Skip to content

Use-case canvas — <Customer>

A one-page scoping instrument for the pre-workshop phase of a partner engagement. Send this to the customer sponsor after the first scoping call; use the answers to decide whether to schedule the discovery workshop.

If the sponsor can't fill this in ~30 minutes of their time, the engagement isn't ready for a workshop. Iterate on framing first.

Engagement slug: <customer-slug> Sponsor: <name · role · email> Partner lead: <your name> Date: <YYYY-MM-DD> Status: Draft / Reviewed with sponsor / Approved for workshop

1. The process in one sentence

Write ONE sentence describing the current-state process this engagement will change. No "AI", no "agent" in this sentence. Example: "Our sales reps spend ~3 hours per account preparing a briefing before a discovery call."

Process: …

2. Who does this today, who pays for the change

  • Primary user persona (the role whose day the agent changes):
  • Volume (how many people × how often):
  • Budget owner (who signs the invoice):
  • Executive sponsor (who loses sleep if this fails):

3. The one metric that matters

Pick one primary metric the sponsor will judge the engagement on. Everything else is secondary. Resist the urge to list five.

  • Metric name:
  • Unit (minutes, $, %, count, …):
  • Baseline today (actual measurement — not a guess):
  • Target (what makes this worth doing):
  • Measurement source (where the number comes from: system of record, time study, sampled observation, …):

4. Must-not-do guardrails

Hard rules the agent must not violate. Compliance, policy, brand, or customer-promise constraints. Think: "If the agent did X, we would pull the plug tomorrow."

  • …
  • …
  • …

5. Data the agent can reach

What the agent can read from on day 1. Check all that apply and note the access mechanism.

  • SharePoint site(s): …
  • SQL / Synapse database(s): …
  • Internal APIs (REST/GraphQL): …
  • Blob / files: …
  • SaaS systems (CRM, ITSM, HRIS, …): …
  • Web search (bounded allow-list?): …
  • None (ask the sponsor why)

6. Side-effects the agent should perform

What the agent writes or sends on day 1. List each; every non-trivial side-effect will be a src/tools/*.py module with an HITL policy per accelerator.yaml:solution.hitl.

  • …
  • …

If the answer is "none" — i.e., the agent only reads and summarizes — that's fine; note the pattern as single-agent in Section 8.

7. Constraints

  • Data residency: US / EU / APAC / custom / no constraint
  • Identity provider: Entra ID / External ID / other
  • Compliance regime: SOC 2 / GDPR / HIPAA / PCI / none / other
  • Tenant topology: Customer tenant / partner-hosted / ISV-hosted
  • Azure AI Landing Zone tier (expected): standalone / avm / alz-integrated — see docs/patterns/azure-ai-landing-zone/
  • Hard deadlines: …

8. Expected solution shape (partner hypothesis)

The partner's guess before the workshop validates it. Three shapes ship in the template; the chatmode /discover-scenario will recommend one after the workshop.

  • supervisor-routing — multiple specialists + aggregator + HITL. Flagship default. Pick when the task has >1 distinct reasoning step.
  • single-agent — one agent + retrieval + 1–2 tools. Pick for narrow, repeated tasks (e.g., triage, summarization).
  • chat-with-actioning — conversational UX with tools. Pick when the human drives the workflow and the agent executes.

Rationale: …

9. Why we're confident this is worth a workshop

Two-sentence answer. If the answer is "the customer asked us to explore AI" — stop. Come back with a framing that ties to the metric in Section 3.

  • …

Go / no-go gate

Before scheduling the discovery workshop, confirm with the sponsor:

  • The process in Section 1 is the same thing everyone in the room thinks we're talking about.
  • Section 3 metric + baseline are real numbers, not placeholders.
  • Budget is identified.
  • The customer can bring the SMEs who do the work today — not just their managers — to the workshop.

If any checkbox is unchecked, do not schedule the workshop yet. Use the next scoping call to close the gaps.

Next step

Once the canvas is approved, schedule the workshop and open docs/discovery/SOLUTION-BRIEF-GUIDE.md to run the session. The workshop fills docs/discovery/discovery-workbook.csv live, then /discover-scenario turns the workbook into docs/discovery/solution-brief.md.