Microsoft

Overview

Public technology position and identified correspondence

Microsoft provides the most explicit public documentation of the agent-platform primitives contemplated by the proposed architecture. Copilot Studio documentation describes agents configured with knowledge sources, tools and actions, connected agents to which work can be delegated, generative orchestration that selects tools and knowledge at run time, and autonomous triggers that allow agents to act on events without a user prompt.

Persistent agent memory, knowledge grounding across enterprise data via Microsoft Graph, Azure AI Search and Microsoft Fabric, and connector-based execution against hundreds of external systems, together constitute strong correspondence with the orchestration, agent, memory and enterprise-integration pillars.

The material distinction identified is subject and domain. Microsoft's documented agents are grounded in enterprise organisational data and act within the enterprise tenant. A creator/viewer duality, cross-platform creator audience intelligence, sponsorship and marketplace negotiation, and content rights lineage are not identified in reviewed public material as part of the documented agent platform. The proposed architecture's Digital Twin construct — a continuously updated behavioural model of a person that generates strategy for that person — is also not identified in reviewed public material in those terms; agent memory is documented as recall of prior context rather than as a predictive behavioural model.

Evidence register

Primary public evidence supporting each entry

Sources are prioritised in the order: official technical and developer documentation, official architecture and product documentation, official announcements and research, published patent material, then reputable secondary sources. All links open in a new tab.

Sections B, A and E

What they have / what we have

Where the reviewed company material demonstrates a capability more developed than the proposed architecture, that is recorded plainly. Where an element of the proposed architecture cannot be located, the finding is expressed as 'not identified in reviewed public material' — not as an assertion that the company lacks it.

What the company publicly demonstrates	What the proposed patent architecture contains	Key difference / potential patent position
Documented agents with knowledge, tools, connected agents, persistent memory, generative orchestration and autonomous triggers, executing against enterprise systems through connectors.	An orchestration layer with the same primitives, but whose principals are individual creators and viewers and whose execution targets are independent consumer platforms.	The agent primitives themselves show strong public correspondence. Differentiation, if any, must come from the coordinated combination with behavioural Digital Twins, cross-platform audience intelligence and rights-aware commercial orchestration. Requires further technical/legal analysis.
Enterprise-grade governance, tenant isolation, compliance and identity control materially beyond anything contemplated in the proposed specification.	Privacy-preserving learning and entitlement controls described at architectural level.	Microsoft demonstrates substantially stronger governance capability. This is an area where the proposed architecture is presently less developed.
Agent memory documented as retained context and personalisation across sessions.	Behavioural memory that continuously updates a Digital Twin, which in turn generates predictions and modifies subsequent autonomous decisions.	Recall versus predictive behavioural modelling with measured outcome write-back. The latter is not identified in reviewed public material.
Multi-agent orchestration between agents under common administrative control.	Agents representing separate principals (creator, brand, marketplace) discovering each other and conducting preliminary commercial negotiation.	Cross-principal commercial negotiation between autonomous agents is not identified in reviewed public material.

Architecture correspondence

Proposed stack mapped against Microsoft's reviewed public architecture

01 · External Digital Platforms / Enterprise Systems

Microsoft 365, Dynamics 365, Azure services and hundreds of documented connectors to external systems.

02 · Intelligent AI Orchestration Layer

Generative orchestration in Copilot Studio and the Foundry Agent Service select tools, knowledge and sub-agents at run time.

03 · Autonomous AI Agents

Agents and connected agents, including delegation to specialised agents.

04 · Digital Twins

Not identified in reviewed public material as behavioural Digital Twins.

05 · Persistent Behavioural Memory

Documented agent memory retaining context across sessions.

06 · Semantic Knowledge Graph

Microsoft Graph and enterprise knowledge sources; a semantic graph across creators, rights and commercial agreements is not identified in reviewed public material.

07 · Predictive / Commercial / Audience / Content / Rights Intelligence

Analytics through Fabric and Dynamics; predictive creator/audience intelligence not identified in reviewed public material.

08 · Autonomous or Assisted Workflow Decisions

Autonomous triggers and human-in-the-loop approval documented.

09 · Execution Across External Platforms

Execution through connectors, tools and API actions.

10 · Observed Outcomes

Telemetry and analytics on agent runs.

11 · Continual Learning and Model Evolution

Evaluation and continuous improvement tooling documented; automatic write-back of measured outcomes into a behavioural twin is not identified in reviewed public material.

Correspondence ratings

Matrix entries for this company

AI orchestrationStrong public correspondence

Creator AI AgentPartial correspondence

Viewer AI AgentPartial correspondence

Specialised agentsStrong public correspondence

Agent-to-agent communicationStrong public correspondence

Digital TwinsPartial correspondence

Behavioural Digital TwinsNot identified in reviewed public material

Digital Content TwinsNot identified in reviewed public material

Persistent memoryStrong public correspondence

Knowledge graphsStrong public correspondence

Predictive intelligenceModerate public correspondence

Content lifecyclePartial correspondence

AI content transformationModerate public correspondence

Multimodal semantic searchStrong public correspondence

Cross-platform audience intelligencePartial correspondence

Commercial orchestrationModerate public correspondence

Sponsorship orchestrationNot identified in reviewed public material

Rights managementPartial correspondence

ProvenancePartial correspondence

AuthenticityPartial correspondence

Intelligent notificationsModerate public correspondence

Enterprise / API orchestrationStrong public correspondence

Distributed AIModerate public correspondence

Privacy-preserving AIModerate public correspondence

Explainable AIModerate public correspondence

Continual learningModerate public correspondence

Autonomous workflow executionStrong public correspondence

Feedback-based optimisationModerate public correspondence

Public patent research

Publicly available patent activity indicating technological direction

Patent references are evidence of technological direction only. The existence of another party's patent does not determine infringement or validity, and no such conclusion is drawn here. Searches below are live public queries for counsel to review and refine.

Agent orchestration and delegation

Public filings concerning multi-agent task orchestration, tool selection and workflow automation indicate sustained technological direction.

Open public patent search (Google Patents)

Persistent memory and personalisation

Public filings concerning conversational memory, personalisation and context retention.

Open public patent search (Google Patents)

Knowledge graph and semantic retrieval

Public filings concerning enterprise knowledge graphs and semantic search grounding.

Open public patent search (Google Patents)

Section F — Where the architecture could evolve

Patent evolution opportunities

Drafting considerations for patent counsel arising from this mapping. These are not draft claims and no final claim language is proposed.

High Priority

Distinguish behavioural memory from conversational/context memory

Technology development observed: Persistent agent memory is now a documented, commoditised platform feature.
Demonstrated by: Microsoft (Copilot Studio, Foundry Agent Service)
Relation to existing architecture: Pillar 5 is at risk of being read as recall of prior context.
Addressed by present proposed specification?: Insufficiently — the proposed specification should be strengthened here.
Additional embodiment that may be useful: Describe memory records that store decision, prediction, executed action and measured outcome as a linked tuple, and a mechanism by which prediction error adjusts subsequent agent behaviour for that specific principal.
Potential use case: The Creator Agent's publishing-time recommendation shifts because previous predictions for that creator under-performed.
Possible future claim family: Potential claim family: outcome-linked behavioural memory conditioning subsequent autonomous decisions.

High Priority

Cross-principal agent negotiation

Technology development observed: Delegation between co-owned agents is publicly documented; negotiation between agents with opposed commercial interests is not identified in reviewed public material.
Demonstrated by: Microsoft (connected agents), Google (A2A), Salesforce (Agentforce)
Relation to existing architecture: Pillars 3 and 13.
Addressed by present proposed specification?: Named in the proposed architecture but requires deeper technical embodiment.
Additional embodiment that may be useful: Describe capability advertisement, offer/counter-offer exchange, constraint and entitlement checking, human approval thresholds, and the recording of negotiation outcomes to both parties' twins.
Potential use case: Creator Agent and Sponsorship Agent converge on preliminary sponsorship terms for human ratification.
Possible future claim family: Potential claim family: autonomous inter-principal commercial negotiation with approval gating.

Medium Priority

Autonomous trigger semantics for behavioural events

Technology development observed: Autonomous event triggers for agents are documented.
Demonstrated by: Microsoft
Relation to existing architecture: Pillars 1, 14 and 18.
Addressed by present proposed specification?: Partially.
Additional embodiment that may be useful: Describe triggers derived from predicted behavioural change (predicted audience migration, predicted churn of a subscriber cohort) rather than from discrete system events.
Potential use case: The Creator Agent acts on predicted audience migration before it is observable in platform analytics.
Possible future claim family: Potential claim family: prediction-derived autonomous workflow triggers.

Watch / Future Continuation

Governance and approval architecture

Technology development observed: Enterprise governance of autonomous agents is comparatively mature.
Demonstrated by: Microsoft
Relation to existing architecture: Pillar 1 and 17.
Addressed by present proposed specification?: Lightly.
Additional embodiment that may be useful: Describe explainability records as the audit substrate for autonomous commercial action, retained per decision.
Potential use case: A creator reviews why an agent declined a sponsorship.
Possible future claim family: Potential future continuation: explainability-linked autonomous action audit.

Scoring

Score derivation for this company

Category	Weight	Score	Weighted
Core architectural correspondence	30%	90	27.0
AI Agent correspondence	15%	92	13.8
Digital Twin / behavioural modelling	10%	40	4.0
Memory / knowledge architecture	10%	92	9.2
Creator / content / audience correspondence	15%	45	6.8
Enterprise / API orchestration	10%	95	9.5
Commercial / rights / provenance correspondence	5%	55	2.8
Continual learning / predictive intelligence	5%	70	3.5
Weighted total			76.5

Microsoft records the strongest agent, memory and enterprise correspondence of any company reviewed. Creator/content/audience correspondence is scored low because creator-ecosystem and audience-intelligence functions are not identified in reviewed public material for this platform.

Continue

Other company mappings

Adobe Meta Google Amazon Web Services Salesforce NVIDIA

Public technology position and identified correspondence

Primary public evidence supporting each entry

New autonomous agents scale your team like never beforeMicrosoft announced autonomous agents that can be triggered by events and act across business processes without a user prompt.Strong21 October 2024

Add other agents to your agent (multi-agent orchestration)Microsoft documents connecting agents so a primary agent can delegate tasks to specialised child agents and receive their results.StrongMicrosoft Learn — reviewed 2026

What is Azure AI Foundry Agent Service?Microsoft documents hosted agents with tools, knowledge grounding, and orchestration of multiple agents.StrongMicrosoft Learn — reviewed 2026

Threads, runs and messages in Agent ServiceMicrosoft documents persistent threads that retain conversation state across runs, with automatic context management.ModerateMicrosoft Learn — reviewed 2026

SeCom: On Memory Construction and Retrieval for Personalized Conversational AgentsMicrosoft Research publishes work on constructing and retrieving memory segments for personalised conversational agents.Moderate2025

Hybrid search in Azure AI SearchMicrosoft documents combined vector, keyword and semantic retrieval with reciprocal rank fusion for grounding agents.StrongMicrosoft Learn — reviewed 2026

Microsoft 365 Copilot connectors and the semantic indexMicrosoft documents ingesting external content into the tenant graph so it becomes retrievable organisational knowledge.StrongMicrosoft Learn — reviewed 2026

Agents are apps — Microsoft 365 Copilot extensibilityMicrosoft documents agents as first-class applications extending Copilot with organisational knowledge and actions.ModerateMicrosoft Learn — reviewed 2026

OneLake, the OneDrive for dataMicrosoft documents a single unified logical data lake underpinning analytics and AI workloads.PartialMicrosoft Learn — reviewed 2026

Transparency note for Azure AI Foundry Agent ServiceMicrosoft publishes a transparency note describing intended uses, limitations and evaluation of agent behaviour.ModerateMicrosoft Learn — reviewed 2026

What they have / what we have

Proposed stack mapped against Microsoft's reviewed public architecture

Matrix entries for this company

Publicly available patent activity indicating technological direction

Patent evolution opportunities

Distinguish behavioural memory from conversational/context memory

Cross-principal agent negotiation

Autonomous trigger semantics for behavioural events

Governance and approval architecture

Score derivation for this company

Other company mappings