Analytics that continuously reassess a safety instrumented system's real-time risk reduction — factoring in bypasses, degraded redundancy and process conditions — rather than relying solely on the static SIL calculated at design time.
Control Systems — DCS · PLC · ESD · HMI
The DCS, PLC, safety and HMI/SCADA layer — where control logic executes, and where AI is arriving fastest.
The control system executes the logic that keeps a plant stable, efficient and safe: the distributed control system, programmable logic controllers, the safety instrumented system, and the operator interface and SCADA layer above them. Its architecture has been stable for a generation, with controllers, engineering tools and networks supplied together as one package.
That package is now being unpicked. Control software is separating from hardware, networks are converging on standard deterministic Ethernet, and data is being opened to other systems through common models. At the same time artificial intelligence is arriving at every level, from engineering tools to operator assistance to the control loop itself. Entries that touch safety systems move slowest, for good reason.
Full analyses in this category
- Software-defined / virtualised DCS & PLC
Control software that runs on standard servers or industrial PCs instead of proprietary controller hardware. The first open-architecture plant and server-based controllers from mainstream suppliers are now in production; the long-term support model is still unproven. - Cloud-native / SaaS SCADA & HMI
SCADA, HMI and historian functions hosted in a cloud region instead of on servers in the plant. Proven for monitoring distributed assets such as water networks and well pads; for closed-loop control of a process unit the local control system remains the place to be. - Unified Namespace & MQTT Sparkplug B architectures
An architecture in which plant systems publish their data once, to a shared event-driven namespace, instead of being wired together pair by pair. The messaging standards are settled; the governance needed to make it work across an enterprise is not. - AI-driven closed-loop advanced process control (APC 2.0)
Machine learning and reinforcement learning applied to closed-loop process optimisation, alongside or above classical model predictive control. Specific commercial results exist; most of the evidence still comes from suppliers and their early customers.
This category has 12 entries in the October 2026 edition, of which 6 are AI/ML-driven. They are listed below by stage, from earliest to most mature. Entries marked "Full analysis" have a sourced, in-depth page; the others are short entries.
Emerging Idea (4)
Entries at the Emerging Idea stage have a working concept and early demonstrations, but few or no production deployments. Evidence at this stage comes mainly from pilots, research and supplier roadmaps.
AI agents that don't just recommend a setpoint change but are given standing authority to execute a bounded set of control actions autonomously.
An operator interface that answers plain-language questions ('why is the reactor trending high?') by querying live process data, alarms and procedures, rather than requiring navigation through faceplates and trend screens.
AI assistants that search across historian data, alarm history, maintenance records and equipment manuals to propose a likely root cause for an abnormal situation — distinct from a general conversational HMI in that they're built specifically for diagnostic reasoning, not open-ended Q&A.
Over-Promise (4)
Entries at the Over-Promise stage are receiving more supplier and media attention than their deployed base yet justifies. Products exist and early adopters are live, but independent evidence of results is still thin.
A single, event-driven data model that every system — DCS, MES, historian, analytics — publishes to and subscribes from, replacing point-to-point integrations between each pair of systems.
Graphical, template-driven configuration environments intended to let process and automation engineers configure standard control strategies without hand-writing IEC 61131-3 logic.
Control logic execution decoupled from proprietary controller hardware and run instead as software on standard, virtualised compute — scaled, patched and failed-over like any other enterprise application.
Generative-AI assistants embedded in engineering tools that draft PLC/DCS logic, translate P&IDs into control narratives, or answer natural-language questions about an existing configuration.
Reality Check (2)
Entries at the Reality Check stage have been through a round of over-promising. Early projects under-delivered, attention has moved on, and what remains is being re-scoped around the uses that work.
Time-Sensitive Networking standards combined with OPC UA FX give standard Ethernet the deterministic, low-jitter behaviour that motion and safety control traffic requires.
Virtual and augmented-reality environments built on a plant's digital twin, used to train operators on startup, shutdown and abnormal-situation procedures without touching the live process.
Industry Acceptance (2)
Entries at the Industry Acceptance stage have repeatable reference deployments and a reasonably well-understood business case. The open questions are about integration, cost and scale more than about whether the technology works.
Machine-learning-based advanced process control that continuously retrains against live plant data, rather than the fixed linear models of first-generation APC.
SCADA and HMI platforms delivered and licensed as a cloud service, with the visualization and historian layer hosted off-premises rather than on a local server.
Stage and time-horizon placements are editorial judgements based on public information as of October 2026. See how placements are made.