Intelligence · Module
Context Graph
How your customers, deals, tickets and suppliers relate, across every app
Context Graph connects entities across One3 apps into one relationship graph, resolves duplicates with steward review and explains paths between things. It also assembles permission-trimmed context for Copilot and Digital Employees.
One graph, no second copy
Every Action in One3 already writes typed relations: this contact belongs to that company, this ticket concerns that order. Context Graph builds on those relations instead of copying them. It is eligible from W2 and used internally by the platform first; v1 is planned for W3 and v2 for W4. In Year 1, the relations table and MDM-lite entity resolution are the substitute.
Entity resolution across apps
The same customer often appears in CRM, Service Desk and Billing. Context Graph proposes match candidates with a confidence score, sends them to a steward for review, and merges through the core merge Actions. Automatic merging happens only on an exact phone and email match within a tenant.
Context for Copilot and agents
When Copilot or a Digital Employee works on a record, Context Graph assembles the record plus one or two hops of related entities and recent events, trimmed to what that actor may see. Graph tools such as graph.neighbours and graph.paths are exposed over API and MCP with hop, row and time limits.
Explore and explain
The graph view shows a neighbourhood of one to three hops, filtered by entity type, relation type and time window. Nodes the viewer may not see are shown as "restricted" with no identifiers. Temporal edges support an "as of" view. "Why" traversal explains paths between two entities, or from an affected metric segment to linked change events such as a campaign pause, price change, stock-out or complaint spike. These links are labelled "linked, not causal".
Inferred relationships, carefully
Relationships mentioned in documents and emails can be inferred with a confidence score. Only low-risk relation types above a threshold are accepted automatically; others go to review. Inferred edges never drive high-risk Actions without a human-confirmed edge or other evidence. Tenants can configure their own entity and relation types from Apps Builder objects, and import or export relations in bulk with a dry run. Impact analysis and graph analytics are planned for W4.
Features
What Context Graph includes
Graph builder from Actions
Consumes Action events and writes typed, idempotent edges to the shared relations table.
Entity registry
Shared core entities and app objects with type metadata from Pack manifests.
Cross-app entity resolution
Match candidates across CRM, Desk and Billing with confidence, steward review and merge Actions.
Object context assembly
Record plus one or two hops and recent events, permission-trimmed, for Copilot and agents.
Graph view
One to three hop exploration filtered by entity, relation type and time; restricted nodes shown without identifiers.
Graph API and MCP tools
graph.neighbours and graph.paths with hop, row and time limits.
Temporal edges
Valid-from and valid-to dates with an "as of" view.
"Why" traversal
Explains paths between entities or from a metric segment to linked change events.
Inferred relationships
Entity mentions from documents and emails with confidence; review for anything but low-risk types.
Tenant ontology
Custom entity and relation types from Apps Builder objects with allowed pairs.
Bulk import and export
CSV and JSON lines with a dry run.
Impact analysis (W4)
What depends on a supplier, customer, product or employee transfer: open orders, projects, revenue at risk.
Workflow
Edges, resolution and use
- 1
Actions in every app write typed relations
- 2
The graph builder registers entities and edges
- 3
Duplicate candidates are scored and sent to a steward
- 4
The steward approves and the merge runs as a core Action
- 5
Copilot and agents receive permission-trimmed context from the graph
AI
AI skills and Digital Employees
Entity resolution candidates
Scores possible duplicates; merges are decided by a steward.
Context assembly
Builds record context within a token budget for Copilot and agents.
Entity mention extraction
Proposes inferred edges from documents and emails with confidence.
"Why" ranking
Ranks linked change events for a metric segment, labelled "linked, not causal".
Digital Employees that use this module
They act through the same governed Actions as people; risky steps wait for approval.
Frequently asked questions
Does Context Graph keep its own copy of my data?
No. It references the shared relations table and adds only resolution, review and view metadata.
Can it merge records on its own?
Only on an exact phone and email match within a tenant. Everything else goes to a steward.
When is it planned?
Eligible from W2 and outside the Year-1 boundary; v1 is planned for W3 and v2 for W4.
See Context Graph in your workflow
Tell us about your team and the outcomes you want. We will walk you through the apps, the Digital Employees and the plan for your industry.