IRIS Foundry — The Industrial AI platform

The industrial AI platform built to act, not just analyze

Most industrial AI stalls at the data layer—fragmented OT, IT, and engineering data that takes months to wrangle before a model runs. IRIS Foundry removes that bottleneck. It connects, contextualizes, and governs your industrial data, then puts it to work through 300+ pre-built agents (IRIS Flows) and a no-code app builder (IRIS Forge)—so predictive models, agents, and copilots go to production in weeks, not quarters.

IRIS Foundry is the industrial data backbone that connects every source—SCADA, historians, MES, ERP, PLCs, P&IDs, edge devices—and transforms raw data into governed, contextualized intelligence. The unified namespace and knowledge graph give every downstream AI application a structured, accurate view of your operations from day one.

Built on open, composable architecture with 100+ pre-built industrial connectors, edge-to-cloud deployment, and enterprise-grade security. IRIS Foundry works alongside your existing systems—no rip-and-replace required. Independently recognized as a leader in Asset Performance Management, Industrial Data Management, Industrial AI Analytics, and Industrial DataOps by key analysts.

The full industrial AI stack—data, agents, and apps in one governed platform

IRIS Foundry, IRIS Flows, and IRIS Forge deliver everything industrial AI needs—built in, not bolted on. Pre-trained models, 300+ industrial agents, a no-code app builder, and an MCP server, all on one governed foundation.

Unified Namespace

A single, real-time source of truth for all industrial data. Connects OT, IT, and engineering sources into one standardized structure — no rigid ISA-95 hierarchies. Every system speaks the same language.

Industrial Cortex (Knowledge Graph)

Asset hierarchies, failure modes, process relationships, and KPIs — connected and contextualized. Makes root cause analysis, fault tracing, and KPI diagnostics navigable.

P&ID ingestion

AI-powered ingestion of piping and instrumentation diagrams transforms static engineering documents into live, queryable data linked to asset models and process flows. A differentiator that generic platforms cannot match.

ML Studio and MLOps

Build, validate, deploy, and monitor AI models in a governed, repeatable workflow. Pre-built engines for predictive maintenance, anomaly detection, process optimization, and forecasting — ready to configure, not build from scratch.

IRIS Forge — prompt-to-production app builder

Plant engineers and operators build and deploy industrial AI applications from natural-language prompts. No data science team. No development ticket. Hours to production. Native Azure deployment and NVIDIA Omniverse integration for real-time 3D digital twin applications.

Open agentic architecture and MCP server

300+ pre-built industrial agents, a low-code flow builder (IRIS Flows), and an MCP server that exposes IRIS Foundry intelligence to Microsoft Teams, M365 Copilot, Claude, Cursor, and any MCP-compatible tool — no custom integration required.

Connect

100+ prebuilt connectors ingest data from every OT, IT, and engineering source—SCADA, historians, ERP, MES, PLCs, edge devices—in real time or batch. No custom connectors, no manual pipeline builds.

Contextualize

The unified namespace and industrial knowledge graph transform raw data into governed, structured intelligence—with asset hierarchies, ontologies, and process relationships built in and maintained automatically.

Build

Deploy pre-built AI models and applications or build your own with IRIS Forge's no-code interface. ML Studio handles model development, validation, and MLOps at scale—from edge to cloud.

Act

Deploy 300+ pre-built agents through IRIS Flows to run workflows, not just inform them. Modular, composable architecture rolls across sites—edge, hybrid, or cloud—without starting over.

Get a demo
Industrial LLM — purpose-built for operations
Trained on 7+ trillion industrial data points. Understands P&IDs, failure modes, and alarm hierarchies at the model level — not the prompt level. Reduces the hallucinations that make generic LLMs unsafe in high-stakes environments.

WHAT ALTERNATIVES REQUIRE

Generic LLMs require months of fine-tuning before the model reliably understands your operations. You pay for that setup time every time.
300+ pre-built agents — production-ready on day one
IRIS Flows ships digital workers for alarm management, root-cause analysis, predictive maintenance, shift intelligence, and production optimization — validated in industrial environments, not lab demos. You configure; you do not build from scratch.

WHAT ALTERNATIVES REQUIRE

Build-your-own means months before a single agent reaches production. Generic agentic platforms provide a framework — not validated industrial logic.
Governed autonomy — three modes to choose from
Assistive, Augmented, and Autonomous modes provide a defined path to closed-loop AI. Start where you are comfortable, expand agent authority as trust builds, and keep every decision auditable by design. Built for safety-critical operations.

WHAT ALTERNATIVES REQUIRE

Governance bolted on after the fact. Audit trails and rollback are separate projects. Agent rationale is rarely explainable by default.
Edge-to-cloud AI — intelligence where the asset lives
Predictive, generative, and agentic AI running at the edge — real-time inference at the machine, without cloud round-trips. Built for latency-sensitive, remote, offshore, and OT-isolated sites.

WHAT ALTERNATIVES REQUIRE

Most platforms are cloud-first. Edge deployment is an afterthought or a separate SKU — not a design principle baked into the architecture.
No-code for the frontline — IRIS Forge
Plant engineers and operators build and deploy workflows in natural language. No data science team. No development ticket. Value compounds across the whole organization, not just a central center of excellence.

WHAT ALTERNATIVES REQUIRE

Every change routes back through a central CoE, a development backlog, or an integrator — slowing the teams that need AI most.
Open, not siloed — works with what you have
IRIS Foundry works alongside AVEVA, AspenTech, SAP, OSIsoft PI, DCS, and SCADA through 100+ pre-built connectors. Open at the AI layer too — any LLM, any cloud, any MCP-compatible tool. No rip-and-replace.

WHAT ALTERNATIVES REQUIRE

Build-your-own means permanent data-engineering, ML, and platform FTEs. Total cost of ownership scales with complexity — not with value delivered.

Production scale, in production environments—not pilot decks

6.7T


Asset data points processed annually. 80,000+ assets, 3 PB of industrial data, every year.

12 weeks


Typical time to first production value. Versus 12–24 months for build-your-own approaches.

Trusted by industrial leaders across 1,000+ plants

IRIS Foundry in production—across energy, manufacturing, food and beverage, and industrial gases.

A culture of innovation and efficiency — Novelis

Novelis, the world's largest flat-rolled aluminum manufacturer, moved from preventive to predictive maintenance with SymphonyAI's Predictive Asset Intelligence, built on IRIS Foundry. Spanning a mix of ERP systems, historians, and equipment generations from state-of-the-art to 1960s legacy. The result: dramatically reduced unplanned downtime across multiple facilities, predictive insights embedded into daily workflows, and a clear path toward Novelis's 'plant of the future' vision.

Read the case study

Multi-site

Unplanned downtime reduced across multiple facilities

Preventing unplanned downtime across European plants — Nippon Gases

Nippon Gases needed a scalable AI platform to keep critical assets running across its European plants—and to catch failure modes that vibration and oil analysis alone could not. SymphonyAI's Predictive Asset Intelligence, built on IRIS Foundry, monitors compressors, high-voltage motors, turbines, heat exchangers, and pre-purifiers with real-time prescriptive monitoring and predictive maintenance. Anomalies caught before unplanned stoppages. Reliability program extended company-wide.

Read the case study

$3M

Annual savings from avoiding unplanned downtime and production loss

$500K

Gain in production throughput per site per year

100% manufacturing compliance through AI — GSK Biologicals

GSK Biologicals needed to simplify inspections, empower frontline inspectors, and maximize manufacturing compliance across a complex, GxP-regulated environment. Using SymphonyAI's Connected Worker platform, built on IRIS Foundry, the team brought AI-powered workflows into daily operations—replacing manual reporting, Excel-based issue tracking, and endless email chains with guided, auditable digital processes. Complete compliance achieved: every task completed by the right person, at the right time, in the right place.

Read the case study

100%

Manufacturing compliance achieved

Questions buyers ask about IRIS Foundry

Straight answers on deployment, integration, and what makes IRIS Foundry different from generic alternatives.

How does IRIS Foundry differ from horizontal AI platforms and from data-only industrial platforms?
Horizontal AI platforms provide compute, models, and tooling, but no industrial domain knowledge, no pre-built OT connectors, and no asset ontologies. Data-only industrial platforms stop at the data layer—they contextualize, but you still build the agents, apps, and workflows on top. IRIS Foundry arrives with all three: a contextualized data foundation (Foundry), 300+ pre-built agents (Flows), and a no-code app builder (Forge), plus industrial data models (ISA-95, ISA-88, CFIHOS) and 100+ OT/IT connectors. We build on the major cloud platforms—including a deep partnership with Microsoft Azure—so the underlying compute stays open, and the industrial intelligence on top is purpose-built. Independently recognized by Verdantix and IDC across Data Management, AI Analytics, APM, and DataOps.
Most customers reach first production value in 8–12 weeks. IRIS Foundry’s modular architecture lets you connect the highest-value data sources first, deploy the relevant pre-built models and agents, and scale from there—without waiting for a full enterprise rollout to put anything in production. A global agri/food customer stood up the standard model in 12 weeks and is now rolling out to hundreds of sites annually. The build-your-own alternative typically takes 12–24 months to the first production insight.
No. IRIS Foundry is designed to work alongside your existing infrastructure through open APIs and 100+ prebuilt connectors. AVEVA, Aspen,Tech SAP, OSIsoft PI, DCS, SCADA — they all continue to operate. IRIS Foundry adds the data governance, AI, and intelligence layer on top without disrupting the systems your operations depend on.
IRIS Foundry supports edge, hybrid, and cloud deployments natively. AI models — predictive, generative, and agentic — run at the edge for latency-sensitive or OT-isolated sites, delivering real-time decisions at the machine without cloud round-trips. The container and API-driven architecture provides flexibility across any deployment topology, from a single plant to a global enterprise.
Through the IRIS Foundry MCP (Model Context Protocol) server. This exposes industrial intelligence — asset health, KPIs, diagnostics, workflows — directly into Teams and M365 Copilot as callable tools. Operators query live plant data, trigger diagnostics, and escalate alerts without leaving Teams. New capabilities added to Foundry are auto-discoverable in Teams, with no custom integration work required.
Scaling Industrial AI: From Pilots to Production Across the Global Value Chain