Connect the sources
Bring agreed documents and system feeds into the knowledge base while preserving their provenance and access boundaries.
Engram
Connect sources, facts and earlier decisions so applications can retrieve useful context and people can review the evidence behind a proposal.
The reason
Useful knowledge includes where a fact came from, when it was true and what it relates to. A search result alone does not preserve that context.
Engram combines retrieval with relationships and temporal records. Domain workflows can use that knowledge to prepare proposals for human review. The general Decision Engine is still in development.
Inside Engram
Select an image to take a closer look.
Bring agreed documents and system feeds into the knowledge base while preserving their provenance and access boundaries.
Keep relationships and time context alongside facts. Distinguish when information was valid from when it became known.
Combine semantic search with related facts instead of treating a retrieved passage as a complete answer.
The supply-chain pilot combines specialist findings and evidence. People review the proposal; general decision workflows remain under development.
A delivery is delayed. Which option fits?
Connected systems provide documents, data and conversation context. Source identity and access are part of the setup.
Illustrative target workflow · no live data. The new decision engine is under development. Available features and connections are agreed in the pilot.
ENGRAM × MICROSOFT 365 COPILOT
A business decision depends on how facts, dependencies and rules produce the result—and whether you can reconstruct that process later.
| Your requirementA concrete decision process | Microsoft 365 CopilotStandard product, without a separately built domain solution | EngramKnowledge architecture + domain decision workflow |
|---|---|---|
| Build on organisational knowledge | Work context for the answer Relevant content from Microsoft 365 and connected sources is assembled for the model and its tools. | Independent organisational memory Sources, extracted facts, relationships and earlier decisions form an independent knowledge base for applications and agents. Architecture |
| Interpret external signals | Use connected information Connectors and tools can bring in external information. Mapping an event to trade lanes, materials and orders requires domain logic. | Connect signals to operational impact The supply-chain demo connects logistics and risk signals with trade lanes, materials, stock and affected products. AIS and flight-data connectors are implemented. Supply-chain demo |
| Examine different perspectives | Agents according to configuration Specialist agents are possible. A council with defined domain roles and a combined result is an additional solution to configure. | A specialist council in the workflow Domain agents investigate a disruption using their respective data. Council synthesis combines findings and evidence into causes, impacts and proposed actions. Supply-chain demo |
| Understand what changed | Analyse historical content Can analyse earlier documents and conversations. This does not automatically create a domain model of validity and superseded statements. | Preserve what was true when The temporal knowledge graph distinguishes when a fact was valid from when it became known. Superseded statements retain their provenance. Architecture |
| Check the options | Model and available tools Can provide recommendations and calculations. Binding domain rules, eligibility checks and approvals must be configured as a workflow. | Explicit rules and judgments Domain logic checks constraints and calculates effects. The new decision engine adds bounded domain judgments; the explanation follows from them. Pilot + development |
| Inspect the result | Response, document or action The form and depth depend on context, instructions and configuration. A prescribed domain decision schema must be supplied. | A structured decision case The supply-chain demo includes a decision ledger with ranked actions, incremental costs, avoided penalties, conditions and reasons. Domain pilot |
| Reuse experience | History and work context Conversations and context can persist. A dedicated loop connecting decisions, actual outcomes and failure checks requires additional implementation. | Connect decisions to outcomes Earlier reasoning, failed attempts and observed outcomes are intended to inform the next decision. The general feedback workflow is being developed. In development |
| Operate the system yourself | Microsoft-operated service The Copilot core runs as a Microsoft service. Custom extensions do not change that deployment model. | Your knowledge base and model choice Engram is designed for self-hosting. Infrastructure, local or external models and data flows are configured for your environment. Deployment architecture |
Engram status: “Architecture” describes the documented design; “Domain pilot” identifies implementation in a bounded use case. The new general decision engine and end-to-end feedback workflow are in development. The available scope is demonstrated for each pilot.
The aim: a decision proposal whose basis you can inspect and whose outcome you can reuse.
BUILT EXAMPLE / SUPPLY CHAIN
This supply-chain application brings demand, suppliers, production, logistics and financial impact into one workflow. It shows what domain-specific decision support looks like as an application.
Vessel and flight feeds, logistics events and risk zones.
Connect affected lanes to materials, products, stock and customer exposure.
Specialist findings and supporting evidence feed a shared root-cause conclusion.
Compare actions, costs, avoided penalties and conditions before human review.
Demo implementation with external-feed connectors and example data. Live-feed availability depends on configuration; simulated fallback data is also supported. Domain logic includes demo assumptions. This demonstrates the supply-chain application, not universal feature availability in every Engram deployment.
Bring a concrete workflow. Let’s identify the knowledge, rules and evidence it needs, and what Engram can deliver for it today.
Microsoft 365 Copilot includes its documented work-context capabilities; separately developed domain solutions are not counted as part of the standard product. Based on Microsoft documentation checked on 22 September 2026, and Engram technical documentation and supply-chain pilot work. This is not a comparative performance test.
Engram is developed around a concrete domain and agreed sources. The supply-chain pilot demonstrates external signals, specialist council analysis and a decision ledger. That implementation is a reference for scoping a pilot, not a claim that every domain is ready out of the box.
Documented architecture, implemented domain pilots and ongoing development are distinct. The general Decision Engine and a complete decision-to-outcome feedback loop are evolving; available integrations and functionality are agreed for each pilot.
Copilot can use company knowledge and external data, with extensions for specialised workflows. Engram’s approach is a dedicated knowledge and decision process configured around your sources, criteria and operating requirements. Evaluate it against your existing workflow.
Start with the sources relevant to your use case. Before implementation, establish which systems can use existing interfaces and where a custom integration is needed.
The new decision engine is under development. It is intended to provide bounded domain judgments while explicit rules and calculations check quantities, costs and hard constraints. The result is a proposal with evidence and open conditions. Demonstrate and agree the available scope for your pilot.
In the workflow shown, Engram presents proposals for human review. Automated follow-up actions require a separately configured workflow with defined responsibilities and approvals.
Yes. Self-hosting and offline operation are possible. Infrastructure, models, access and data flows are configured for your environment. External model or research connections are explicitly agreed.
Evaluate the current product against your use case and establish the available features, required integrations and any additional development. The pilot starts with a bounded business question and measurable success criteria.
Engram guides