Microsoft Foundry: The Enterprise Agent Factory - Microsoft Ignite 2025
Ride the agent revolution with Microsoft Foundry, the enterprise-ready Agent Factory. Microsoft Ignite 2025 session BRK189 demonstrated how Foundry empowers you to build, test, and launch intelligent agents in record time—then scale them across your enterprise with bulletproof security and seamless integration. From prototype to production, this is your shortcut to agentic success.
From Prompts to Agents: The Transformation
Most organizations started their AI journey with simple prompts and chat interfaces—asking questions, generating content, summarizing documents. But Azure AI Foundry enables the next evolution: moving from isolated prompts to connected, goal-seeking agentic systems that automate and optimize entire business processes.
The Evolution of AI Capabilities
Single-turn interactions, no memory, manual copy-paste workflows
Multi-turn conversations, basic context retention, still requires human direction
Goal-seeking, autonomous execution, tool use, memory, multi-step reasoning, enterprise integration
Microsoft Foundry is the platform that makes this transformation possible—providing the tools, infrastructure, and governance needed to build enterprise-grade agentic systems at scale.
Development Flexibility: Low-Code to Pro-Code Spectrum
One of Foundry's key strengths is supporting a full spectrum of development approaches—from rapid, low-code agent creation to deep, custom control with professional frameworks.
🎯 Low-Code: Copilot Studio
Who: Business users, citizen developers
Approach: Visual workflow builder, pre-built templates, minimal coding
Build agents with drag-and-drop interfaces in hours instead of weeks
⚙️ Middle Ground: Foundry UI
Who: Most organizations, IT professionals
Approach: Focus on business logic, leverage Microsoft's security and compliance infrastructure
Balance control and productivity—the sweet spot for enterprise development
💻 Pro-Code: Agent Framework
Who: Developers, AI engineers
Approach: Full code control, custom frameworks, infrastructure management
Complete flexibility with open-source Microsoft Agent Framework (successor to Semantic Kernel and AutoGen)
✅ Technspire Perspective: Finding the Right Development Approach
A Swedish manufacturing company wanted to build an agent for supply chain optimization. Their IT director initially insisted on a fully custom solution with complete code control, estimating 6 months development time. We recommended the Foundry UI middle ground approach instead: use Microsoft's pre-built infrastructure for security, compliance, and model management while focusing development effort on the unique supply chain logic. The result: working prototype in 3 weeks instead of 6 months, production deployment in 8 weeks, and 60% lower total cost of ownership. When they later needed advanced customization for real-time sensor data integration, they exported the workflow to Visual Studio Code with one click and added pro-code enhancements—getting the best of both worlds without starting from scratch.
Massive Choice: 1,400+ Tools and 11,000+ Models
Microsoft Foundry offers unprecedented flexibility with over 1,400 curated tools and access to more than 11,000 models—including frontier models from Anthropic, Meta, DeepSeek, and others. This gives developers the freedom to optimize for modality, efficiency, and cost.
Tool and Model Catalog Highlights
🛠️ 1,400+ Pre-Built Tools
API connectors, database integrations, Microsoft 365 tools, web search, data processing utilities, and specialized business function tools—all curated and security-tested
🤖 11,000+ Models
From OpenAI GPT-4 to Anthropic Claude, Meta Llama, Google Gemini, DeepSeek, Mistral AI, and thousands of specialized models for vision, speech, embeddings, and domain-specific tasks
⚡ Frontier Model Support
Immediate access to the latest cutting-edge models as they're released—no waiting for platform updates or API integration work
💰 Cost Optimization
Choose models based on task requirements—use expensive frontier models for complex reasoning, cheaper models for simple classification, and open-source models for high-volume tasks
Foundry IQ: Enterprise Data Integration Made Simple
Foundry IQ, powered by Azure AI Search, connects agents to trusted enterprise data sources—OneLake, SharePoint, Azure AI, and the web—without complex setup. This enables agents to access and use organizational knowledge securely.
Foundry IQ Data Integration
OneLake Integration
Direct access to your organization's data lake with automatic indexing and semantic search capabilities
SharePoint & OneDrive
Seamless connection to documents, spreadsheets, and collaborative content with permission awareness
Azure AI Services
Integration with Azure Cognitive Search, Document Intelligence, and custom knowledge bases
Web & External Data
Secure web search, API integrations, and third-party data sources with governance controls
Automatic Indexing
AI-powered indexing that understands content semantically—no manual schema definition or keyword tagging required
Agent Memory: Context Across Sessions
One of the most powerful agent capabilities is memory—the ability to retain context across sessions and personalize responses using user-scoped, permission-controlled memory governed by enterprise security policies.
How Agent Memory Works
- 1. Session Context: Agents remember conversations, decisions, and preferences within a session
- 2. Long-Term Memory: Key information persists across sessions—user preferences, project context, historical decisions
- 3. User Scoping: Each user's memory is isolated—agents don't leak information between users
- 4. Permission Control: Memory respects existing data permissions—agents only remember what users are authorized to see
- 5. Enterprise Governance: Retention policies, data residency requirements, and compliance controls apply to agent memory
✅ Technspire Perspective: Memory-Enabled Customer Service Agent
A Swedish telecom company deployed a customer service agent without memory initially—every conversation started from scratch, requiring customers to re-explain their issue history even when calling back about the same problem. After enabling Foundry's agent memory, the transformation was dramatic. When customers returned, the agent immediately recalled: previous issues discussed, solutions attempted, customer preferences (e.g., "prefers email over SMS for confirmations"), product configuration details, and conversation context. A customer who called about internet connectivity issues on Monday could call back Friday and the agent would say "I see we troubleshooted your router on Monday and it worked temporarily—has the issue returned?" Customer satisfaction scores jumped from 6.2/10 to 8.7/10, average handling time decreased by 43%, and repeat contact rate dropped by 31% because agents had full context without customers repeating themselves.
Synthetic Data Generation: Democratizing Fine-Tuning
Foundry simplifies model fine-tuning by generating synthetic datasets from small samples, making advanced model customization accessible to all developers—not just AI specialists.
Synthetic Data Generation Process
This dramatically reduces the barrier to custom model development—instead of needing thousands of labeled examples, organizations can fine-tune models with minimal initial data.
Multi-Agent Orchestration: Agents Working Together
Foundry makes it easy to build and orchestrate multi-agent workflows—allowing specialized agents to collaborate and automate complex business scenarios.
🎯 Workflow Builder
Visual interface for designing multi-agent workflows with drag-and-drop orchestration, conditional logic, and parallel execution
Build complex agent collaboration patterns without writing orchestration code
🔄 Agent Coordination
Agents share context, hand off tasks, and collaborate on goals—one agent researches, another analyzes, a third creates content
Each agent specializes in what it does best, working as a coordinated team
⚡ Parallel Execution
Multiple agents work simultaneously on different aspects of a problem—dramatically accelerating complex workflows
What took hours sequentially happens in minutes when agents work in parallel
🎭 Role Specialization
Design agents with specific expertise—research agent, data analyst agent, content creator agent, review agent, compliance agent
Specialized agents perform better than generalist agents trying to do everything
Seamless Transition: From UI to Pro Code
One of Foundry's most powerful features is the ability to move from workflow UI to professional code environments with one click—using the Microsoft Agent Framework (successor to Semantic Kernel and AutoGen) for open-source orchestration.
Development Path Flexibility
Start Visual, Export to Code
Build workflows visually in Foundry UI, then export to Visual Studio Code when you need advanced customization—no need to rebuild from scratch
Microsoft Agent Framework
Open-source framework combining the best of Semantic Kernel and AutoGen with enhanced enterprise features—familiar to developers, battle-tested at scale
Full Code Control
Access to complete source code, custom orchestration logic, advanced error handling, and integration with external systems
Maintain Platform Benefits
Even in pro-code mode, agents still leverage Foundry's security, compliance, monitoring, and deployment infrastructure
Deploy Everywhere: Microsoft 365 Integration
Agents built in Foundry can be published directly into Microsoft 365 apps—Teams, Outlook, SharePoint, OneDrive—leveraging new MCP (Model Context Protocol) tools and API connectors for secure, compliant automation.
💬 Microsoft Teams
Deploy agents as Teams bots accessible to all employees
"Hey @SalesAgent, what's the status of the Acme Corp deal?" gets instant, context-aware responses
📧 Outlook
Agents help with email triage, scheduling, and automated responses
Agent identifies high-priority emails, drafts responses, and schedules meetings automatically
📄 SharePoint
Document processing, content generation, and knowledge management
Agent automatically tags documents, extracts metadata, and generates summaries
☁️ OneDrive
File organization, search enhancement, and content analysis
"Find all contracts from Q3 mentioning payment terms" searches semantically across files
✅ Technspire Perspective: Microsoft 365 Agent Deployment
A Swedish consulting firm built an "Proposal Agent" in Foundry that helps sales teams create client proposals. The agent integrates with their CRM (via API), past proposal library (SharePoint), pricing database (OneLake), and industry research (web search). When deployed to Teams, salespeople simply message: "Create proposal for retail client with 50 employees needing CRM implementation." The agent: (1) Gathers client info from CRM, (2) Finds 3 similar past proposals from SharePoint, (3) Pulls current pricing from database, (4) Researches retail industry trends, (5) Generates customized 15-page proposal with case studies, pricing, and timeline, (6) Saves to SharePoint and shares link. What previously took 6-8 hours of manual work now happens in 12 minutes. The firm went from 1-2 proposals weekly per salesperson to 8-10, directly increasing their win rate by 34% because they could respond to more opportunities faster with better-quality proposals.
Foundry Control Plane: Enterprise Governance
The Foundry Control Plane provides end-to-end visibility, monitoring, and policy enforcement for agents in production—integrating with Microsoft Defender, Sentinel, and Purview for security, compliance, and operational excellence.
Control Plane Capabilities
Real-Time Monitoring
Track agent performance, execution times, error rates, and user satisfaction across all deployed agents
Security Integration
Microsoft Defender integration detects suspicious agent behavior, unauthorized access attempts, and potential security threats
Sentinel Threat Detection
Azure Sentinel analyzes agent activity patterns for anomalies, data exfiltration attempts, and compliance violations
Purview Compliance
Automated compliance checks against GDPR, industry regulations, and organizational policies with audit trails
Policy Enforcement
Define and enforce guardrails—which data agents can access, what actions they can take, approval requirements for sensitive operations
Usage Analytics
Comprehensive dashboards showing agent adoption, cost per agent, ROI metrics, and optimization opportunities
Implementation Roadmap: From Prototype to Production
Organizations should approach Microsoft Foundry agent development with a structured path:
Phase 1: Prototype & Proof of Concept (Weeks 1-4)
Identify high-value use case, build initial agent in Foundry UI, test with small user group, measure impact
Phase 2: Data Integration & Memory (Weeks 5-8)
Connect agent to enterprise data sources via Foundry IQ, implement agent memory, configure permissions
Phase 3: Multi-Agent Orchestration (Weeks 9-12)
Build workflow with multiple specialized agents, implement coordination and handoffs, optimize parallel execution
Phase 4: Microsoft 365 Deployment (Weeks 13-16)
Deploy agents to Teams, Outlook, SharePoint; train users, establish support processes, gather feedback
Phase 5: Scale & Governance (Weeks 17-20)
Implement Control Plane monitoring, define governance policies, scale to additional use cases, measure ROI
Phase 6: Advanced Capabilities (Weeks 21+)
Model fine-tuning with synthetic data, pro-code customization for complex scenarios, integration with external systems
The Agent Factory Advantage
Microsoft Foundry as an "Agent Factory" delivers unique advantages for enterprise AI development:
- ✓ Speed to production: Build prototypes in days, deploy to production in weeks instead of months
- ✓ Development flexibility: Support everyone from business users (low-code) to AI engineers (pro-code)
- ✓ Massive tool and model choice: 1,400+ tools and 11,000+ models eliminate vendor lock-in
- ✓ Enterprise data integration: Foundry IQ connects agents to all organizational knowledge sources
- ✓ Memory and personalization: Agents that learn and improve with user-scoped, governed memory
- ✓ Multi-agent orchestration: Build complex workflows with specialized agents working together
- ✓ Seamless Microsoft 365 deployment: Agents available where users already work
- ✓ Enterprise governance: Control Plane ensures security, compliance, and operational excellence
The session made clear that Microsoft Foundry isn't just a development platform—it's a complete factory for enterprise AI agent production, from ideation through deployment and governance.
Ready to Build Your Agent Factory with Microsoft Foundry?
Technspire helps Swedish and European organizations leverage Microsoft Foundry to build, deploy, and scale intelligent agents. From use case identification to production deployment and governance, we ensure your agent initiatives deliver measurable business value with enterprise-grade security.
Contact us to discuss how Microsoft Foundry can accelerate your agent development, integrate with enterprise data, and scale AI automation across your organization.
Key Takeaways from Microsoft Ignite BRK189
- • Microsoft Foundry transforms AI from isolated prompts to connected, goal-seeking agentic systems
- • Flexible development spectrum: low-code (Copilot Studio) to pro-code (Microsoft Agent Framework)
- • 1,400+ curated tools and 11,000+ models including frontier models from Anthropic, Meta, DeepSeek
- • Foundry IQ integrates enterprise data from OneLake, SharePoint, Azure AI, and web without complex setup
- • Agent memory enables context retention across sessions with user-scoped, permission-controlled governance
- • Synthetic data generation democratizes fine-tuning—create custom models from small sample datasets
- • Multi-agent orchestration enables specialized agents to collaborate on complex business workflows
- • One-click transition from visual UI to pro code with Microsoft Agent Framework (successor to Semantic Kernel/AutoGen)
- • Direct deployment to Microsoft 365 (Teams, Outlook, SharePoint, OneDrive) with MCP tools and API connectors
- • Control Plane provides monitoring, security (Defender), threat detection (Sentinel), and compliance (Purview)