ADR-2026-09-04-15 — IntegrateWise Updated Full Form Architecture
Status: ACCEPTED — 2026-09-04 (founder updated full form canon; continuity layer between AI host, business ecosystem, and human workbench)
Context
Previous iterations treated the Spine alternatively as a database, an ETL repository, or a static registry. This failed to capture the dynamic reality:
- Companies already have their tools connected to hosts and SaaS suites.
- The user repeatedly faces "Day 0" context bankruptcy in AI tools, paying an immense "context tax".
- AI cannot be granted 100% blind authority, nor should humans be forced to act as manual routers.
- The Spine is not an ETL replica of all third-party data; it holds the SSOT anchor, business meaning, and relationship topology.
Decision
-
The Continuity Layer Invariant:
The complete platform is a continuity layer between the user's AI environment, the user's business ecosystem, and the user's workbench.- Core thesis: Know the user. Learn the user. Grow with the user.
- Core problem solved: Don't start from zero. Start from context.
-
Prima Facie Discovery at AI Host Marketplace Entry:
- Entry point: AI Host Marketplace (ChatGPT, Claude, Cursor, Slack).
- First operation: Discovery (User, Ecosystem, Work History), not raw data migration.
- Reusable connection bindings rather than credential duplication.
-
SSOT Anchor + Communication Framework over Heavy ETL:
- Flow A extracts structured business truth into the Spine; the operational depth remains in source systems.
- When on-demand deep context is required, the Co-worker queries source platform agents across the Communication Framework (A2A over ACP / Hermes / MCP) using the Spine's Account ID and source bindings.
-
The Account as the Gravitational Center:
- In B2B work, the Account anchors commercial, operational, financial, and communicative threads.
- Cross-system issues (e.g. Jira delay → Zendesk ticket → usage decline → unpaid Stripe invoice) are synthesized around the Account.
-
Stable Spine, Adaptive Projection:
- The Core Spine schema remains stable and additive-only.
- Role, department, and industry determine the projection on the Workbench (Sales sees pipeline; CSM sees retention).
-
The Complete Runtime Loop:
$$\text{User Intent} \longrightarrow \text{Spine Context} \longrightarrow \text{A2A Gap Fill} \longrightarrow \text{Context Assembly} \longrightarrow \text{Human Sixth Sense Decision} \longrightarrow \text{Governance Gate} \longrightarrow \text{Hermes Execution} \longrightarrow \text{KPI Reconciliation} \longrightarrow \text{Spine Growth}$$
Consequences
- The platform eliminates context rebuilding across sessions.
- Ingestion routes strictly via the 4 Flows (A, B, C, D) and 3 Memory Tiers (Organizational, Personal, Agentic).
- All AI work is bounded, evidence-linked, and reconciled against business KPIs.