Capability&Consequence

Decision tools

Frameworks

Portable ways to inspect the obligations that conventional technology decisions leave out.

Each framework begins with a recurring analytical error: treating the artifact as the product, ownership as a single right, labour removal as pure gain, a model file as a durable asset, generation speed as productivity, or technical progress as realised value.

The standard

A framework should change the document in the room.

A useful framework does more than give a problem a memorable name. It changes what an investment memo measures, what an architecture review requires, what a contract preserves, or what a leader must fund. The test is whether another person can apply it to a new case and produce a decision that is more complete, more inspectable and harder to evade.

01

CAC-001 Read the coining essay

The Five Scarcities

When a plausible artifact becomes cheap, advantage moves to the capabilities around it that remain difficult to reproduce.

Decision

Use this framework to decide what a professional-services firm should protect, build or charge for when analysis and content production are increasingly automated.

The test

  1. Proprietary inputsData, benchmarks, case patterns and domain knowledge that a general-purpose model cannot readily reproduce.
  2. Reusable executionSoftware, agents and workflows that perform defined work repeatedly at low marginal cost.
  3. Client integrationThe engineering and operating work that connects a capability to systems, controls, incentives and roles.
  4. VerificationThe evidence, evaluation, traceability and expert review required before an output can be trusted.
  5. Accountable judgementA person or institution prepared to interpret, endorse and defend the conclusion.

Required output

A value map that identifies which scarcity carries the offer, which ones the firm genuinely controls, and where a generic artifact is being mistaken for an advantage.

02

CAC-002 Read the coining essay

The Four Rights

Owning an advisory product does not mean owning the business around it. Platform power is divided among four separable rights.

Decision

Use this framework before funding a platform, agent or productised offer to determine which rights the business model requires and how each will be retained.

The test

  1. The right to distributeWho controls where the client discovers, invokes, combines and substitutes the capability?
  2. The right to use client contextWho can reach the data, permissions, workflow state and decision history needed to produce a useful answer?
  3. The right to improveWho receives the corrections, edge cases, outcomes and usage evidence generated by deployment?
  4. The right to endorseWho is prepared to say the output is good enough for the decision and stand behind that judgement?

Required output

A rights-retention plan that names the owner, technical design and contractual mechanism for every right on which the economics depend.

03

CAC-003 Read the coining essay

The Capability-Depreciation Ledger

Automation can improve the current project margin while consuming the developmental experience from which future professional judgement is built.

Decision

Use this ledger whenever AI removes junior research, analysis, drafting, reconciliation, diligence or other work that produced both a deliverable and a practitioner.

The test

  1. Production removedName the work, hours removed, checking added and net effect on accepted output.
  2. Experience removedIdentify the source judgement, ambiguity, assumption testing, challenge or client exposure that disappears with the task.
  3. Replacement fundedSpecify how that experience will be recreated rather than assigning it vaguely to training.
  4. Senior timePrice the observation, challenge and teaching required, including staffing ratios and incentives.
  5. Evidence of capabilityMeasure demonstrated judgement in realistic work, not course completion or tool usage.

Required output

An automation business case that records capability consumed as explicitly as labour saved, with an owner and budget for replacing it.

04

CAC-004 Read the coining essay

The executable-stack serviceable-part lens

The durability unit for embedded AI is not the model file. It is the serviceable stack of model, runtime, libraries, hardware, evidence and replacement procedure.

Decision

Use this lens when an AI component sits inside a product or governed workflow whose support obligation is longer than the component lifecycle.

The test

  1. Fixed interfaceKeep the surrounding system from depending on every proprietary behaviour of the present model.
  2. Designed degradationSpecify and test what happens when the component is missing, unavailable, changed or replaced.
  3. Executable replacementName the owner, sequence, evidence, elapsed time and expected cost of a swap.
  4. External protective controlsKeep limits such as payment, torque, dosage or permission thresholds in deterministic controls that survive model change.

Required output

A serviceability specification with a reproducible stack, replacement trigger, tested fallback and measured time and cost to change the component.

05

CAC-005 Read the coining essay

The Verification Shadow

Generation creates an artifact. Value begins only after the organisation is willing to accept, use and stand behind it; the work between those moments is the verification shadow.

Decision

Use this framework to replace pilot claims about draft speed with the economics of accepted output across the full workflow.

The test

  1. Before-AI productionTime required to produce the artifact under the current process.
  2. Before-AI acceptanceTime required to review, reconcile and approve it today.
  3. AI generationOperator and system time required to create the proposed output.
  4. AI verificationTime required to test, correct, challenge and approve the generated output.
  5. ExceptionsThe rate of cases outside the happy path and the time required to resolve them.
  6. Post-acceptance reworkErrors or omissions discovered after approval, including their consequence and correction cost.

Required output

A before-and-after measure of accepted-output effort, separated by workflow consequence, with deterministic controls assigned wherever they are cheaper than human review.

06

CAC-006 Read the coining essay

The Three Clocks

AI business cases fail when they compress the speed of technical change, organisational absorption and financial conversion into one schedule.

Decision

Use this framework for investment approval so that a capability improvement is not counted as value before the operating model can absorb it and management can convert it.

The test

  1. Capability clockModel lifecycle, pricing, replacement triggers, portability assets and expected refresh cost.
  2. Operating clockIntegration, acceptance, exceptions, role changes, time to scale and revalidation procedure.
  3. Economic clockNet accepted-output improvement, conversion mechanism, accountable owner, recognition timing and useful-life assumption.

Required output

Three linked schedules in which every technical change creates dated operating work and every operating gain names the cash, revenue, risk or capital mechanism that converts it.