Patterns applied to the B2B case¶
Stopped demonstration. The case uses simulated data and remains instrumented-not-executed. Conversation may structure review, but cannot turn absent owner, action, expectation or outcome into facts.
The MVP source concludes that a conversational interface would be decorative for one cycle. This page applies patterns as a meeting or guided-review protocol. A checklist, brief or competent facilitator is currently simpler and sufficient.
Entry state¶
| Element | Available evidence | Status |
|---|---|---|
| Priority | Recover compliance after April–June demo deterioration | Provisional demonstration only |
| Decision | Where to intervene first without unnecessary acquisition expansion | Formulated |
| Owner | Proposed Sales Director | Unconfirmed |
| Action | Depends on selected option | Not executed |
| Metric | Proposal conversion and compliance | Simulated baseline available |
| Risk | Regression to mean, confounding, small samples, overload | Identified |
| Outcome | None after intervention | Not observed |
Move 1 — Decision contract¶
The facilitator asks whether to intervene in Proposal, reallocate channels, combine actions or wait, then asks who has the right to choose and by when. Options A–D exist and A is provisionally preferred for proximity and reversibility. The owner remains to be confirmed, so CDP-001 exposes a blocker and does not authorize commitment.
Move 2 — Evidence boundary¶
Keep separate: the source states simulated data; June compliance is 80.11% and Proposal falls 11.50 percentage points between observed quarters; deterioration appears more concentrated in conversion than volume; intervention might improve conversion; real operations, owner, action, expectation, comparability and outcome remain unknown. Event and referral have higher observed rates than digital inbound, but wide overlapping intervals do not establish robust superiority or authorize budget reallocation.
Move 3 — Option challenge¶
| Option | Best case | Strong objection | What changes status |
|---|---|---|---|
| A · Improve Proposal | Proximate signal, reversible action | Pricing or mix may explain decline | Operational diagnosis, owner, capacity |
| B · Reallocate channels | Higher observed event/referral rates | Small sample, inconclusive evidence | More observations, comparison, incremental cost |
| C · A + B | Addresses conversion and origin | Attribution becomes impossible | Factorial design or component measurement |
| D · No action | Avoids reacting to anomaly | Delay may deepen the gap | Seasonality evidence or new time cut |
A remains a preference for a first test, not automatic execution. B and C stay blocked; D is an active comparator.
Move 4 — Decision–action handoff¶
Stop conditions remain: no confirmed owner, formal choice, action owner/date, frozen expectation or assigned real scope/comparator. The only valid exit is: Prepared for owner review; not authorized to execute. “Execute A” would trigger ANTI-CDI-004.
Move 5 — Expectation–result review¶
CDP-005 does not apply. No frozen expectation, executed action, closed horizon or outcome exists. A learning narrative would manufacture evidence and trigger ANTI-CDI-006. Next: confirm owner and rights; select or reject an option; register action and scope; freeze range and horizon; define comparison, guardrails and stops; approve and execute; then observe the agreed horizon.
What it demonstrates—and does not¶
Within the artifact, patterns can reveal omissions, preserve evidence classes and stop before fabricated commitment or learning. It does not show that conversation improves the decision, A works, PULSE creates commercial impact or the catalog beats competent facilitation.
Potentially superior alternative¶
For the current state, completing the decision contract with the owner is more defensible than building a copilot. Consider conversation only after recurring cycles reveal access or deliberation friction that the manual protocol cannot solve.