AI discovery

Map as-is processes and pain after reading existing platform strategy. Do not re-ask fields already filled on the business platform.

Kind

Sequential procedure.

Depends on

ai-adoption-data-model.

Prerequisites

Import contract

Procedure

Stages: bind_platform → read_strategy → interview → optional_scan → write_as_is.

DoD: at least three as_is_processes rows (or the advisor confirmed a smaller set) and a strategy recap written under the engagement folder. Workshop participants are not in Trace chat; Trace asks the advisor structured questions while they are in the room.

Do not call strategy-evaluate. That scores strategy prose, not process pain.

Tenant SKILL.md (copy this body)

# AI discovery

Load `platforms`, then `datapool`. Bound schemas: `{{engagements_schema}}`, `{{as_is_processes_schema}}`, `{{stakeholder_inputs_schema}}`. Markdown under `{{engagement_folder}}`.

## Sequence

1. `list_platforms` then `get_platform`. Summarize filled `strategy` sections (typically `business_model`, `market_analysis`, `implementation`; also any richer keys present). Bind `platform_id` on the engagement.
2. Do **not** re-ask `value_proposition`, `target_market`, `revenue_streams`, or other filled strategy fields. Cite them.
3. If strategy is empty, `request_human_input`: fill strategy first (Platforms UI / `update_platform_strategy`) **or** proceed thin. New facts go only to DataPool.
4. Interview for as-is inventory: name, owner, volume, systems, pain, scale pressure (growth vs headcount). Workshop mode: ask the **advisor** via `request_human_input` (`kind: text` or `choice`); they relay stakeholder answers.
5. Optional context scan: `list_integrations`, `list_processes`. Use results to avoid re-asking “which systems exist”. Do not create processes.
6. Confirm, then `insert_datapool_rows` into `{{as_is_processes_schema}}`. Optional `{{stakeholder_inputs_schema}}` with `source=interview` or `workshop`.
7. Set engagement `status=discovery`. Write `{{engagement_folder}}/strategy-recap.md` (citations, not duplicated strategy fields).

## Stop

- Do not invent processes or systems the advisor did not confirm.
- Do not write `implementation_links`.
- Do not create ProcessFlow, WebApps, or Autonomic Agents.
- Do not call `strategy-evaluate`.

Optional skill.json

{
  "name": "ai-discovery",
  "version": "1.0.0",
  "description": "Read platform strategy first, then map as-is processes and pain into DataPool.",
  "capabilities": {
    "prompt": [
      "Read get_platform strategy before interviewing. Do not re-ask filled strategy fields.",
      "Do not create ProcessFlow or Autonomic Agents."
    ],
    "resources": [],
    "tools": [
      "list_platforms",
      "get_platform",
      "describe_datapool_table",
      "query_datapool",
      "insert_datapool_rows",
      "update_datapool_rows",
      "list_integrations",
      "list_processes",
      "request_human_input",
      "write_file"
    ]
  }
}

See also

• AI adoption data model
• AI opportunity assessment
• AI adoption advisor (user guide)