Skip to content

Solution Brief —

STATUS: TEMPLATE. This is the unmodified template shipped with the accelerator. Fill it via /discover-scenario (or /ingest-prd then /discover-scenario for gap-fill). The chatmodes overwrite this banner on completion; lint exempts the brief while the banner is present.

This document is the canonical source of truth for this engagement. Every downstream artifact (agent prompts, tools, retrieval, HITL, evals, telemetry, manifest) derives from it. Fill it with the customer during discovery (/discover-scenario), then run /scaffold-from-brief to apply it across the repo.

Engagement: <customer-slug> Status: Draft / Reviewed / Approved Last updated: <YYYY-MM-DD>

1. Business context

  • Industry / segment:
  • Customer name + size:
  • Executive sponsor:
  • Decision maker:
  • Problem statement (one sentence):
  • In scope:
  • …
  • Out of scope (explicit):
  • …

2. Target users & journeys

  • Primary persona: (role · daily workflow · tools they live in)
  • Secondary persona (if any):
  • Top 3 user journeys:
  • …
  • …
  • …

3. Success criteria (measurable)

Metric Current Target How measured
Time per task
Volume / throughput
Quality bar

Must-not-do guardrails: - …

4. ROI hypothesis

  • Baseline cost of status quo: (FTE × rate, or $/transaction × volume)
  • Target savings ($/yr):
  • Payback target:
  • KPIs to instrument (these become telemetry events in src/accelerator_baseline/telemetry.py and charts in infra/dashboards/roi-kpis.json):
KPI event name Type Baseline Target
e.g. review_completion_time_ms duration_ms
e.g. escalation_rate ratio

5. Solution shape

  • Pattern: supervisor-routing / single-agent / chat-with-actioning
  • Rationale:
  • Grounding sources: SharePoint / SQL / APIs / blob / other
  • Side-effect tools (list every one):
Tool name External system Operation Reversible? HITL policy
  • Out-of-scope tools (explicit):

5b. UX shape

  • ux_shape: TBD — pick one of:
  • Structured form + report — user fills a form, agent produces a structured briefing/analysis. Next step: fork patterns/sales-research-frontend/ as your starter and adapt the form to your scenario's request schema.
  • Chat — multi-turn conversational UX. No chat UI pattern shipped yet; the chat-with-actioning backend pattern supports this shape — you build the UI on top (or use any chat UI framework).
  • Dashboard / viewer — agent output renders in the customer's existing app. No UI pattern needed from the accelerator — consume the SSE endpoint directly from the customer's app.
  • API-only / embed — another system calls the agent programmatically (Power Automate, n8n, partner platform). No UI. The accelerator's hosted SSE endpoint IS the deliverable.
  • Rationale (one line):

5c. UX inputs

Filled only when ux_shape is Structured form + report. Each row should match a field in src/scenarios/<pkg>/schema.py (ScenarioRequest).

Field Type Description Required
TBD text TBD yes

5d. UX output sections

Filled only when ux_shape is Structured form + report. Each section is a panel rendered by the result UI; Source agent is the worker whose output populates it (or supervisor if composed).

Section Content Source agent
TBD TBD TBD

6. Constraints & risks

  • Data residency: US / EU / APAC / custom
  • Identity: Entra ID / External ID / custom
  • Compliance regime: SOC 2 / GDPR / HIPAA / PCI / none / custom
  • RAI risks (3–5 specific to this scenario; become redteam cases):
  • …
  • …
  • …

7. Acceptance evals

Gate Threshold Wired to
Quality (golden agreement) evals/quality/golden_cases.jsonl
Groundedness evals/quality/
Safety (redteam) must pass evals/redteam/
Latency (P50 ms / P95 ms) / App Insights
Cost per call ($) cost attribution telemetry

Next step: run /scaffold-from-brief in Copilot Chat to apply this brief across src/, infra/, evals/, accelerator.yaml, and infra/dashboards/.