SymphonyAI’s IRIS Foundry is now available inside Microsoft Teams and Microsoft 365 Copilot, bringing conversational access to industrial data — and the ability to trigger maintenance workflows — directly to the collaboration chat used by plant-floor operators. The integration, announced this week, relies on the Model Context Protocol (MCP), a standard that allows Copilot to securely query live factory systems and take actions like creating work orders, without requiring workers to juggle multiple screens.

What IRIS Foundry in Teams Actually Delivers

The integration turns Microsoft Teams into a unified front end for industrial operations. Instead of toggling between data historians, SCADA systems, and maintenance applications, frontline workers can now use natural language to get answers and initiate actions from a single chat interface. The key capabilities include:

  • Conversational querying: An operator can type, “Show me recent heat exchanger anomalies at Plant 7,” and Copilot pulls real-time data from IRIS Foundry, returning a summary with time-series trends, recent alarms, and maintenance history.
  • Automated workflows: Copilot agents can create work orders in a CMMS, alert specific teams in a named channel, or post runbook excerpts — all from the Teams conversation, with human confirmation steps built in.
  • Custom industrial copilots: Using Copilot Studio, organizations can build role-based assistants (e.g., for rotating equipment or energy optimization) that are scoped to specific assets, standard operating procedures, and compliance rules.
  • Edge and cloud deployment: IRIS Foundry can run on Azure or at the edge via IoT modules, allowing low-latency responses even in environments with strict OT isolation.

Under the hood, IRIS Foundry ingests data from historians, PLCs, MES, and CMMS, organizing it into a semantic knowledge graph. The platform then exposes governed MCP endpoints that Copilot Studio discovers as tools — making the connection secure and auditable.

What the Integration Means for You

The impact of this announcement differs depending on your role. Here’s a breakdown.

For Frontline Operators and Engineers

If you spend your day switching between screens to monitor equipment and respond to alarms, the change is immediate. You can now query asset status, view anomalies, and kick off corrective actions without leaving Teams. This reduces context switching and can shave hours off mean time to repair (MTTR) — especially during high-pressure incidents. The natural-language interface also lowers the training barrier for newer staff.

For IT and OT Administrators

You’re on the hook for governance. Exposing operational data through MCP servers creates a new attack surface. You’ll need to configure least-privilege access policies in Copilot Studio, ensure MCP endpoints run behind private networks or virtual networks, and enable immutable audit logging. Microsoft’s Copilot Studio allows you to control which MCP tools are available to which agents, but that requires careful setup. Any automated action that could change process parameters must remain human-in-the-loop until validated.

For Plant and Business Leadership

The promise is a faster, more informed frontline — but only if the rollout is disciplined. SymphonyAI’s materials cite directional ROI numbers (like reduced downtime), but those should be treated as estimates until proven in a controlled pilot. You’ll need to model the total cost of ownership: Microsoft 365 Copilot licenses, Copilot Studio access, IRIS Foundry subscriptions, Azure infrastructure, and the engineering effort for data onboarding and change management. The integration is a tool, not a magic wand; its success will depend on clean data, accurate asset models, and the team’s willingness to adopt new workflows.

How We Got Here: The Road to Teams-Based Industrial AI

The announcement doesn’t come out of nowhere. SymphonyAI has been building IRIS Foundry for years as a DataOps platform that unifies operational technology (OT) and IT data. The company already offered Azure integration and edge deployment, making it a natural candidate for deeper Microsoft ecosystem ties.

Microsoft’s side of the equation evolved rapidly in 2024. The company added MCP support to Copilot Studio, allowing its agents to call out to external “knowledge servers” in a standardized, governable way. MCP — the Model Context Protocol — acts as an integration fabric, letting developers expose tools (data readers, actions) that AI agents can discover. This replaced brittle, one-off connectors with a centralized management and auditing layer.

SymphonyAI then built MCP servers for IRIS Foundry, packaging the platform’s data and actions into discoverable endpoints. The final piece was Copilot Studio’s ability to add these MCP servers as agent actions, complete with metadata and consent prompts. The result is the tight, chat-based experience now rolling out.

What You Should Do Now

If your organization is considering this integration, the following steps will help you move from announcement to value without stumbling:

  1. Inventory and prioritize your data estate
    Map out critical assets, data sources (historians, PLCs, CMMS), and the quality of your asset models. Choose a high-value, well-instrumented asset — such as compressors or heat exchangers — for a first pilot.

  2. Start small with a scoped pilot
    Define clear metrics: MTTR, downtime minutes, work order lead time, and operator time saved. Run the pilot on a single line or asset family for six to twelve weeks. Record pre- and post-deployment numbers to gauge actual impact.

  3. Bake governance into the pilot from day one
    - Configure MCP server scopes and per-agent consent prompts.
    - Enforce private endpoints, VNet controls, and full audit logging.
    - Validate that all actions leave a replayable trail before allowing any automated writes.

  4. Keep a human in the loop for prescriptive actions
    Any recommendation that could change a control setpoint, trigger a safety bypass, or alter a process parameter must require operator confirmation. Run simulations to test suggested changes before production rollout.

  5. Align procurement across Microsoft and SymphonyAI
    Combine all license, cloud, and integration costs into a single TCO model. Factor in training, change management, and potential edge hardware. Expect payback periods to extend beyond marketing claims, but plan for real gains when executed well.

  6. Build a cross-functional team
    Include data scientists, integration engineers, cloud architects, and — crucially — veteran operators and maintenance leads. The domain knowledge sitting in frontline workers’ heads is irreplaceable.

Outlook: What to Watch for Next

This integration is likely the first of many. Because MCP is an open protocol, we can expect other industrial software vendors to build similar bridges, creating a marketplace of certified connectors that reduce vendor lock-in. Edge-hardened MCP servers will become more common, targeting sites with intermittent connectivity, and regulatory bodies may soon issue formal guidance on the use of automated decision agents in safety-critical sectors.

For now, the IRIS Foundry integration is a concrete example of how AI can move from analytics dashboards into the flow of work — right where the people who keep plants running already spend their day.