AI adoption advisor

Use Trace AI to run a practical AI-adoption engagement: reuse existing platform strategy, map as-is work, score opportunities, publish a small-steps roadmap, and only later (if you choose) draft Tealfabric automations. Executives and workshop participants use a logged-in engagement app, not Trace chat.

Who this is for

• Tenant advisors (operations, HR, transformation): tenant administrators who can open Trace AI.
• Consulting partners: the same playbooks, installed into the client tenant, with deliverables under engagements/{client}/.
• Executives and workshop contributors: platform users with role user. They sign in to the engagement app (/ai-adoption). They do not need Trace.

Trace AI is typically tenant-admin only. Do not expect HR or finance colleagues to chat with Trace.

Two steps (do not mix them)

Assessment is a hook so the work happens on Tealfabric. Building agents here is a later, explicit choice (your Sales motion, or the customer asking Trace to implement). Recommended work stays vendor-neutral (it could be built on AWS, Azure, Tealfabric, or in-house).

Install the Library pack

In Trace AI, ask to install the AI adoption playbooks from the model skills library. Trace loads load_platform_skill with skill_id: model-skill-import, creates or binds DataPool schemas, and writes tenant SKILLS/.

Install order:

1. ai-adoption-data-model (schemas)
2. ai-discovery
3. ai-opportunity-assessment
4. ai-roadmap
5. Optional: ai-operating-model, ai-hr-backoffice
6. Later, only if you will implement on Tealfabric: ai-tealfabric-bridge

Optional unattended chaser (draft): AI engagement followup watcher.

Partner path: when Trace asks for engagement_folder, bind engagements/{client_slug}/ instead of engagements/current/.

Flow

flowchart LR
  Strategy[Platform_strategy]
  Advisor[Advisor_in_Trace]
  Stakeholders[Users_on_engagement_app]
  Store[DataPool]
  Board[Board_pack]
  Bridge[ai_tealfabric_bridge]
  Drafts[Draft_platform_resources]

  Strategy -->|"get_platform"| Advisor
  Advisor -->|"discovery and scoring"| Store
  Stakeholders -->|"login JWT then intake"| Store
  Store --> Board
  Advisor -->|"only if asked"| Bridge
  Bridge --> Drafts

1. Reuse platform strategy

Trace loads platforms, lists business platforms, and reads get_platform strategy. Filled fields (value proposition, market, milestones, and any richer sections) are not asked again. New facts (process volume, pain, scale pressure) go to DataPool only.

If strategy is empty, decide whether to fill it in Platforms first or proceed with a thin engagement.

Do not use strategy-evaluate as the adoption scorer. That tool scores strategy prose quality.

2. Discovery (as-is)

Ask Trace to load tenant skill ai-discovery. In a workshop, Trace asks you structured questions; you relay answers from the room. Optional: Trace lists existing integrations and processes as context — it must not create automations.

Stakeholders who are not in the workshop can submit pain notes in the engagement app Intake page after login.

3. Score opportunities

Ask Trace to load ai-opportunity-assessment. Use cases use capability types (document_ai, conversational_agent, workflow_automation, decision_support, data_integration, rpa) — not product component names. Scores are 1–5 for value, feasibility, time-to-pilot, and risk. Confirm before Trace writes rows.

4. Roadmap and executive pack

Ask Trace to load ai-roadmap. Confirm 2–5 now items (weeks, not a long transformation program). Trace writes roadmap.md under the engagement folder and draws working-session Mermaid in chat.

Presenting to leadership: open the engagement app (public URI /ai-adoption after deploy). Sign in as a regular platform user (role: user). The Board pack page reads the same DataPool rows. Print or export PDF from the browser. Chat diagrams are not the board pack.

Authentication matches other tenant apps: product /api/v1/auth/login verifies email and password; the WebApp then issues a webapp_access JWT. The browser stores that JWT (Bearer), not a console type: access token. See WebApps Authentication.

5. Operating model (optional)

ai-operating-model writes a short operating-model.md (goals, human review, owners, cadence).

6. Implement on Tealfabric (optional bridge)

Only when you (or Sales) explicitly want the first experiment on this platform, ask Trace to load ai-tealfabric-bridge. Trace proposes drafts (for example workflow → ProcessFlow), you confirm, it creates draft resources only, and writes implementation_links. It does not enable executable_by_ai_agents or activate Autonomic Agents unless you ask.

Engagement app (JWT SPA)

The engagement UI is a Tealfabric WebApp that loads a React SPA from tenant CDN. It is forked from the platform skeleton app: login, session, and BFF router are already in place.

Deploy and CSP notes live with the app in the apps repository (tf-ai-adoption). Operators: auth_transport is bearer; auth_require stays false so login works; the router enforces session on protected actions.

Create user accounts for executives who should see the pack. tenant_admin is rejected at the app (same as the skeleton).

Data

Canonical tables: ai_engagements, as_is_processes, stakeholder_inputs, ai_use_cases, roadmap_items, implementation_links. See AI adoption data model.

See also

• Model skills library
• Trace AI Agent
• Tenant skills user guide
• WebApps Authentication
• Platforms (skill_id: platforms)