Skip to main content

AI Transformation Case Method

How do you turn a spec into a case for change?

Start with the business. End with a decision. AI is the redesign lens, not the subject.

Outcome​

Produce a confidence case that makes one reader able to approve, reject, or investigate one bounded AI transformation bet. The work is rhetoric in service of wisdom: public judgment under pressure, aimed at a better decision.

Inputs​

InputPurpose
Business specNames the company, model, goals, constraints, and known pain.
Company evidenceGrounds the case in live public and private facts.
Workflow inventoryNames what work is real, artifact, or hybrid.
Market signalsShows why the window is open now.
Existing playbookProvides reusable method, not client-specific proof.

Steps​

  1. Internalize the spec. Name the reader, the business model, the decision, and the first workflow candidate.
  2. Research reality. Gather company evidence, tech/data flow, customer signals, competitor signals, and operational constraints.
  3. Develop the rhetoric. Use Ethos, Logos, Pathos, Topos, and Kairos to pressure the claim, objections, proof path, timing, and shared frame before writing report copy.
  4. Backcast the bridge. Define the 24-month future state, 12-month milestone, 90-day proof, and day-14 prerequisites.
  5. Assemble the report. Write present-day reality, desired future, bridge with confidence, evidence, rhetoric, and decision as one chain.
  6. Gate the case. Every major claim must trace to the evidence ledger or argument map.
  7. Close the loop. After dogfood, update the template or module that produced the weak output. Do not patch only the rendered page.

Artifact Build Matrix​

Use this practice when the case needs a set of decision artifacts, not one report. Each row must trace from a counterparty's demand to the model that could carry its state. Unsupported mappings stay UNKNOWN, N/A, or GAP. A similar name is not evidence of capability.

  1. Derive the demand story. Read the five ecosystem counterparties and the case's local Dream stories. Name the participant, job story, decision, desired behaviour, and success signal. biz-dev-ai-transformation owns this judgment.
  2. Select the role and mindset. Choose the smallest blend from the Archetypes hub. Archetypes are modes a person or agent can enter, not fixed identities. Record why the blend fits and its shadow or kill signal. drmg-thinking-router owns the selection.
  3. Select the toolkit, skill matrix, and TACO mode. Choose zero or one primary creation tool from the AI Toolkit. Classify required skill across modality I/O, industry expertise, business function, and tool execution. Link one active owning skill. Assign one TACO mode: Tasker, Automater, Collaborator, or Orchestrator. agency-power-tools owns the route.
  4. Map the Stackmates domain model. Run a fresh sm-capability-map read. Name the bounded domain module and the entity, value object, domain service, port, or aggregate that supports the artifact. Record REALITY, DREAM, or GAP, the code evidence path, missing semantics, and the next Dream question. sm-capability-map owns the evidence read.
  5. Build, score, and commission. Build with the selected skill and tool contract. Check canonical blueprint coverage, evidence state, decision, gauge, lever, kill signal, and next action. Populate the scorecard and commission at L2. L3 is blocked until the served person uses the artifact and their feedback is recorded. biz-dev-ai-transformation owns the gate.

Run the five composable prompt cards in that order. The conductor binds the sequence; each card remains a cold-start input contract and does not repeat its owning skill's procedure.

Rhetoric As Wisdom Work​

Rhetoric is not polish. It is the skill that turns private belief into public judgment another mind can test. It develops wisdom because it forces the case to answer when, where, with whom, and under what proof condition the next action is right.

Use the belief articles as the philosophical spine:

  • Rhetoric — public judgment under pressure: Ethos, Logos, Pathos, Topos, and Kairos.
  • Wisdom — knowing when, where, and with whom to act.

Beyond AI Transformation​

The method is not specific to AI, or to businesses. AI is one redesign lens; the deeper engine is Reality → Dream → Bridge → Decision backed by a source-tagged evidence ledger. The same spine builds a confidence case for any high-stakes decision where a person can pay but is held back by dis-ease — a major purchase, a relocation, a medical or legal choice, a high-stakes holiday booking under uncertainty.

What changes when the subject is a decision rather than a transformation:

  • Binding constraint. It may be confidence / dis-ease rather than leverage or distribution — the risk is losing money and still feeling unwise, or spending months of peace.
  • Artifacts rename, roles hold. transformation-plan becomes a decision-plan (Go / Wait / Don't, with payment gates and kill signals); one-page-commercial-plan becomes a one-page decision. The evidence ledger, objection map, and confidence case are unchanged.
  • The value exchange is the same shape. The guide's worth is intelligence arbitrage — seeing the clauses, routes, and exclusions the decider cannot — priced against avoided loss/regret, not hours. Both sides reading the same evidence makes the price fair.

The test still holds: a cold reader should be able to name the decision, the acceptable-loss line, the kill signal, the support chain, and the missing evidence in under 90 seconds.

Outputs​

OutputDone When
evidence-ledger.mdEvery major claim has source, date, status, and confidence tag.
argument-map.mdThe thesis, objections, counters, and reader action are explicit.
confidence-case.mdReality-Dream-Bridge is written as a decision argument.
_case-manifest.mdInput, research, rhetoric, presentation, and privacy status are mapped.

Checks​

  • The case has one named reader.
  • The first workflow is named.
  • The 90-day proof has a metric, owner, budget, and kill signal.
  • Each major claim is tagged VERIFIED, INFERRED, ASSUMPTION, or CONTRADICTED.
  • The report contains no fresh claim that is absent from the evidence ledger or argument map.
  • Private client proof is not linked from public discovery surfaces without explicit consent.

Failure Modes​

  • AI-first framing. The report sells tools instead of business change.
  • Research without rhetoric. Sources pile up but no belief changes and no decision becomes easier.
  • Optimistic bridge. The plan has no kill signal, owner, or early proof.
  • Output patching. A page improves once, then regresses because the module was not changed.

Context​

Questions​

  • What does the reader need to decide after reading?
  • Which first workflow can prove the whole thesis without betting the whole business?