Diagnosis
Drafting is solved. Agreement is the bottleneck.
Tailor compresses regulated agreement work by turning scattered review activity into a governed decision fabric: what was proposed, who reviewed it, what evidence was cited, which rules applied, what changed, and how agreement was reached.
The ecosystem below shows how that fabric is built: sovereign infrastructure, captured work signals, organisational constraints, agentic orchestration, model routing, and vertical products.
Sovereignty boundary
AloomU
The physical and logical boundary for sovereign workloads. AloomU defines what can stay in-country, what must remain in-substrate, and which AI tasks can safely route to external providers.
Decision capture
Tailor captures the work that usually disappears between meetings, documents, comments and approvals.
Audio
Conversations, calls, classrooms and meetings become transcripts, segments and proposed actions.
Visual
Walkthroughs, builds, environments and physical context become annotated scene intelligence.
Operational
Review events, comments, approvals, ownership changes, document versions and decision states become indexed decision memory.
Constraint layer
Axioms
Axioms are the rules Tailor reasons against: policies, procedures, delegations, legislation, contract requirements and approval rubrics.
They constrain topology proposals, guide GoalNode evaluation, and make every AI-assisted recommendation auditable against the organisation's own operating rules.
Tailor Fabric
Tailor Fabric is the governed orchestration layer. It turns captured context and organisational constraints into human-controlled, agent-run workflows.
Fabric lets teams observe, plan, execute, evaluate and adapt review work while preserving an audit trail of how each decision was reached.
Core capabilities
Human-authored workflow graph
Open Topology
Humans control the macro-DAG through nodes, edges and layout, with deterministic validators and optimistic concurrency control.
Bounded agentic workspaces
GoalNodes
Cyclic ReAct loops can run inside without breaking the acyclic macro-graph.
Safe rollback of graph and memory
Timeline
High-confidence actions auto-commit; the timeline allows safe rollback of graph topology and semantic memory.
Pre-flight validation of AI proposals
Inner Monologue
The LLM tests and corrects topology proposals against validators before surfacing them to the user.
Autonomous recovery suggestions
Self-Healing Pipelines
Failed nodes trigger structural patch proposals that bypass outages or repair execution paths.
Unaffected review streams keep moving
Parallel Branches
Independent execution branches continue while only the affected branch waits for human clarification.
Interfaces & protocols
- REST API — declarative graph mutations for nodes, edges and layout.
- SignalR — real-time graph mutation + agent telemetry.
- MCP — model context protocol integration.
- CLI — operator and developer access.
- WOPI — document workflow integration.
- PACT — proposal and contract layer.
- BAINK — sovereign billing layer.
Model routing
Every AI task has a sovereignty footprint.
Tailor routes reasoning, vision, document generation, GraphDiff proposals, embeddings and GoalNode evaluation to the right model based on sensitivity, latency, context length, modality and tenant policy.
Anthropic Claude
US-hostedDeep reasoning · long-context orchestration · complex GoalNode work · cleanup passes
Azure OpenAI
Azure tenant — region configurableVision · structured GraphDiff output · fast correction loops · embeddings · tenant-configurable regional hosting
Sovereign LLM on AloomU
In-substrate (TBD)Sensitive workloads · PII sanitisation · high-classification Axiom evaluation · self-healing context that must remain in-substrate
Routing policy · key design decision
Default reasoning · vision · long-context · low-latency · per-tenant overrides · sovereignty-only execution all need explicit rules.
Vertical products
The same decision fabric can be packaged into vertical products. Each one inherits Capture + Axioms + Fabric + Routing on the same sovereign substrate.
Live / emerging
Vantage
verticalCapital Programme Intelligence
Traide
verticalTrade Quoting Intelligence
TailorVision
platformVisual Intelligence
BAINK
platformSovereign Billing
Future / reserved
Praxis
verticalHealthcare Intelligence
Shield
verticalLaw Enforcement Intelligence
Spark
verticalEducation Intelligence
Host
verticalHospitality Intelligence
Sentinel
verticalManaged Sovereign Security
Decision register
The decisions still to make. Separated from the product story so they read as architectural calls, not uncertainty embedded in the substrate.
- Is AloomU the same entity as Loom, or a lower-level substrate?
- Are Axioms purely constraints, or also a capture modality?
- Should Threads, Stitches and Tailors become formal Fabric concepts?
- Which workloads must remain inside AloomU?
- Does TailorVision conflict with "Visual" as a capture-layer term?
- Is Spatial a platform module, a product, or only an internal capability?
- What is the final routing policy across Claude, Azure OpenAI and sovereign models?
Source data: src/frontend/src/data/ecosystem.json (#1176) · 4 live · 5 reserved · Anduril inline. Architecture overlay pending taxonomy lock.