The foundation of autonomy matters most. Organizations need a unified data architecture - one that can connect legacy OT systems, contextualize data through knowledge graphs and make it accessible across control, plant and enterprise layers.
Why 2026 Will Redefine Manufacturing Autonomy & What Leaders Must Prepare for Now
Q&A with Rahul Negi, Executive | Honeywell Process Automation
You’ve described 2026 as the point where autonomy becomes an operating model rather than a pilot. In asset-intensive sectors like oil and gas, what does that shift look like in practice – whether in upstream operations, refining or LNG facilities?
In 2026, autonomy stops being an overlay and becomes embedded in how operations are run. In the oil and gas value chain, AI systems are continuously evaluating process conditions, recommending adjustments and, within defined limits, executing them. In refining, for example, advanced algorithms can dynamically adjust fuel blends, temperatures and flow rates across interconnected units, analyzing thousands of variables per second. The shift is from reactive control to self-adjusting systems that operate against optimal benchmarks.
What makes this possible is physical AI – connecting real-time operational data with decades of contextual plant knowledge. Platforms like Honeywell Forge and Experion Cognition bring together edge, cloud and agentic AI so decisions improve over time. Autonomy is embedded at the control room, plant and enterprise level, as standard operating procedure.
For example, at Chevron’s El Segundo refinery, AI-powered alarm guidance gets activated when a critical alarm is limit is reached and provides recommended actions directly within the control console, ultimately reducing alarm fatigue and improving response precision. At ADNOC Borouge, Honeywell piloted a fully autonomous control room capable of operating a furnace and gas redistillation unit specific failure scenarios without human intervention, while staying within defined safety parameters.
As autonomy increases, how are leading organizations redesigning workflows and decision-making between field operators, control room teams and centralized digital centers?
The biggest redesign is redistributing cognitive load. In leading organizations, routine and time-sensitive decisions are being handled by agentic AI, while humans focus on oversight, exceptions and higher-level judgment. In the control room, systems like Experion Cognition will act as a reasoning engine, comparing live data against digital twin models and historical performance, then presenting prescriptive guidance. Operators move from monitoring alarms to supervising outcomes, which reduces fatigue and improves consistency. At TotalEnergies, next-generation operator guidance leverages pattern recognition and historical data to detect early warning signs and recommend recovery steps before abnormal situations escalate.
At the plant level, tools like Honeywell Production Intelligence and Asset Performance Management connect OT and IT data streams, so field teams receive guided workflows and predictive maintenance recommendations in real time. In centralized operations centers, executives gain enterprise visibility and contextualized intelligence that supports cross-site decisions. The workflow becomes a closed loop where AI analyzes, recommends and acts, while humans validate, refine and escalate when needed. That partnership is what enables autonomy to scale responsibly.
When automation scales across complex, high-risk operations such as petrochemical plants or offshore platforms, how should executives rethink ROI – particularly around safety, uptime and risk mitigation?
Traditional ROI models focus heavily on throughput gains or cost reduction. In high-risk environments, that’s too narrow. Executives need to account for avoided losses, like unplanned downtime, safety incidents, environmental penalties and cyber vulnerabilities. For example, predictive asset management in offshore operations can prevent equipment failures that would otherwise cost millions per day in lost production. That risk-adjusted value often outweighs incremental efficiency gains.
There’s also workforce ROI – in many process plants, up to 90% of operational errors are tied to human variability due to skill gaps. Digital cognition reduces that variability by providing contextual recommendations and automated safeguards, which translates into more stable operations and fewer catastrophic surprises.
This is why autonomy should be evaluated over the lifecycle. Systems that continuously learn and optimize create compounding returns where safety improvements, energy efficiency gains and maintenance savings build year over year. The right lens is resilience and long-term value creation.
For facilities that have moved beyond proof-of-concept, what performance benchmarks are emerging – whether in throughput, energy efficiency, predictive maintenance or emissions management?
Facilities that have scaled autonomy are benchmarking against optimal parameters drawn from global performance data. In refineries, we’re seeing tighter control over fuel blending and process variability, driving measurable throughput improvements without sacrificing safety. In offshore energy operations, AI-enabled asset performance reduces downtime and extends equipment life.
Energy efficiency is another area where we’re seeing significant performance benchmarks, like continuous optimization of temperatures, flow rates and pressure reduces energy intensity and associated emissions. The key benchmark is output per unit of energy and risk. The common thread is moving from reactive maintenance and static thresholds to dynamic, continuously improving performance baselines.
As companies plan for 2026 capital cycles, what foundational investments are critical to sustain autonomy in heavy industrial environments – from data infrastructure and cybersecurity to edge computing in remote sites?
The foundation of autonomy matters most. Organizations need a unified data architecture – one that can connect legacy OT systems, contextualize data through knowledge graphs and make it accessible across control, plant and enterprise layers. Without that visibility, AI models operate in isolation. Edge and cloud infrastructure must work together as well. Remote sites like offshore platforms and LNG terminals require edge processing for real-time responsiveness, while the cloud provides enterprise-scale analytics and continuous learning. Cybersecurity also has to be embedded from the start. As OT becomes more connected, the attack surface expands. Industrial AI must be built with secure connectivity and governance in mind.
And finally, domain expertise is critical – AI models only create value when grounded in operational reality. Honeywell’s decades of experience across energy and industrial sectors – connecting more than 100 million assets globally – enable benchmarking and optimization that generic AI providers simply can’t replicate. Companies that treat autonomy as infrastructure will be the ones positioned to sustain it.
Rahul Negi is an executive at Honeywell Process Automation leading Digitalization, Autonomous, and AI initiatives, with over 20 years of expertise in Strategy, Consulting, Business Development, and AI/ML. His leadership in Honeywell’s strategic collaborations with industry leaders such as Chevron, Total Energy, and ADNOC and others has resulted in groundbreaking AI/ML deployments in Operational Technology layer. These innovative solutions have significantly enhanced safety, operator productivity, and asset reliability, driving substantial improvements in operational efficiency and reduced downtime. Additionally, he has established a Digital Consulting practice that crafts tailored digital transformation roadmaps towards autonomous operations including AI/ML technologies, aligning client objectives and vision with real life gains and ROI.
The content & opinions in this article are the author’s and do not necessarily represent the views of ManufacturingTomorrow
Featured Product
