Unclassified

IntegrateWise — Workbench Experience Package

Status: DRAFT — for product/experience review
Source of truth: docs/architecture/WORKBENCH_MATRIX.md, docs/architecture/PLATFORM_LOCK_v1.0.md
Mandatory constraint: One Workbench surface. No L1/L2/L4/L7 UI labels.

1. Core belief

The Workbench is not a dashboard. It is an operational surface.

The user does not switch contexts to work.
They stay in one place. Connected systems appear here when needed.
AI stays in the background until it is valuable.

2. What the user experiences

2.1 Layout

┌───────────────────────────────────────────────────────────────┐
│ Workspace shell                                              │
│ ┌─────────────────┐ ┌───────────────────────────────────────┐ │
│ │ Sidebar         │ │ Content                               │ │
│ │                 │ │                                       │ │
│ │ [Brand]         │ │ Current view:                         │ │
│ │                 │ │  • Home                                │ │
│ │ --- Work ---    │ │  • Accounts / Contacts                │ │
│ │ > Work          │ │  • Tasks / Calendar / Docs            │ │
│ │   Accounts      │ │  • Analytics / Projects               │ │
│ │   Contacts      │ │                                       │ │
│ │   Deals         │ │ ┌───────────────────────────────────┐ │ │
│ │   Tickets       │ │ │ Operational capability bar        │ │ │
│ │                 │ │ │ [Call] [Email] [Slack] [Schedule] │ │ │
│ │ --- Self ---    │ │ │ [Create task] [Open file] [Share] │ │ │
│ │   Tasks         │ │ └───────────────────────────────────┘ │ │
│ │   Calendar      │ │                                       │ │
│ │   Notes         │ │ Detail cards:                         │ │
│ │   Knowledge     │ │ • Summary / health / signals          │ │
│ │                 │ │ • Recent activity timeline            │ │
│ │ --- Platform ---│ │ • Connected entity list               │ │
│ │ > Activation    │ │                                       │ │
│ │   Integrations  │ │                                       │ │
│ │   Configuration │ │                                       │ │
│ │   Governance    │ │                                       │ │
│ │   Continuity    │ │                                       │ │
│ │   Network       │ │                                       │ │
│ │                 │ │                                       │ │
│ │ © Twin          │ │                                       │ │
│ └─────────────────┘ └───────────────────────────────────────┘ │
└───────────────────────────────────────────────────────────────┘

Rules:

Sidebar shows real control surfaces, never architecture labels.
Twin is not a sidebar destination. It is a capability inside the Workbench.
Content area changes with the active module. Capability bar stays contextual.

2.2 User mental model

The user thinks...	The platform does...
Where is everything for this account?	One Account view, all connected context
How do I contact this person?	`Call` / `Email` / `Message` in capability bar
What is happening right now?	Signals + timeline via Continuity
Can I delegate this?	`Assign your Twin` via OODA grammar
Is this safe to execute?	`Approve Twin’s Action` via Governance
Who do I become here?	Role projection changes modules/capabilities

3. Module anatomy

3.1 Home

Default landing inside /app.

Shows:

Active work summary for the selected role
Entities needing attention
Signals from Continuity
Suggested actions

No duplicate navigation. Home is a dashboard, not a menu.

3.2 Work modules

Department-specific operational views:

Accounts
Contacts
Deals / Opportunities
Tickets / Cases
Projects / Tasks
Meetings / Calendar
Documents / Knowledge
Analytics

Each module shows the same entity grammar:

Entity header
  • Identity fields
  • Status / health
  • Key numbers

Context panel
  • Connected records
  • Activity timeline
  • Linked entities

Operational bar
  • Contextual verbs only
  • Disabled when no provider is connected
  • Deep link when provider supports richer action

3.3 Self modules

Personal workspace:

Tasks
Calendar
Notes
Knowledge / Memory
Personal signals

These are personal projections of the same Spine data. Not a separate product.

3.4 Platform modules

Real control surfaces, always available:

Activation Hub — finish onboarding, enable intelligence
Integration Manager — connect providers, manage auth/sync
Configuration Manager — tenant schema, depth matrix, policies
Network Graphs — Spine relationships
Governance — approvals, audit, policy gates
Continuity — memory, context timeline

No “settings.” Everything has a name.

4. Operational capability grammar

The capability bar is the same grammar everywhere.
Only the available verbs change by entity type and connected providers.

4.1 Universal verbs

Verb	Meaning	Provider targets
View	Open canonical entity	CRM, GitHub, Notion, Jira
Open	Open in native app	Salesforce, HubSpot, Linear
Create	Create linked record	Any connected provider
Message	Send message	Slack, Teams, Gmail, Outlook
Call	Place call / open dialer	Zoom, Teams, Airwatch
Email	Compose email	Gmail, Outlook
Schedule	Create meeting	Calendar, Calendly, Zoom
Share	Share record / link	Any provider with share action
Assign	Delegate via Twin	Twin runtime
Approve	Execute approved action	Governance + Hermes

4.2 Rules

Max 6 visible verbs per context. Rest hidden under More.
Disabled state only when no provider supports the verb. Never hide capability entirely.
Deep link always preferred over embedding. Workbench opens native surface when action is non-trivial.
Twin verbs (Assign, Approve) are governed. They route through the OODA grammar, not direct execution.

5. Twin placement

Twin is not:

a left sidebar icon
a right panel
a chat widget
a button in the capability bar

Twin is:

ambient signal in the header when there is material output
contextual suggestion inside the capability bar when user is stuck
surfaced via Ask your Twin / Assign your Twin buttons

Default silence law applies.

6. Experience quality rules

One mental model: “What am I trying to do?”
No app switching penalty: Workbench opens native app only when needed, with return path.
No duplicate data entry: Spine is the source of truth. Connected app edits sync back.
No feature discovery tax: Platform modules are real controls, not settings.
No dead screens: Every view has actions or explicitly explains why actions are unavailable.

7. Minimum lovable Workbench (v0)

Deliver exactly these views first:

Home — work summary
Accounts — operational surface
Contacts — operational surface
Tasks — self operational surface
Calendar — self operational surface
Knowledge — personal + team artifacts
Integration Manager — connect providers
Activation Hub — onboarding completion
Governance — approvals
Continuity — memory

Everything else is Phase 1+.

8. Non-negotiables

One Workbench. No L1/L2/L4/L7 in the UI.
Sidebar shows Platform modules and Work modules. Nothing else.
Capability bar is contextual, not global.
Twin is silent by default.
AI is infrastructure. User never sees model selection, providers, tokens, or config unless explicitly needed.

Tagline:
Truth you own. AI you rent. Approval in between.