Objectives
- Deliver an onboarding wizard that maps Sahil Bloom’s wealth pillars to the AI coaches the student selects (defaulting to all personas per area).
- Stand up the TickTick (Tic-Tac MCP) integration so action items can be synchronized, tagged, and monitored.
- Persist the full AI coach matrix, wealth areas, and to-do list associations in the database for downstream phases.
- Provide administrative controls for managing AI coach personas, embeddings, and integration credentials.
Functional Scope
Wealth Area & AI Coach Selection
- Build a multi-step onboarding wizard that educates users on Time, Finance, Relationship, Career, and optional Personal Growth wealth types.
- Present available AI coaches (Atomic Habits, The Magic of Thinking Big, Richard Feynman, Elon Musk) with persona summaries and allow multi-select per wealth area.
- Support default assignments (all AI personas) with the option to reorder or remove personas per area.
- Store selections in
user_coach_assignments
with priority ordering and metadata (e.g., rationale notes from the student) to drive AI routing. - Provide edit controls in the
/coach/home
dashboard so users can revisit AI persona choices.
TickTick / Tic-Tac MCP Setup
- Implement OAuth login with TickTick (or configure Tic-Tac MCP API credentials) and secure token storage.
- Fetch available TickTick lists and allow the user to map or auto-create one list per wealth area.
- Configure default tags for each AI coach persona and wealth area combination to support analytics.
- Build backend service for task CRUD that conforms to the MCP tool contract (list tasks, create, update, complete, apply tags).
- Document fallback plan if TickTick is unavailable (local queue, manual completion logging).
Data Model & Services
- Extend database with tables:
coach_profiles
(AI persona metadata),wealth_areas
,user_coach_assignments
,ticktick_accounts
,ticktick_lists
,action_items
(skeleton for future phases). - Seed canonical AI coach data (persona prompts, source material references, knowledge embedding pointers).
- Provide admin UI to edit AI coach metadata, toggle availability, and upload knowledge sources for future retrieval.
- Add service layer modules for AI coach assignment CRUD, wealth area queries, and TickTick API interactions.
Experience Glue & Analytics
- Update personalization dashboard with a summary of selected AI coaches per wealth area and TickTick linkage status.
- Emit analytics events when users change AI coach selections or connect/disconnect TickTick.
- Add health indicators for integration status (last sync time, outstanding errors).
- Create migration scripts to backfill wealth area data for existing Phase 1 users.
Technical Considerations
- Handle AI coach selection concurrency: optimistic UI with background validation, server resolves conflicting updates.
- Encrypt TickTick access/refresh tokens using KMS or libsodium before storage.
- Normalize tags for TickTick tasks (e.g.,
ai_coach:atomic-habits
,wealth:time
) and document taxonomy for analytics consumers. - Use MCP tool schema definitions stored in
agents/contracts
so multi-agent workers can reuse the integration reliably. - Rate limit TickTick API calls and implement exponential backoff with jitter for retries.
Multi-Agent Workstream
| Agent | Responsibilities | Deliverables | | --- | --- | --- | | AI Coach Matrix Architect | Design onboarding wizard UX, AI persona selection logic, and persistence APIs. | Wizard flows, selection components, API schemas. | | Integration Engineer | Build TickTick OAuth, MCP toolset, and synchronization services. | OAuth routes, MCP tool definitions, sync jobs. | | Knowledge Curator | Seed AI coach personas, maintain embeddings, and author persona prompts. | AI coach dataset, embedding pipeline docs. | | Database Engineer | Implement migrations for wealth areas, assignments, TickTick tables, and seeding scripts. | SQL migrations, seed CLI. | | Analytics Lead | Instrument events and dashboards for AI coach selection and integration health. | Event catalog, dashboard configs. | | DevOps Steward | Manage secrets rotation, webhook endpoints, and monitoring alerts. | Secret management plan, uptime dashboards. |
Exit Criteria
- Users can select and save AI coaches per wealth area during onboarding and edit them later.
- TickTick accounts connect successfully, with at least one list mapped to each wealth area and verified via API call.
- MCP toolset exposes
listTasks
,createTask
,updateTask
, andcompleteTask
functions with schema validation and tests. - Admins can view and update AI coach metadata through a secured interface or CLI.
- Analytics dashboard highlights AI coach selection distribution and TickTick integration health metrics.
Risks & Mitigations
| Risk | Mitigation | | --- | --- | | Users overwhelmed by AI coach persona choices. | Provide recommended defaults, highlight most popular personas, allow “decide later” option with reminders. | | TickTick OAuth complexity or rate limits. | Cache metadata, defer heavy sync tasks to background jobs, communicate sync status clearly to users. | | Coach data drift (persona updates). | Version persona prompts, run regression tests on knowledge responses, and require review before publishing changes. | | Data consistency across wealth areas and TickTick lists. | Enforce constraints ensuring each area has exactly one default list; add nightly reconciliation job. |
Dependencies & Notes
- Requires Phase 1 tables (
users
,favorites
, base analytics) and environment scaffolding to be in place. - Coordinate with legal/compliance on storing TickTick tokens and presenting consent copy.
- Prepare support documentation so users understand why TickTick access is needed and how data is used.
- Share AI coach selection data schema with Phase 3 team to seed session context retrieval.