

    DelayedDecision: Hold

    Figures: Coding assistant
    Annual cost
    £98k
    Claimed
    £108k
    Realised
    £26k
    Evidence
    E2
    Next review
    15 January 2027
  • Contact-centre summarisation

    Gate changeDecision: Redesign

    Figures: Contact-centre summarisation
    Annual cost
    £180k
    Claimed
    £240k
    Realised
    £60k
    Evidence
    E2
    Next review
    31 January 2027
  • Customer-service agent pilot

    CancelledDecision: Stop

    Figures: Customer-service agent pilot
    Annual cost
    £60k
    Claimed
    £90k
    Realised
    £0k
    Evidence
    E0
    Next review
    Overdue since 30 September 2026

Read the cases

  • Illustrative · LegalScale: Contract review assistant

    A holdout comparison showed the quality floor held and most of the claimed value was booked.

    Evidence E3 · Next review: 31 March 2027

  • Illustrative · All functionsHold: Workforce assistant

    Strong adoption and self-reported savings, but no outcome baseline and nothing booked.

    Evidence E1 · Next review: 10 December 2026

  • Illustrative · Customer operationsRedesign: Contact-centre summarisation

    Helps on simple contacts, but correction time on complex ones cancels most of the saving.

    Evidence E2 · Next review: 31 January 2027

  • Illustrative · Customer operationsStop: Customer-service agent pilot

    No owner, no baseline and a vendor-demo benchmark, with an agent able to act on accounts.

    Evidence E0 · Next review: Overdue since 30 September 2026

How the cases are built

Every case has the same parts: executive summary, situation, economic baseline, full cost, consumption and workflow, outcome and quality floor, evidence assessment, net position, decision, and lessons with limitations. Each shows realised value separately from claimed value, and ends with a decision record in the format of the AI Value Ledger.

Case status labels

Illustrative (in use):
Synthetic figures used to explain the mechanics. No real organisation.
Representative:
Built from common programme patterns, not from one organisation's data.
Validated:
Based on programme data that has been reviewed.
Organisation-supplied:
Provided and approved by the named organisation.
Benchmark case:
Suitable for repeated comparison, with clear scope and limitations.

Only illustrative cases are published today. A case is not relabelled representative or validated until it rests on reviewed data.