Visual Framework

Interactive decision trees to guide Microsoft AI technology selection.

Use these diagrams after working through the Decision Framework and Evaluation Criteria. They reinforce the nine critical questions and are designed to support workshops, architecture reviews, and executive readouts.

Diagram Index
Complete Decision Flow
Persona-Based Flow
Data Grounding Decision
Complexity Assessment Flow
Budget & Timeline Tradeoffs
Governance & Compliance Path
Multi-Agent Orchestration
Upgrade Paths

Diagram Index

Diagram	Purpose	Maps To Framework
1. Complete Decision Flow	End-to-end technology selection	Phase 2: Q1-Q9 - All nine critical questions
2. Data Grounding Decision	Data strategy and knowledge sources	Phase 2: Q4 - Data grounding patterns
3. Persona-Based Flow	Selection by role and skill level	Phase 2: Q2 - Build approach + Scenarios
4. Complexity Assessment	Technical complexity evaluation	Evaluation Criteria: Complexity
5. Budget & Timeline	Cost and time-to-production paths	Evaluation Criteria: Budget & Time
6. Governance & Compliance	Security and compliance requirements	Evaluation Criteria: Governance
7. Multi-Agent Orchestration	Multi-agent patterns and frameworks	Quick Reference: Orchestration Complexity
8. Upgrade Paths	Migration and progressive enhancement	Implementation Patterns: Progressive Enhancement

Complete Decision Flow

flowchart TD
    Start([Start: Need AI Solution]) --> Q1{User interaction<br/>pattern?}
    
    Q1 -->|Conversational/Chat UI| UI_Path[UI-Based Agent]
    Q1 -->|Autonomous/Event-driven| Auto_Path[Autonomous Agent]
    Q1 -->|API/Headless Service| API_Path[Headless Service]
    
    UI_Path --> Q2{Where will users<br/>interact?}
    Q2 -->|M365 apps only| M365[M365 Copilot]
    Q2 -->|Teams + custom| Q3{Build approach?}
    Q2 -->|Multi-channel| Q3
    
    Q3 -->|Low-code| Studio[Copilot Studio]
    Q3 -->|Pro-code| Q4{Primary platform?}
    
    Q4 -->|M365-centric| SDK[M365 Agents SDK]
    Q4 -->|Azure-centric| Q4Azure{Need custom UI<br/>protocol?}

    Q4Azure -->|Yes| AGUI[Agent Framework + AG-UI<br/><i>Preview</i>]
    Q4Azure -->|No| Foundry[Microsoft Foundry - Azure]
    
    Auto_Path --> Q5{Build approach?}
    Q5 -->|Low-code| StudioAuto[Copilot Studio<br/>with Event Triggers]
    Q5 -->|Pro-code| Q6{Primary workflow type?}
    
    Q6 -->|Enterprise integration| LogicApps[Azure Logic Apps<br/>AI Agent Workflows<br/><i>Preview</i>]
    Q6 -->|Custom orchestration| Foundry
    
    API_Path --> Q7{Hosting preference?}
    Q7 -->|Managed PaaS| AgentService[Foundry Agent Service]
    Q7 -->|Self-hosted| Foundry
    Q7 -->|Local/Edge| WinAI[Microsoft Foundry - Windows<br/><i>Local</i>]
    
    M365 --> DataQ{Need custom data?}
    Studio --> DataQ
    StudioAuto --> DataQ
    SDK --> DataQ
    AGUI --> DataQ
    Foundry --> DataQ
    LogicApps --> DataQ
    AgentService --> DataQ
    
    DataQ -->|M365 data| GraphConn[Graph Connectors]
    DataQ -->|Documents| AISearch[Azure AI Search<br/>Agentic retrieval - ACL/labels]
    DataQ -->|Structured data| VectorDB{Vector DB choice?}
    DataQ -->|Analytics data| Fabric[Microsoft Fabric]
    DataQ -->|No grounding| DirectDeploy
    
    VectorDB -->|Global scale NoSQL| CosmosDB[Cosmos DB<br/>IVF/HNSW/DiskANN]
    VectorDB -->|Relational, OSS| PostgreSQL[PostgreSQL<br/>pgvector]
    VectorDB -->|SQL Database Engine| SQLFamily{Where does<br/>SQL run?}
    
    SQLFamily -->|Cloud PaaS| AzureSQL[Azure SQL Database<br/>VECTOR type<br/>ANN Index <i>Preview</i>]
    SQLFamily -->|Lift-and-shift PaaS| SQLMI[Azure SQL MI<br/>VECTOR type<br/>ANN Index <i>Preview</i>]
    SQLFamily -->|On-prem / VM| SQLServer[SQL Server 2025<br/>VECTOR type<br/>ANN Index <i>Preview</i>]
    SQLFamily -->|Fabric-native| SQLFabric[SQL database in Fabric<br/>VECTOR type<br/>ANN Index <i>Preview</i>]
    
    GraphConn --> DeployConfig
    AISearch --> DeployConfig
    CosmosDB --> DeployConfig
    PostgreSQL --> DeployConfig
    AzureSQL --> DeployConfig
    SQLMI --> DeployConfig
    SQLServer --> DeployConfig
    SQLFabric --> DeployConfig
    Fabric --> DeployConfig
    DirectDeploy --> DeployConfig
    
    DeployConfig[Configure Deployment] --> Q8{Deployment needs?}
    Q8 -->|M365 Copilot| Deploy_M365[Publish to M365 Copilot]
    Q8 -->|Teams| Deploy_Teams[Publish to Teams]
    Q8 -->|Web/Mobile| Deploy_Web[Web/Mobile Channels]
    Q8 -->|Multiple channels| Deploy_Multi[Multi-channel:<br/>10+ channels via SDK]
    Q8 -->|Azure service| Deploy_Azure[Azure Container Apps/<br/>App Service/AKS]
    
    Deploy_M365 --> Monitor[Monitor & Govern]
    Deploy_Teams --> Monitor
    Deploy_Web --> Monitor
    Deploy_Multi --> Monitor
    Deploy_Azure --> Monitor
    
    Monitor --> End([Production])
    
    style M365 fill:#004578,color:#fff
    style Studio fill:#004578,color:#fff
    style StudioAuto fill:#004578,color:#fff
    style SDK fill:#4b2070,color:#fff
    style Foundry fill:#a52617,color:#fff
    style AGUI fill:#4b2070,color:#fff
    style LogicApps fill:#004578,color:#fff
    style AgentService fill:#a52617,color:#fff
    style WinAI fill:#004578,color:#fff

Validation Summary

Last Validated: February 2026

UI-Based Agents (GA unless noted)

Technology	Action Safety	Proactive	Description
M365 Copilot	🔒 User-in-the-loop always	🔄 Reactive only	Conversational chat in M365 apps (docs)
Copilot Studio	⚠️ Actions can execute (add approval workflows)	🔄 Reactive (conversational) or ✅ Autonomous (event triggers)	Low-code, 13+ channels (docs)
M365 Agents SDK	⚠️ Custom action safety design	✅ Proactive capable	Pro-code, 10+ channels, C#/JS/Python, BYO orchestrator (docs)
Microsoft Foundry (Azure)	⚠️ Autonomous planning loops	✅ Proactive capable	Custom UI deployment (docs)
Agent Framework + AG-UI (Preview)	⚠️ Approvals via AG-UI middleware	✅ Proactive capable (inherits host orchestration)	Protocol bridges agents to web/mobile UI with SSE streaming, backend tool rendering, shared state, and CopilotKit components (docs)

Autonomous Agents

Technology	Action Safety	Proactive	Description
Copilot Studio (with event triggers)	⚠️ Actions can execute	✅ Autonomous (event-driven)	Event triggers: SharePoint, OneDrive, Planner, Recurrence (docs)
Logic Apps AI Agent Workflows (Preview)	⚠️ Autonomous execution	✅ Proactive (event-driven)	1,400+ connectors (docs)
Microsoft Foundry (Azure) Agent Service	⚠️ Autonomous planning loops	✅ Proactive capable	Custom orchestration (docs)

API/Headless Services (GA)

Technology	Action Safety	Proactive	Description
Foundry Agent Service	⚠️ Autonomous planning loops	✅ Proactive capable	REST API, managed PaaS (docs)
Microsoft Foundry (Azure)	⚠️ Autonomous planning loops	✅ Proactive capable	REST API deployment (docs)

Vector Databases

Technology	Status	Capabilities
Cosmos DB	GA	IVF, HNSW, DiskANN algorithms (docs)
PostgreSQL pgvector	GA	Extension 0.7.0 (docs)
Azure SQL Database	GA (ANN Index Preview)	Native VECTOR type, VECTOR_DISTANCE, DiskANN index, 1,998 dims, LangChain + Semantic Kernel integrations (docs)
Azure SQL MI	GA (ANN Index Preview)	Same SQL Database Engine; requires Always-up-to-date or SQL Server 2025 update policy (docs)
SQL Server 2025	GA (ANN Index Preview)	Native VECTOR type, float32/float16, on-prem or VM (docs)
SQL database in Fabric	GA (ANN Index Preview)	Fabric-native SQL with VECTOR type, same engine capabilities (docs)

Persona-Based Flow

flowchart TD
    Start([Who are you?]) --> P1{Your role?}
    
    P1 -->|End user| User[Use M365 Copilot]
    P1 -->|Business maker| Maker{Dev support?}
    P1 -->|Developer| Dev{Focus area?}
    P1 -->|Data scientist/analyst| DS{Primary workload?}
    P1 -->|Integration specialist| IntSpec[Azure Logic Apps<br/>AI Agent Workflows<br/><i>Preview</i>]
    
    Maker -->|No devs| MakerStudio[Copilot Studio]
    Maker -->|Occasional help| MakerPlus[Studio + Custom Actions]
    
    Dev -->|M365 integration| DevM365[M365 Agents SDK]
    Dev -->|Azure services| DevAzure[Microsoft Foundry - Azure]
    Dev -->|Multi-agent orchestration| DevOrch[Agent Framework<br/><i>Public Preview</i>]
    Dev -->|Multi-platform| DevChoice{Skill level?}
    Dev -->|Autonomous agents| DevAuto{Event-driven?}
    Dev -->|Custom UI streaming| DevAGUI[Agent Framework + AG-UI<br/><i>Preview</i>]
    
    DevChoice -->|Mid-level| DevStudio[Copilot Studio<br/>+ Custom Actions]
    DevChoice -->|Senior| DevSDK[M365 SDK or Foundry]
    
    DevAuto -->|Yes| DevLogic[Logic Apps<br/>AI Agent Workflows<br/><i>Preview</i>]
    DevAuto -->|Custom orchestration| DevFoundry[Microsoft Foundry - Azure<br/>+ Agent Service]
    
    DS -->|Analytics/BI| DSFabric[Fabric Data Agents<br/><i>Preview</i>]
    DS -->|ML/Custom models| DSFoundry[Microsoft Foundry - Azure]
    
    User --> UserEnd([Start using Copilot])
    MakerStudio --> MakerEnd([Build in Studio])
    MakerPlus --> MakerPlusEnd([Studio + code])
    DevM365 --> DevM365End([Build with SDK])
    DevAzure --> DevAzureEnd([Build in Foundry])
    DevOrch --> DevOrchEnd([Multi-agent workflows])
    DevStudio --> DevStudioEnd([Studio with code])
    DevSDK --> DevSDKEnd([Full code solution])
    DevLogic --> DevLogicEnd([Event-driven agent])
    DevFoundry --> DevFoundryEnd([Custom agent])
    DevAGUI --> DevAGUIEnd([Custom streaming UI])
    DSFabric --> DSFabricEnd([Data agent])
    DSFoundry --> DSFoundryEnd([ML pipeline])
    IntSpec --> IntSpecEnd([Enterprise workflows])
    
    style User fill:#0b6a0b,color:#fff
    style MakerStudio fill:#004578,color:#fff
    style MakerPlus fill:#004578,color:#fff
    style DevM365 fill:#4b2070,color:#fff
    style DevAzure fill:#4b2070,color:#fff
    style DevOrch fill:#4b2070,color:#fff
    style DevLogic fill:#004578,color:#fff
    style DevFoundry fill:#a52617,color:#fff
    style DevAGUI fill:#4b2070,color:#fff
    style DSFabric fill:#8c5e00,color:#fff
    style DSFoundry fill:#a52617,color:#fff
    style IntSpec fill:#004578,color:#fff

Validation Summary - Persona-Based Flow

Last Validated: January 28, 2026

End User (GA)

Technology	Description	Documentation
M365 Copilot	Built-in AI in M365 apps, no setup required	M365 Copilot

Business Maker (GA)

Technology	Description	Documentation
Copilot Studio	Low-code platform, no dev support needed	Copilot Studio
Copilot Studio + Custom Actions	Low-code with occasional developer support for custom connectors/flows	Custom Actions

Developer (GA unless noted)

Technology	Description	Documentation
M365 Agents SDK	Pro-code for M365-centric solutions, C#/JavaScript/Python, 10+ channels, BYO orchestrator	M365 Agents SDK
Microsoft Foundry (Azure)	Pro-code for Azure-centric solutions, custom models, full control	Microsoft Foundry
Microsoft Agent Framework (Public Preview)	Microsoft’s investment direction - Next-gen multi-agent orchestration framework (5 patterns: Sequential, Concurrent, Handoff, Group Chat, Magentic). Use instead of Semantic Kernel (maintenance mode). Works with M365 SDK or standalone.	Agent Framework
Agent Framework + AG-UI (Preview)	Protocol layer for web/mobile clients, supports SSE streaming, backend tool rendering, human approvals, shared/predictive state, and CopilotKit components.	AG-UI Integration
Copilot Studio + Custom Actions	Mid-level developers, low-code with custom code extensibility	Copilot Studio Extensibility
Logic Apps AI Agent Workflows (Preview)	Event-driven autonomous agents, 1,400+ connectors	Logic Apps Agent Workflows

Data Scientist/Analyst

Technology	Description	Documentation
Fabric Data Agents (Preview)	Analytics/BI focus, Python SDK, evaluation capabilities, Power BI/semantic models	Fabric Data Agents \| Python SDK
Microsoft Foundry (Azure)	ML/custom models, full AI/ML pipeline control	Microsoft Foundry

Integration Specialist

Technology	Description	Documentation
Logic Apps AI Agent Workflows (Preview)	Enterprise integration focus, 1,400+ connectors, workflow automation	Logic Apps Overview \| AI Agent Workflows

Data Grounding Decision

flowchart TD
    Start([Need to ground AI]) --> Q1{Where is<br/>your data?}
    
    Q1 -->|M365 only| M365Data{Data type?}
    Q1 -->|Azure only| AzureData{Data type?}
    Q1 -->|Both| Hybrid{Primary<br/>location?}
    Q1 -->|Analytics platform| FabricData[Microsoft Fabric]
    
    M365Data -->|SharePoint, OneDrive| Graph[Microsoft Graph Connectors]
    M365Data -->|Teams messages| GraphTeams[Graph Connectors]
    M365Data -->|Custom M365 app| GraphCustom[Custom Graph Connector]
    
    AzureData -->|Documents, PDFs| Q2{File count?}
    AzureData -->|Structured DB| DB{Database?}
    AzureData -->|Unstructured| Blob[Blob Storage with AI Search]
    
    Q2 -->|Under 10,000 files<br/>Simple setup| FileSearch{Platform?}
    Q2 -->|Production scale<br/>Custom chunking| Search[Azure AI Search]
    Q2 -->|Multimodal content| ContentUnderstanding[Azure AI Content Understanding<br/>with AI Search<br/>Preview]
    
    FileSearch -->|Azure Agent Service| AgentFileSearch[Agent Service File Search<br/>Auto chunk and embed]
    FileSearch -->|Copilot Studio| StudioKnowledge[Studio Knowledge Base<br/>Up to 1,000 files<br/>SharePoint or OneDrive]
    
    DB -->|Global scale, NoSQL| Cosmos{Vector algorithm?}
    DB -->|Relational, OSS| Postgres[PostgreSQL with pgvector]
    DB -->|SQL Database Engine| SQLFamily{Where does<br/>your SQL run?}
    
    SQLFamily -->|Cloud PaaS| AzureSQL[Azure SQL Database<br/>VECTOR type, DiskANN<br/>ANN Index Preview]
    SQLFamily -->|Lift-and-shift PaaS| SQLMI[Azure SQL MI<br/>VECTOR type<br/>ANN Index Preview]
    SQLFamily -->|On-prem / VM| SQL[SQL Server 2025<br/>VECTOR type<br/>ANN Index Preview]
    SQLFamily -->|Fabric-native| SQLFabric[SQL database in Fabric<br/>VECTOR type<br/>ANN Index Preview]
    
    Cosmos -->|Flat index| CosmosIVF[Cosmos DB IVF]
    Cosmos -->|Graph-based| CosmosHNSW[Cosmos DB HNSW]
    Cosmos -->|Disk-optimized| CosmosDiskANN[Cosmos DB DiskANN]
    
    Hybrid -->|M365 primary| HybridM365[Graph with BYOK]
    Hybrid -->|Azure primary| HybridAzure[AI Search with Graph API]
    
    FabricData --> Q3{Access method?}
    Q3 -->|Direct platform access| FabricPlatform[Fabric Lakehouse and Warehouse<br/>OneLake with SQL endpoint]
    Q3 -->|Conversational agent| FabricAgent[Fabric Data Agents<br/>Preview]
    
    Graph --> Platform{Which<br/>platform?}
    GraphTeams --> Platform
    GraphCustom --> Platform
    AgentFileSearch --> Platform
    StudioKnowledge --> Platform
    Search --> Platform
    ContentUnderstanding --> Platform
    CosmosIVF --> Platform
    CosmosHNSW --> Platform
    CosmosDiskANN --> Platform
    Postgres --> Platform
    AzureSQL --> Platform
    SQLMI --> Platform
    SQL --> Platform
    SQLFabric --> Platform
    Blob --> Platform
    HybridM365 --> Platform
    HybridAzure --> Platform
    FabricPlatform --> Platform
    FabricAgent --> Platform
    
    Platform -->|M365 Copilot| PlatformM365([Use built-in])
    Platform -->|Copilot Studio| PlatformStudio([Configure Studio])
    Platform -->|M365 SDK| PlatformSDK([Code integration])
    Platform -->|Microsoft Foundry Azure| PlatformFoundry([Configure Foundry])
    Platform -->|Logic Apps| PlatformLogicApps([MCP Server<br/>Connector])
    
    style Graph fill:#004578,color:#fff
    style AgentFileSearch fill:#004578,color:#fff
    style StudioKnowledge fill:#4b2070,color:#fff
    style Search fill:#004578,color:#fff
    style ContentUnderstanding fill:#006192,color:#fff
    style CosmosIVF fill:#004578,color:#fff
    style CosmosHNSW fill:#004578,color:#fff
    style CosmosDiskANN fill:#004578,color:#fff
    style Postgres fill:#004578,color:#fff
    style AzureSQL fill:#004578,color:#fff
    style SQLMI fill:#004578,color:#fff
    style SQL fill:#004578,color:#fff
    style SQLFabric fill:#8c5e00,color:#fff
    style FabricPlatform fill:#8c5e00,color:#fff
    style FabricAgent fill:#8c5e00,color:#fff

Validation Summary - Data Grounding Decision

Last Validated: February 6, 2026

M365 Data Sources (GA)

Technology	Capabilities	Documentation
Microsoft Graph Connectors	M365 data sources (SharePoint, OneDrive, Teams)	Graph Connectors Overview

Document Processing - File Search (GA)

Technology	Capabilities	Documentation
Foundry Agent Service File Search Tool	Built-in file search with automatic parsing, chunking (800 tokens/400 overlap), embedding (text-embedding-3-large), keyword + semantic search, reranking. Supports up to 10,000 files per vector store (max 512 MB/file). Two modes: Basic (Microsoft-managed) vs Standard (BYO Azure AI Search + Blob Storage). Supported formats: .doc, .docx, .pdf, .pptx, .py, .md, .txt, .json, .html, .java, .cs, .cpp, and more. Service handles entire ingestion automatically.	Agent Service File Search
Copilot Studio Knowledge Base	File upload from local/OneDrive/SharePoint. Supports .doc, .docx, .ppt, .pptx, .pdf, .xls, .xlsx, .txt, .md, .html, .csv, .xml. Max 512 MB per file. Direct uploads allow up to 500 files per agent, while SharePoint/OneDrive document sources now support up to 1,000 files (GA Oct 6, 2025). Automatic chunking and vectorization into Dataverse with semantic indexing. OneDrive/SharePoint: Auto-sync (updates reflected automatically) vs Upload: Static files. SharePoint: User-scoped permissions (only files user has access to).	Copilot Studio Knowledge \| File Upload \| SharePoint Files \| Use up to 1000 files

Document Processing - Production Scale (GA)

Technology	Capabilities	Documentation
Azure AI Search	Document indexing, full-text search, vector search, hybrid queries, custom chunking strategies (fixed-size, variable-size, Document Layout skill). Requires manual setup of indexers, skillsets, chunking strategy. Production-scale scenarios with millions of documents.	AI Search Overview \| Chunking Strategies

Document Processing - Multimodal (Preview)

Technology	Capabilities	Documentation
Azure AI Content Understanding	Multimodal processing (documents/images/audio/video), RAG-ready Markdown output, AI Search custom skill integration, built-in chunking, standard/pro modes. API version: 2025-05-01-preview.	Content Understanding Overview \| Multimodal Search

Structured Databases - Vector Search (GA)

Technology	Status	Capabilities	Documentation
Cosmos DB Vector Search	GA	IVF/HNSW/DiskANN algorithms, NoSQL & MongoDB vCore APIs	Cosmos DB Vector Search
PostgreSQL pgvector	GA	Extension version 0.7.0, HNSW/IVF indexes	PostgreSQL Vector Search
Azure SQL Database	GA (ANN Index Preview)	Native VECTOR type, VECTOR_DISTANCE, DiskANN index, 1,998 dims, LangChain + Semantic Kernel connectors	Azure SQL AI
Azure SQL Managed Instance	GA (ANN Index Preview)	Same SQL Database Engine; requires Always-up-to-date or SQL Server 2025 update policy	SQL MI Update Policy
SQL Server 2025	GA (ANN Index Preview)	Native VECTOR data type, float32 (1,998 dims)/float16 (3,996 dims)	SQL Server Vector
SQL database in Fabric	GA (ANN Index Preview)	Fabric-native SQL with VECTOR type, same engine as Azure SQL Database	SQL database in Fabric

Analytics Platform (GA with Preview features)

Technology	Status	Capabilities	Documentation
Microsoft Fabric Platform	GA	Direct knowledge source access via Lakehouse (Delta tables, Spark), Warehouse (T-SQL), OneLake (ADLS Gen2 APIs), KQL databases. Microsoft Foundry (Azure) integration for RAG	Fabric Overview \| Microsoft Foundry (Azure) FAQ
Fabric Data Agents	Preview	Analytics data grounding (warehouses, lakehouses, Power BI semantic models, KQL databases), Copilot Studio connected agents, Foundry Agent Service integration	Fabric Data Agents \| Copilot Studio Integration

MCP Integration (Preview)

Technology	Capabilities	Documentation
Logic Apps MCP Server	Standard logic apps as remote MCP servers, 1,400+ connectors, OAuth 2.0 auth, Streamable HTTP/SSE transports	Logic Apps MCP Server \| API Center Integration

When to Use File Search vs Azure AI Search

Solution	Use When	Cost
File Search (Agent Service/Copilot Studio)	Up to 10,000 files, simple setup (no manual indexer/chunking config), automatic embedding, suitable for smaller document sets, internal knowledge bases, rapid prototyping	Included in Agent Service consumption or Copilot Studio credits
Azure AI Search	Production scale (millions of documents), custom chunking strategies required, advanced features (analyzers, scoring profiles, faceting), complex indexing pipelines, enterprise search	Dedicated AI Search tier (Basic ~$75/mo to S3 ~$3K/mo)

Complexity Assessment Flow

flowchart TD
    Start([Assess Use Case]) --> Q1{How many<br/>data sources?}
    
    Q1 -->|1| Simple
    Q1 -->|2 to 3| Q2{Workflow<br/>complexity?}
    Q1 -->|4 or more| Complex
    
    Q2 -->|Linear| Medium
    Q2 -->|Branching| Complex
    
    Simple[Low Complexity] --> S_Tech{Skills?}
    S_Tech -->|Makers| S_M365[M365 Copilot with Graph]
    S_Tech -->|Devs| S_Studio[Copilot Studio]
    
    Medium[Medium Complexity] --> M_Tech{Skills?}
    M_Tech -->|Makers| M_Studio[Copilot Studio]
    M_Tech -->|Devs| M_SDK[Studio with Custom Actions]
    
    Complex[High Complexity] --> C_Tech{Need custom<br/>models?}
    C_Tech -->|No| C_SDK[M365 Agents SDK]
    C_Tech -->|Yes| C_Foundry[Microsoft Foundry Azure]
    
    S_M365 --> Timeline1[Days to deploy]
    S_Studio --> Timeline2[1 to 2 weeks]
    M_Studio --> Timeline3[2 to 4 weeks]
    M_SDK --> Timeline4[1 to 2 months]
    C_SDK --> Timeline5[2 to 3 months]
    C_Foundry --> Timeline6[3 to 6 months]
    
    style Simple fill:#0b6a0b,color:#fff
    style Medium fill:#8c5e00,color:#fff
    style Complex fill:#a52617,color:#fff

Budget & Timeline Tradeoffs

flowchart TD
    Start([Project Constraints]) --> Q1{AI setup<br/>budget?}
    
    Q1 -->|Existing M365| M365Only[M365-Only Path]
    Q1 -->|USD 200 to 500 per month| Starter[Starter Budget]
    Q1 -->|USD 1K to 5K per month| Growth[Growth Budget]
    Q1 -->|USD 5K or more per month| Enterprise[Enterprise Budget]
    
    M365Only --> M365T{Timeline?}
    M365T -->|Days| M365_Fast[M365 Copilot Chat<br/>0 addl cost included with M365<br/>Instruction-based agents]
    M365T -->|1 to 2 Weeks| M365_Med[M365 Copilot with Graph<br/>0 addl cost<br/>Knowledge grounding]
    M365T -->|1 Month or more| M365_Slow[Declarative Agents<br/>0 addl cost or PAYG<br/>Custom instructions and data]
    
    Starter --> StarterT{Timeline?}
    StarterT -->|1 to 2 Weeks| S_Fast[Copilot Studio PAYG<br/>USD 200 to 500 per month avg<br/>USD 0.01 per credit, low-code]
    StarterT -->|1 to 2 Months| S_Med[Studio with AI Builder<br/>USD 200 to 500 per month<br/>Document processing]
    StarterT -->|3 or more Months| S_Slow[Logic Apps AI Workflows<br/>Approx 200 to 400 per month<br/>Preview, event-driven]
    
    Growth --> GrowthT{Timeline?}
    GrowthT -->|2 to 4 Weeks| G_Fast[Copilot Studio Capacity<br/>200 or more per month prepaid<br/>25K credits per pack]
    GrowthT -->|1 to 3 Months| G_Med[M365 SDK with Azure Basic<br/>USD 1 to 3K per month<br/>Custom agents, Azure hosting]
    GrowthT -->|3 to 6 Months| G_Slow[Microsoft Foundry Azure Starter<br/>USD 1 to 5K per month<br/>PAYG tokens with AI Search Basic]
    
    Enterprise --> EntT{Timeline?}
    EntT -->|4 to 8 Weeks| E_Fast[Foundry Serverless<br/>Approx 5 to 15K per month est<br/>PAYG with AI Search Standard]
    EntT -->|3 to 6 Months| E_Med[Foundry with Agent Service<br/>USD 10 to 30K per month est<br/>Managed orchestration]
    EntT -->|6 or more Months| E_Slow[Foundry PTU with Premium<br/>30K or more per month<br/>PTU reservation with S2 or S3 Search]
    
    style M365_Fast fill:#0b6a0b,color:#fff
    style M365_Med fill:#0b6a0b,color:#fff
    style S_Fast fill:#0b6a0b,color:#fff
    style G_Fast fill:#0b6a0b,color:#fff
    style G_Med fill:#0b6a0b,color:#fff
    style E_Med fill:#0b6a0b,color:#fff

    style M365_Slow fill:#8c5e00,color:#fff
    style S_Med fill:#8c5e00,color:#fff
    style G_Slow fill:#8c5e00,color:#fff
    style S_Slow fill:#9a4300,color:#fff

Validation Summary - Budget & Timeline Tradeoffs

Last Validated: January 28, 2026

M365-Only ($0 AI infrastructure add’l)

Solution	Monthly Cost	Documentation
M365 Copilot Chat	Free (included) with eligible M365 subscription	Web-grounded chat and instruction-based agents (docs)
M365 Copilot + Graph Connectors	$30/user/month M365 Copilot license	Graph Connectors included at no extra charge (docs)
Declarative Agents	$0 or PAYG	Instruction-based or public-web grounded = $0; shared tenant data = PAYG (docs)

Starter ($200-500/mo)

Solution	Monthly Cost	Documentation
Copilot Studio PAYG	$200-500/mo (typical)	$0.01 per Copilot Credit (docs)
AI Builder	Included in Power Platform	Document processing (invoices, receipts, contracts) (docs)
Logic Apps AI Workflows (Preview)	$200-400/mo (typical)	Consumption pricing, event-driven agents (docs)

Growth ($1K-5K/mo)

Solution	Monthly Cost	Documentation
Copilot Studio Capacity Packs	$200/month per 25,000 credits	Prepaid (docs)
M365 SDK + Azure	$600-2.3K/mo	SDK free; Azure hosting (App Service ~$100-300/mo) + Azure OpenAI PAYG (~$500-2K/mo tokens) (docs)
Microsoft Foundry (Azure) Starter	$1-5K/mo estimate	PAYG tokens + AI Search Basic (~$75/mo) (OpenAI pricing) \| (AI Search pricing)

Enterprise ($5K+/mo)

Solution	Monthly Cost	Documentation
Foundry Serverless	$5-15K/mo	PAYG tokens at scale + AI Search Standard S1 (~$250/mo) (docs)
Foundry + Agent Service	$10-30K/mo	Managed orchestration PaaS + AI Search S2 (~$1K/mo) (docs)
Foundry PTU + Premium	$30K+/mo	PTU reservations (50+ PTUs minimum) + AI Search S2/S3 (docs)

Timeline Estimates

Timeline	Use Case	Example Scenario
Days	M365 built-in features, no development	(scenarios)
1-2 Weeks	Low-code platforms (Copilot Studio Employee Self-Service, Logic Apps)	(HR Knowledge Base scenario)
1-3 Months	Custom agents with SDKs, moderate complexity	(Customer Support scenario)
3-6 Months	Microsoft Foundry (Azure) custom solutions, complex orchestration	(evaluation-criteria)
6+ Months	Enterprise-scale with PTU, fine-tuning, advanced patterns

Cross-Tier: Unified Pre-Purchase (P3)

Organizations running workloads across both Copilot Studio and Microsoft Foundry can purchase Agent Commit Units (ACUs) in a single pool starting at ~$19K/year. This eliminates separate procurement for each platform and is best for “Better Together” architectures that span Growth and Enterprise tiers. See Evaluation Criteria: Budget and Agent Pre-Purchase Plan.

Cost Calculation Notes

M365 per-user costs ($30/user/month) NOT included in bands - these are AI infrastructure costs only
Estimates assume moderate usage (not high-scale production)
Azure consumption highly variable based on tokens, requests, storage
PTU (Provisioned Throughput Units) require Azure Reservations for cost optimization

Sources

Copilot Studio Licensing (Updated: 2025-11-05)
M365 Copilot Cost Considerations (Updated: 2025-11-25)
Azure OpenAI Pricing (Updated: 2025)
Azure AI Search Tiers (Updated: 2025-11-06)
Logic Apps AI Agent Workflows (Preview, Updated: 2025-11-18)
AI Builder Overview (Updated: 2026-01-14)

Governance & Compliance Path

%%{init: {'theme':'dark'}}%%
flowchart TD
    Start([Governance Requirements]) --> Q1{Data residency<br/>requirement?}
    
    Q1 -->|M365 tenant only<br/>Strict boundary| R1[M365 Tenant Boundary]
    Q1 -->|Azure region<br/>Configurable| R2[Azure Region Control]
    Q1 -->|Flexible<br/>Multi-region OK| R3[Flexible Residency]
    
    R1 --> Q2{Built-in DLP<br/>sufficient?}
    Q2 -->|Yes, inherit M365| G1[M365 Copilot<br/>M365 Boundary with DLP]
    Q2 -->|Need custom agents| G2[Copilot Studio<br/>Power Platform Governance]
    
    R2 --> Q3{Network<br/>isolation?}
    Q3 -->|VNet with private<br/>endpoints required| G3[Microsoft Foundry Azure<br/>VNet with Private Endpoints]
    Q3 -->|Managed runtime<br/>with VNet needed| G4[AI Agent Service<br/>Full Azure Governance]
    Q3 -->|Power Platform<br/>governance sufficient| G5[AI Builder<br/>Power Platform DLP]
    
    R3 --> Q4{Development<br/>model?}
    Q4 -->|Enterprise workflows<br/>with integration| G6[Logic Apps Standard<br/>Azure RBAC with VNet]
    Q4 -->|Multi-channel<br/>custom agents| G7[M365 Agents SDK<br/>Custom Auth with Hosting]
    Q4 -->|Orchestration<br/>library only| G8[Agent Framework<br/>Host inherited governance]
    
    style G1 fill:#0b6a0b,color:#fff
    style G2 fill:#004578,color:#fff
    style G3 fill:#004578,color:#fff
    style G4 fill:#004578,color:#fff
    style G5 fill:#8c5e00,color:#fff
    style G6 fill:#004578,color:#fff
    style G7 fill:#4b2070,color:#fff
    style G8 fill:#a52617,color:#fff

Validation Summary: Governance & Compliance Path

Last Validated: January 2026

M365 Tenant Boundary Technologies

Technology	Action Safety	Proactive	Data Grounding/Memory/Analytics	Key Governance Features	Documentation
M365 Copilot (GA)	✅ User-in-the-loop (drafts only)	❌ Reactive only	Grounding only (M365 Graph per-request), no extractable memory	M365 trust boundary, auto DLP/sensitivity labels, user-scoped permissions, Purview audit, no training on tenant data	M365 Copilot Security
Copilot Studio (GA)	⚠️ Actions execute (flows/connectors)	✅ Autonomous (event triggers)	Grounding + Dataverse persistence (transcripts, variables), full analytics	Power Platform RBAC + DLP, environment-level governance, connector controls, ⚠️ web search leaves boundary, ⚠️ dual auth (user/service)	Copilot Studio Security

Azure Region Control Technologies

Technology	Action Safety	Proactive	Data Grounding/Memory/Analytics	Key Governance Features	Documentation
Microsoft Foundry (Azure) (GA)	⚠️ Tool calling with autonomous planning	✅ Proactive (Azure Functions, Logic Apps)	Grounding + BYO thread storage (Cosmos DB), Azure Monitor + OpenTelemetry	Azure RBAC (control + data plane), VNet + private endpoints, managed identity, CMK optional, ⚠️ API key OR Entra ID (Entra recommended)	Microsoft Foundry security
AI Agent Service	⚠️ Autonomous with action tools (Logic Apps, Functions, MCP)	✅ Proactive (event triggers)	BYO thread storage (Cosmos DB), Azure Monitor project-scoped	Full RBAC (project + resource), VNet + private endpoints, BYO storage, no public egress by default, container injection for VNet	Agent Service Security
AI Builder (GA)	Varies by model type	❌ Reactive	Dataverse storage	Power Platform DLP, Dataverse RBAC, environment location	AI Builder Governance

Flexible Residency Technologies

Technology	Action Safety	Proactive	Data Grounding/Memory/Analytics	Key Governance Features	Documentation
Logic Apps Standard (GA)	⚠️ Autonomous execution (workflows)	✅ Proactive (event-driven, 1,400+ connectors)	Workflow state + connector data	Granular Azure RBAC, VNet + private endpoints (Standard), managed identity, Customer Lockbox, FedRAMP/HIPAA/ISO 27001	Logic Apps Security
M365 Agents SDK (GA)	⚠️ Custom design (full developer responsibility)	✅ Proactive (custom event handling)	Custom (developer implements)	Custom auth (MSAL, Entra ID), hosting platform RBAC, ⚠️ delegated OR application permissions, self-hosted = full network control	M365 SDK Auth
Agent Framework (Preview)	Inherits from host app	Inherits from host app	Inherits from host app	No built-in governance (library only), inherits from hosting platform	Agent Framework

Key Decision Criteria:

M365 tenant boundary required → M365 Copilot (strict) or Copilot Studio (custom agents)
Azure region control + VNet isolation → Microsoft Foundry (Azure) or Foundry Agent Service
Enterprise workflow automation → Logic Apps Standard
Custom multi-channel agents → M365 Agents SDK (full control, full responsibility)
Orchestration library only → Agent Framework (no built-in governance)

Multi-Agent Orchestration

flowchart TD
    Start([Need Multi-Agent?]) --> Q1{Pattern Type?}
    
    Q1 -->|Connected agents<br/>Mesh A2A| Connected[Connected mesh pattern]
    Q1 -->|Sequential or parallel<br/>workflows| Workflows[Agent Workflow Orchestration]
    Q1 -->|Event triggers| EventDriven[Event-Driven Agents]
    
    Connected --> C_Platform{Platform?}
    C_Platform -->|Low-code| C_Studio[Copilot Studio Preview<br/>Agent2Agent A2A<br/>Decentralized Mesh]
    C_Platform -->|Azure| C_Foundry[Microsoft Foundry Azure GA<br/>Connected agents<br/>Sub-agent delegation]
    
    Workflows --> W_Framework{Framework?}
    W_Framework -->|Microsoft| W_AgentFW[Microsoft Agent Framework Preview<br/>Sequential, Concurrent, Handoff, Magentic]
    W_Framework -->|Open source| W_SK[Semantic Kernel<br/>Sequential, Concurrent, Group Chat<br/>Handoff, Magentic]
    W_Framework -->|Bring your own| W_SDK[M365 Agents SDK<br/>Integrate Agent Framework or SK]
    W_Framework -->|Third-party state| W_LangGraph[LangGraph Third-Party<br/>State graphs]
    
    EventDriven --> E_Type{Event source?}
    E_Type -->|Enterprise systems| E_Logic[Logic Apps Preview<br/>AI Agent Workflows<br/>MCP Server<br/>Single agent event triggered]
    E_Type -->|Azure events| E_Functions[Azure Functions<br/>Agent Service<br/>Single agent event triggered]
    E_Type -->|Custom events| E_Custom[Event Grid with Foundry<br/>Event routing to agents]
    
    C_Studio --> Note1[Can connect Fabric Data Agents<br/>as data grounding participants]
    C_Foundry --> Note2[Fabric Data Agents can integrate<br/>as connected agents]
    
    C_Studio --> Deploy1([Deploy])
    C_Foundry --> Deploy2([Deploy])
    W_AgentFW --> Deploy3([Deploy])
    W_SK --> Deploy4([Deploy])
    W_SDK --> Deploy5([Deploy])
    W_LangGraph --> Deploy6([Deploy])
    E_Logic --> Deploy7([Deploy])
    E_Functions --> Deploy8([Deploy])
    E_Custom --> Deploy9([Deploy])
    
    style Connected fill:#0b6a0b,color:#fff
    style Workflows fill:#8c5e00,color:#fff
    style EventDriven fill:#4b2070,color:#fff

Validation Summary: Multi-Agent Orchestration

Last Validated: January 2026

Connected Agents / Sub-Agent Pattern

Technology	Status	Capabilities	Documentation
Copilot Studio	Preview	Agent2Agent (A2A) decentralized mesh, Connected agents, child agents, handoffs	Connected Agents
Foundry Agent Service	GA	Connected agents with natural language sub-agent delegation	Connected Agents
Fabric Data Agents	Preview	Consumed by other agents for data grounding (NOT orchestrator)	Fabric Integration

Agent Workflow Orchestration

Technology	Status	Orchestration Patterns	Documentation
Microsoft Agent Framework	Public Preview	Microsoft’s investment direction - Sequential, Concurrent, Handoff, Magentic (C#, Python)	Agent Framework
Semantic Kernel	Maintenance mode	Sequential, Concurrent, Group Chat, Handoff, Magentic (use Agent Framework for new projects)	Semantic Kernel Agents
M365 Agents SDK	GA	BYO orchestrator (integrate Agent Framework or third-party)	M365 SDK
LangGraph	Third-party	State graph management for complex workflows	Third-party framework

Event-Driven Agents (Single Agent Triggering)

Technology	Status	Event Handling	Documentation
Logic Apps AI Agent Workflows	Preview	Event triggers + MCP Server (triggers SINGLE agent, NOT multi-agent orchestration)	Logic Apps Workflows
Azure Functions + Agent Service	GA	Event-driven single agent invocation (event routing, NOT coordination)	Agent Service
Event Grid + Foundry	GA	Event routing to trigger agents independently (NOT orchestration)	Azure Event Grid

Key Distinctions:

Connected/Mesh: Agents discover and invoke each other (A2A) or delegate to sub-agents
Workflow Orchestration: Sequential/parallel/concurrent agent coordination patterns
Event-Driven: Single agent triggered by events (NOT multi-agent orchestration)

Upgrade Paths

flowchart TD
    Start([Start: Low-Code Agent]) --> Q1{Need more<br/>capability?}
    
    Q1 -->|No| Stay[Copilot Studio<br/>Standard]
    Q1 -->|Yes| Q2{What kind of<br/>capability?}
    
    Q2 -->|Better Reasoning/LLM| Path1[Enhance<br/>BYO Model - Foundry<br/><i>Preview</i>]
    Q2 -->|Complex Orchestration| Path2[Extend<br/>Handoff to Agent Service<br/><i>Preview</i>]
    Q2 -->|Custom UI/Protocol| Path3[Migrate<br/>M365 Agents SDK<br/>or Foundry]
    
    Path1 --> Result1([Copilot Studio<br/>+ Specialized Model])
    Path2 --> Result2([Copilot Studio UI<br/>+ Azure Agent Brain])
    Path3 --> Result3([Full Pro-Code<br/>Solution])
    
    style Stay fill:#004578,color:#fff
    style Path1 fill:#8c5e00,color:#fff
    style Path2 fill:#a52617,color:#fff
    style Path3 fill:#4b2070,color:#fff

Validation Summary: Upgrade Paths

Last Validated: January 2026

Progressive Enhancement Mechanisms

Mechanism	Status	Description	Documentation
BYO Model	Preview	Swap default Copilot Studio model for a specialized Foundry model (1,900+ options)	BYO Model
Agent Handoff	Preview	Copilot Studio delegates to Foundry Agent Service for complex tasks	Agent Handoff
Shared Knowledge	GA	Both Copilot Studio and Custom Apps can consume the same Azure AI Search index	Azure AI Search

Using This Guide

How to navigate: Start with Complete Decision Flow for end-to-end guidance, or jump to specific sections: Persona-Based Flow (by role), Data Grounding (data architecture), Complexity Assessment (effort estimate), Budget & Timeline (cost/time), Governance Path (compliance), or Multi-Agent (orchestration).

Diagram conventions: Diamonds = decisions, Rectangles = technologies, Circles = start/end. Colors: Blue = Microsoft core, Purple = developer tools, Green = low complexity, Orange = medium complexity, Red = high complexity/enterprise.

Next Steps

Detailed comparisons:
→ Feature Comparison

Real-world examples:
→ Scenarios

Evaluate readiness:
→ Evaluation Criteria

Last Updated: January 28, 2026
Next: Evaluation Criteria - Score complexity, skills, budget, and governance after selecting a path

Visual Framework

Table of contents

Diagram Index

Complete Decision Flow

Validation Summary

UI-Based Agents (GA unless noted)

Autonomous Agents

API/Headless Services (GA)

Vector Databases

Persona-Based Flow

Validation Summary - Persona-Based Flow

End User (GA)

Business Maker (GA)

Developer (GA unless noted)

Data Scientist/Analyst

Integration Specialist

Data Grounding Decision

Validation Summary - Data Grounding Decision

M365 Data Sources (GA)

Document Processing - File Search (GA)

Document Processing - Production Scale (GA)

Document Processing - Multimodal (Preview)

Structured Databases - Vector Search (GA)

Analytics Platform (GA with Preview features)

MCP Integration (Preview)

When to Use File Search vs Azure AI Search

Complexity Assessment Flow

Budget & Timeline Tradeoffs

Validation Summary - Budget & Timeline Tradeoffs

M365-Only ($0 AI infrastructure add’l)

Starter ($200-500/mo)

Growth ($1K-5K/mo)

Enterprise ($5K+/mo)

Timeline Estimates

Cross-Tier: Unified Pre-Purchase (P3)

Cost Calculation Notes

Sources

Governance & Compliance Path

Validation Summary: Governance & Compliance Path

M365 Tenant Boundary Technologies

Azure Region Control Technologies

Flexible Residency Technologies

Multi-Agent Orchestration

Validation Summary: Multi-Agent Orchestration

Connected Agents / Sub-Agent Pattern

Agent Workflow Orchestration

Event-Driven Agents (Single Agent Triggering)

Upgrade Paths

Validation Summary: Upgrade Paths

Progressive Enhancement Mechanisms

Using This Guide

Next Steps