Product Mapping: From Current to Recovered State

Tier C — Historical. This document is a point-in-time snapshot preserved for reference. For current architecture, see CANONICAL_PLATFORM_ARCHITECTURE.md.

Understanding the Drift

Dimension	Jan 2025 (Strong)	Current (Diluted)	Fixed (Recovered)
Center of Gravity	Human Workbench	AI Agents	Human Workbench
Primary Product	Adaptive workspace	MCP/ADK infrastructure	Continuity surface
What customers buy	"One place to work"	"AI orchestration"	"Nothing lost in context"
Customer question	"How do I stop context switching?"	"How do I build AI agents?"	"How do I stop being a human API?"
AI role	Support layer	Primary actor	Invisible infrastructure
User spends time in	Work surfaces (8h/day)	AI chat/approval (30min/day)	Work surfaces (8h/day)
Moat	Continuity + Entity360	Multi-model routing	Continuity + Spine

---

Current State Diagnosis

What Exists (Good)

• ✅ Backend Contract Solid

  • Auth with password enabled
  • Gateway routing correct
  • Twin model route operational
  • Knowledge service + queue + cron

• ✅ Real Customer Zero Infrastructure

  • Tenant: iw-customer-zero
  • User: Nirmal Prince
  • Verification component wired

• ✅ Frontend Foundation

  • Shell and routing architecture
  • L1 domain surfaces (dashboard, contacts, companies, etc.)
  • L2 overlay system for collaboration
  • Real Customer Zero integration

What's Wrong (Problems)

• ❌ Wrong Focus

  • AI treated as primary product
  • Surfaces treated as overlays
  • Continuity not positioned as moat
  • Humans treated as approval layer

• ❌ Wrong Architecture

  • AI on top, humans below
  • Enables "agent autonomy" but kills adoption
  • Surfaces not configurable
  • No clear workbench primacy

• ❌ Wrong Value Prop

  • "Here's another AI" ← resonates poorly
  • "Solve your fragmentation" ← resonates well but lost

---

Mapping: Current Components to Recovered Structure

Level 1: Human Workbench (What exists, needs focus)

Current State:

• Dashboard (exists but not primary)
• Contacts/Companies (buried in UI)
• Tasks/Calendar (secondary to AI)
• Communications (viewed through AI)

Recovered State:

• Dashboard → Primary entry point (always first)
• Accounts → Your profile, preferences
• Contacts → People you work with
• Companies → Customers, prospects, partners
• Opportunities → Deals, pipeline
• Tasks → Personal and team work
• Calendar → Meetings, availability
• Notes → Quick capture
• Meetings → Recording, summary, actions
• Communications → Email, messages, Slack history
• Documents → Files, shared, versions
• Knowledge → FAQs, processes, learnings
• Goals → Quarterly, tracking
• Values → Principles, culture
• Reports → Custom dashboards
• Settings → Integrations, preferences

What Changes:

• Surfaces moved from "overlay" to "primary"
• Left sidebar shows all surfaces (configurable)
• User chooses where to work
• AI recommendation appears as side panel, not center
• "Stop Being Human APIs" becomes visible again

Level 2: AI Workbench (Needs to be built separately)

Current State:

• Twin integrated into overlay
• No persistent home for AI
• No daily routines
• Memory not persistent

Recovered State:

• AI's Own Environment (separate from human surfaces)

  • Observations dashboard (what Twin sees today)
  • Memory consolidation (learnings from sessions)
  • Daily routines (morning standup, EOD summary)
  • Recommendation queue (prepared options)
  • Approval queue (waiting human decision)
  • Collaboration history (past conversations)

• AI Visibility (human can see Twin's work)

  • "Twin saw X today. Should we Y?"
  • "Your patterns: You code 6-9pm. Block calendar?"
  • "Team insight: 5 PRs waiting >2 days. Should we review?"
  • "Follow-up: Customer complained Monday. Haven't heard back?"

• AI Autonomy (Twin works continuously)

  • Morning: Reviews overnight emails, flags urgent
  • Hourly: Monitors for anomalies
  • Evening: Consolidates learning
  • Weekly: Generates insights

What Changes:

• AI not idle (has real responsibilities)
• AI has persistent memory (learns over time)
• AI has boundaries (can't override human work)
• AI has visibility (human always sees what it's doing)

Level 3: Continuity Engine (Always IntegrateWise)

Current State:

• Spine exists but not featured
• Memory not front-and-center
• Auth hidden
• Governance invisible

Recovered State:

• Identity (Always IW)

  • Authentication (not delegated)
  • Authorization (IntegrateWise governs)
  • Organizations, Teams, Roles
  • Permissions, Boundaries
  • Governance policies

• Spine Schema (Always IW)

  • Canonical truth for all entities
  • Relationship definitions
  • Memory scoping rules
  • Continuity rules
  • Lineage tracking

• Shared Memory (Always IW)

  • Org Memory (principles, goals, rules)
  • Work Memory (current projects, priorities)
  • Personal Memory (user preferences, patterns)
  • AI Memory (learnings, continuity facts)

• Invisible Continuity

  • Context follows the user
  • Nothing is forgotten
  • Connections are hidden
  • Intelligence is obvious

What Changes:

• Continuity becomes the moat (not MCP or agents)
• Spine defines what connections mean (not just endpoints)
• Memory scopes defined by governance (not data access)
• Users experience continuity but don't think about it

Level 4: AI Infrastructure (Invisible)

Current State:

• MCP highlighted
• ADK discussed frequently
• Multi-model routing complex
• Reflection loops debated

Recovered State:

• Still exists, but hidden

  • MCP routers (how tools communicate)
  • ADK discovery (how Twin finds capabilities)
  • Multi-model orchestration (Claude → GPT → Gemini)
  • Reflection & consolidation (memory promotion)
  • Voice & multimodal (speech input/output)

• Users don't need to know about:

  • Which model is processing
  • How MCP works
  • ADK routing details
  • Reflection mechanisms
  • These are implementation details

What Changes:

• Infrastructure is means, not end
• Product story doesn't mention MCP
• AI model choice is transparent to user
• Continuity is what users experience, not infrastructure

Level 5: Connected Applications (100+)

Current State:

• Apps treated as data sources
• Continuity applied post-connection
• User doesn't see connection logic

Recovered State:

• Apps remain external

  • Gmail (email, calendar)
  • Slack (messages)
  • GitHub (projects, code)
  • Figma (designs)
  • Notion (docs)
  • Salesforce (CRM)
  • HubSpot (leads)
  • Linear (tasks)
  • Drive (files)
  • Calendar (availability)
  • WhatsApp (personal)
  • and 90+ more

• Connected through Spine

  • Not just synced
  • Relationships mapped
  • Entities normalized
  • Memory extracted

What Changes:

• Users see fewer duplicate entities
• Connections are federated (not replicated)
• Continuity flows across all apps
• No context lost between tools

---

Configuration Architecture

Surface Registry (Built)

Each surface is independently functional:

interface Surface {
  id: string                    // 'dashboard', 'contacts', 'tasks'
  label: string                 // 'Dashboard', 'Contacts', 'Tasks'
  icon: string                  // 'home', 'users', 'checkbox'
  component: React.ComponentType
  enabled: boolean              // Can user disable?
  requiresTools?: string[]      // Which tools it depends on
  order: number                 // Default position in sidebar
  pinned?: boolean              // User's preference
}

Sidebar Configuration (User-Driven)

interface SidebarConfig {
  order: string[]               // ['dashboard', 'accounts', 'contacts', ...]
  hidden: Set<string>           // User hidden surfaces
  pinned: string[]              // User pinned favorites
  collapsed: Set<string>        // Collapsed sections
  theme: 'default' | 'compact'
}

// User actions:
useSurfaceConfig()
  .toggleSurface('notes')       // Hide/show
  .reorderSurfaces([...])       // Drag-drop
  .pin('opportunities')         // Pin to top
  .reset()                      // Back to factory

Per-Role Configuration (Admin)

interface RoleConfig {
  surfaces: {
    [surfaceId]: {
      visible: boolean          // Can they see it?
      editable: boolean         // Can they edit?
      connectedTools: string[]  // Which integrations show?
    }
  }
}

// Example: Sales role
{
  surfaces: {
    dashboard: { visible: true, editable: false },
    accounts: { visible: true, editable: true },
    contacts: { visible: true, editable: true },
    companies: { visible: true, editable: true },
    opportunities: { visible: true, editable: true },
    tasks: { visible: true, editable: true },
    calendar: { visible: true, editable: false },
    goals: { visible: false, editable: false },  // Hidden for sales
    values: { visible: false, editable: false },  // Hidden for sales
  }
}

---

Implementation Checklist: Current → Recovered

Phase 0: Cleanup (TODAY)

• Remove components/l1/entity-stretch.tsx
• Remove components/l1/domains/customer-zero/
• Flag /pages/Metrics.tsx for Replit reconciliation
• Verify tsc --noEmit clean

Phase 1: Deploy (THIS WEEK)

• Deploy cleaned workbench to preview
• Verify all backend services online
• Confirm Customer Zero can access
• Test each L1 surface (dashboard, contacts, companies, etc.)

Phase 2: Activate (WEEK 1)

• Customer Zero starts daily usage
• First tool connection (Gmail)
• Second tool connection (Slack)
• Third tool connection (GitHub)
• Fourth tool connection (Calendar)
• Spine ingests first entities

Phase 3: Schema (WEEK 2)

• Document Spine Schema v1.0
• Map entities (Person, Org, Work, Communication)
• Map relationships (50+ types)
• Map memory scopes (Org, Work, Personal, AI)
• Governance rules (access, boundaries)

Phase 4: AI Home (WEEK 2-3)

• Build AI Workbench separately
• Implement daily routines
• Connect memory consolidation
• Enable human-AI collaboration

Phase 5: Surfaces (WEEK 3)

• Implement Surface Registry
• Build Sidebar Configuration UI
• Enable per-user customization
• Support drag-drop reordering
• Add hide/show functionality

Phase 6: Integration (ONGOING)

• Customer Zero connects 10+ tools
• Spine handles duplicates/conflicts
• Memory consolidation working
• Human-AI working together
• Continuity invisible but omnipresent

---

What NOT to Do

❌ Don't rebuild the frontend from scratch

• Existing foundation is solid
• Just shift focus from AI to human

❌ Don't move AI to the center

• Keep it UNDERNEATH as infrastructure
• Surface it only in collaboration overlay

❌ Don't hide the workbench surfaces

• Make them PRIMARY and prominent
• Left sidebar is the main navigation

❌ Don't make Spine visible

• Continuity is the moat
• Users shouldn't think about it

❌ Don't build AI autonomy first

• Build human workbench first
• AI gets its own environment separately

---

What TO Do

✅ Deploy the existing workbench (cleanup only)
✅ Let Customer Zero work naturally
✅ Let Spine evolve from real usage
✅ Build AI home separately and persistently
✅ Focus on invisible continuity, not AI visibility

---

Success = Back to Original Vision

Original (Jan 2025):

"Stop Being Human APIs. All your work appears in one adaptive workspace. Every conversation, ticket, customer, document and action stays connected. Nothing is lost when switching tools."

Today (Current):

"Use AI agents to orchestrate your tools. Create MCP flows. Configure reflection loops."

Recovered (June 2026+):

"All your work in one place. Your Twin understands everything. Nothing is forgotten. No context switching. One unified surface."

---

Timeline Summary

TODAY:
  Cleanup files, deploy workbench

WEEK 1:
  Customer Zero uses, connects tools

WEEK 2:
  Spine Schema documented, AI ready

WEEK 3+:
  Continuity invisible, productivity obvious

END STATE:
  Product focused.
  Humans supported by AI.
  AI infrastructure hidden.
  Continuity omnipresent.
  "Stop Being Human APIs" vindicated.

---

Status: Ready to Implement

The mapping is clear. The infrastructure exists. All that remains is execution.
First step: Clean files → Deploy → Activate Customer Zero → Build from reality.