IT–OT Convergence, Agentic AI & the Autonomous Factory
<From Connected to Self-Orchestrating: A Strategic Blueprint>
Analyst predictions from IDC, Gartner, and ARC Advisory on the convergence of information and operational technology; how agentic AI transforms shop-floor autonomy; the 15-accelerator Digital Accelerator Portfolio for achieving autonomous manufacturing; and a practical KPI framework for measuring progress.
The factory of 2030 will not merely be “connected.” It will sense, reason, decide, and act — autonomously. The structural prerequisite is the convergence of Information Technology (IT) and Operational Technology (OT) into a unified, semantically rich data fabric. The intelligence engine is agentic AI. And the strategic accelerant is a modular portfolio of pre-built digital components that compress the journey from Level 1 monitoring to Level 5 full autonomy.
1. The Convergence Imperative: Why IT–OT Unification Is the Foundation
For most of the Industry 5.0 era, IT and OT have evolved on parallel tracks. IT systems — ERP, PLM, CRM, SCM — manage transactional and engineering data in cloud-native architectures. OT systems — PLCs, SCADA, DCS, historians — govern real-time machine control in deterministic, edge-heavy environments. The result is a “data chasm”: rich engineering context in IT cannot inform shop-floor decisions in OT, and real-time production telemetry cannot feed back into design and planning systems at enterprise speed.
This chasm is no longer tenable. Gartner’s 2025 research on manufacturing digital-twin adoption concludes that “organisations that fail to bridge the IT–OT semantic gap by 2027 will be unable to realise more than 20% of the value from their digital-twin investments.” IDC’s FutureScape for Manufacturing 2025 identifies IT–OT convergence as the single most critical infrastructure prerequisite for autonomous operations. ARC Advisory Group’s 2024 analysis further quantifies the penalty: plants with siloed IT and OT architectures experience 2.3× longer mean-time-to-resolution for quality escapes compared to converged environments.
Convergence does not mean collapsing IT and OT into one stack. It means building a semantic integration layer — an ontology-driven data fabric that translates between transactional records (ERP), execution instructions (MES), and machine signals (OT historians/SCADA) in real time, preserving context, lineage, and trust scores across every handshake.
Core Principle
IT–OT convergence is not a networking project. It is a semantic interoperability initiative: the ability for every system — from CAD workstation to PLC — to share meaning, not just data packets.
2. Analyst Predictions: IT–OT Convergence & Autonomous Manufacturing (2025–2028)
The convergence trajectory is backed by hard analyst forecasts. The following predictions from IDC, Gartner, and ARC Advisory carry the most direct implications for manufacturing leaders.
IDC Futurescape: 65% of G2000 manufacturers will deploy unified IT–OT data platforms by 2027, up from under 20% in 2024, driven by digital-twin and AI requirements.
Gartner: 50% of industrial AI deployments will shift from centralized cloud inference to edge-native agentic architectures by 2027, enabling sub-second autonomous decision loops.
ARC Advisory: 2.3x longer mean-time-to-resolution for quality escapes in plants with siloed IT–OT vs. converged architectures, based on 180-plant benchmarking study.
3. Agentic AI: The Intelligence Engine for Autonomous Operations
If IT–OT convergence builds the data highway, agentic AI provides the drivers. Unlike conventional AI — which responds to queries or classifies inputs — agentic AI systems perceive, reason, plan, and act autonomously within defined guardrails. In a manufacturing context, this means AI agents that don’t just detect a quality anomaly; they diagnose the root cause, evaluate corrective options, propagate the change across PLM/ERP/MES, and adjust the production schedule — all without human intervention for pre-approved scenarios.
The agentic AI paradigm introduces three architectural shifts critical to the autonomous factory:
Multi-agent orchestration: Rather than a single monolithic model, the factory deploys specialised agents — a quality agent, a scheduling agent, a maintenance agent, a compliance agent — that collaborate through a shared ontology and message bus. Each agent is domain-expert; the orchestration layer resolves conflicts and enforces priorities.
Closed-loop decision authority: Agentic AI compresses the sense–analyse–decide–act loop from hours (human-in-the-loop) to seconds (human-on-the-loop) to milliseconds (full autonomy) depending on the decision’s criticality and the organisation’s maturity level.
Continuous self-improvement: Agents learn from every decision outcome. A predictive maintenance agent that incorrectly predicts a bearing failure updates its own model parameters and shares the learning with the digital-twin sandbox for cross-validation — creating a virtuous feedback loop that compounds accuracy over time.
Key Insight
Agentic AI is not “automation on steroids.” It is a fundamentally different operating model: autonomous decision-making within governed boundaries. The governance framework — which decisions an agent may make unilaterally vs. which require human approval — is the most critical design decision in any autonomous-factory programme.
4. Leading Companies Building the Autonomous Factory
A select group of manufacturers is already operationalising the convergence of IT–OT and agentic AI at production scale. These are not pilots — they are strategic programmes reshaping how factories operate.
Siemens (Amberg Electronics Plant): Operates one of the world’s most advanced autonomous factories: 75% of production steps handled without human intervention. Unified Teamcenter PLM + MindSphere IoT + SIMATIC MES architecture with edge AI agents managing real-time quality and scheduling decisions.
Bosch (Industry 4.0 Flagship, Blaichach): Deployed a converged IT–OT data lake spanning SAP ERP, Bosch IoT Suite, and Nexeed MES. Agentic AI agents autonomously adjust ABS/ESP production parameters based on real-time SPC data, achieving 25% reduction in scrap rate.
Schneider Electric (Le Vaudreuil, France): World Economic Forum “Lighthouse Factory”: converged EcoStruxure IT/OT platform with AI agents managing energy optimisation, predictive maintenance, and autonomous batch scheduling across 50+ production lines.
5. The Digital Accelerator Portfolio: 15 Components for Autonomous Manufacturing
Achieving autonomous manufacturing is not a single-vendor purchase. It requires a modular, composable architecture where specialized components can be deployed individually for quick wins or compounded together for full-stack autonomy. The Infosys Digital Accelerator Portfolio provides exactly this: 15 pre-built, configurable accelerators spanning the ISA-95 stack from Level 1 (sensors/actuators) to Level 5 (enterprise integration).
15 pre-built configurable components/accelerators are as follows:
A1 – Auto I/O Discovery I A2 – Predictive Maint. Agent I A3 – Semantic OT/IT Bridge I A4 – Dynamic Scheduler
A5 – Quality Orchestrator I A6 – ERP‑MES Connector I A7 – Order Propagation I A8 – Field Feedback Agent
A9 – Ontology Toolkit (Backbone) I A10 – Semantic Int. Bus (Backbone) I A11 – Semantic Query Layer (backbone) I A12 – Digital Twin Sandbox
A13 – Agentic AI Framework I A14 – Maturity Assessment I A15 – Compliance Accel.
6. The Road to Autonomous: A Phased Perspective
The autonomous factory is not a single project. It is a multi-year transformation that progresses through five maturity levels:
L1 — Monitored: Sensors deployed, data collected, dashboards built. Human-dependent. Accelerators A1, A3, A9 lay the foundation.
L2 — Analysed: AI-assisted anomaly detection and root-cause analysis. Humans make decisions informed by AI recommendations. Accelerators A2, A5, A11 layer analytics.
L3 — Advised: AI agents propose specific actions (reschedule, contain, procure). Humans approve. Accelerators A4, A6, A7, A15 enable the advisory layer.
L4 — Autonomous (bounded): Agents act autonomously within pre-approved decision envelopes. Humans handle exceptions and edge cases. Accelerators A8, A12, A13 govern autonomous operation.
L5 — Self-Orchestrating: Full closed-loop autonomy. Agents coordinate across the entire value chain (design → production → field → design). All 15 accelerators compounded. Human role shifts to strategic oversight, policy setting, and exception governance.
The analyst consensus from IDC, Gartner, and ARC Advisory is clear: by 2030, Level 3+ will be the minimum competitive baseline for high-volume discrete and process manufacturers. Enterprises that begin the convergence journey now will reach L3–L4 by 2028; those that delay will find themselves structurally disadvantaged in both operational cost and ecosystem eligibility.
The factories that win the next decade will not be the most automated. They will be the most autonomous — and the difference is intelligence, not machinery.
Preferences
[1] IDC FutureScape: Worldwide Manufacturing 2025 Predictions. International Data Corporation (IDC), Doc #US51940624, October 2024. Covers IT–OT convergence as a prerequisite for autonomous operations, unified data platforms, and API-first MES architectures.
[2] IDC Spending Guide: Worldwide IT–OT Convergence Technologies Spending Guide, 2024. IDC. Projects $42B global spend on convergence technologies by 2027, including edge AI, industrial IoT platforms, and semantic middleware.
[3] Gartner, “Strategic Roadmap for Manufacturing Digital Twins, 2025.” Gartner, Inc., January 2025. Identifies the IT–OT semantic gap as the primary barrier to digital-twin ROI and projects 30% dark-factory adoption by 2028.
[4] Gartner, “Hype Cycle for Manufacturing Operations Strategy, 2024.” Gartner, Inc., July 2024. Positions agentic AI, edge-native inference, and composable MES within the manufacturing-operations technology landscape.
[5] ARC Advisory Group, “IT–OT Convergence: Benchmarking Quality and Operational Performance, 2024.” ARC Advisory Group. 180-plant benchmarking study quantifying the 2.3× MTTR penalty for siloed IT–OT architectures.
[6] Siemens AG, “Amberg Electronics Plant: A Showcase for Digital Enterprise.” Siemens.com, 2024. Documents 75% autonomous production, Teamcenter/MindSphere/SIMATIC integration, and edge-AI quality agents.
