The chemicals sector has a talent gap. Putting humans in the lead unlocks productivity gains, says Stephen Reynolds, Industry Principal, Chemicals at AVEVA.
When a pre-sales engineer on our team used a generative artificial intelligence (AI) assistant to create a company briefing about a chemicals partner ahead of a recent meeting, that first draft was surprisingly accurate.
After validation and some correction, we used it. That’s where industry is with AI today: astonishing capable assistants that depend on humans in the lead.
AI tools have quickly begun to take care of the grunt work, enabling us to work faster and smarter. Yet, human acumen remains integral to final approval. That’s for both regulatory and safety reasons, as well to bring real-world awareness to bear on decision making.
For today’s complex and intertwined chemical operations, then, the engineer or plant operator must remain in charge.
Industrial AI’s role is to shorten decision cycles with quicker operational insights. Whether adjusting processes to changing feedstock levels, optimising maintenance schedules, or bridging the chasm between information, operations and enterprise technology, industrial AI streamlines decisions so plants stay humming.
Efficient assistants that handle the heavy lifting in chemicals
In practical terms, AI’s applications for the chemical sector fall into two buckets: efficiency drivers and digital assistants.
On the efficiency side, AI models maximise the potential of an organisation’s data.
Chemical manufacturers and producers have no shortage of industrial data today, but all too often this valuable base substrate—to borrow from ancient alchemists—is locked up in silos.
To be transformed into usable gold, data feeds across the enterprise must be contextualised and connected up, such as into a digital twin. This is where AI models and agents are functional catalysts.
They sift through years of historian data from continuous or batch units to spot patterns indicating fouling, drift quality or heat leaks—well before alarms are triggered. The control-room operator still decides whether to change a setpoint or request a shutdown – but is now armed with focused data and recommendations.
As an assistant, AI handles the heavy lifting, easing operators’ burden.
By synthesising information from across an organisation’s operational ecosystem into context-specific, industrial AI tools deliver actionable insights from shopfloor to top floor within a single pane of glass.
With the kind of comprehensive, high-speed data visibility AI can deliver in near real-time, engineers, managers and even sales teams have an authenticated source of industrial intelligence that clarifies decision trees and shortens innovation cycles.
These could be meeting briefs that equip teams with valuable insights, or comprehensive plant analytics that surface new asset correlations. Consequently, humans can focus on what we’re good at: incisive judgement.
AI assistants lighten teams’ analytical workloads, reduce rework and slash manhours. Efficiency solutions, including AI agents, close gaps between opportunities and action. In both cases, humans stay in the lead, but some of the groundwork is automated.
Reduced downtime and fewer critical asset failures
Teams at chemicals major Methanex experienced these gains firsthand. With production sites in six countries, the world’s largest supplier of methanol knew AI could provide visibility into the tremendous amount of operational data it produces on a regular basis.
It turned to a cloud-based industrial intelligence platform to bring together operational data with customised AI solutions. With AI handling the heavy lifting, Methanex teams were able to take quick, empirical decisions in lockstep with new market opportunities—across departments and national borders.
Non-technical strategic users gained the most, because they can pull up consolidated data across different parameters at any time. Costly downtime has been reduced and critical equipment failures are fewer in number, while efficiency has increased across the board.
Fine-tuning business models around AI-assisted operations while keeping humans in the lead meet the moment in more ways than one. Besides gains such as real-time supply chains adjustments and shorter business cycles, human-AI operations close the chemicals talent gap at a time when experienced operators are set to retire faster than we can replace them1, and new demand could boost compound annual growth rates by 11% to 2028.2
Smaller talent gaps when AI helps human workers
About 31% of chemical industry manhours could be reduced by augmentation and automation as a result of generative AI, the subset behind intelligent AI assistants.3 Industrial AI solutions – such as predictive analytics and asset optimisation – will produce further gains.4
The sector seems slow to respond, even as the majority of younger workers already use and expect to leverage AI at work.5 BCG reported in September that chemicals and machinery firms plan to upskill fewer than 15% of their staff over the coming year – as against 55% in the software sector.6
That’s despite AI’s ability to support leaner manufacturing. Work that once involved large workforces can now be executed by the fewer trained operators available, because digital systems cover more of the monitoring and analytics.
Training is an obvious – and related – example. AI and virtual reality are already supporting process training without exposing new recruits to live plant risks, such as with BASF’s Mobile Operator.
Three categories to maximise AI value towards Industry 5.0
We’re headed to a future where intelligent machines work alongside humans; a development referred to as Industry 5.0. As a practical starting point, look to designing AI-supported value chains with humans in the lead. McKinsey data shows that the companies realising significant revenue from AI programs are far more likely to have formal processes governing when model outputs require human validation: 65% versus 23% of others.7 In the chemicals sector, that governance layer bolsters deployment in high-hazard environments. Human-certified engenders customer confidence.
When looking to AI implementations, a useful exercise is to map tasks into three categories. First, where agentic AI can act autonomously within strictly pre-defined limits.
A second category grouping would see AI-enabled tech propose operations for human supervisors to confirm. Finally, a third class of tasks would see AI provide information without direct control.
But success in each area goes beyond treating AI pilots as technology experiments – a guaranteed pathway to join the 95% failure statistic.8 As digital and analytics tools are being adopted, companies need end-to-end approaches to turn analytics into operational improvements, as Deloitte pointed out in its recent chemical sector outlook.9
Such a breakdown of expectations can go a long way to avoiding the pilot purgatory arising from poorly executed AI initiatives.10 Only a stepwise approach for human-AI teams, backed by clean data and cultural change can move AI from theoretical concepts to real operational and R&D outcomes.
In a sector where a single poor decision can affect safety, sustainability and profit, industrial AI should be implemented so that it supports human expertise, makes work more attractive to a new generation and defers to the realities of high-hazard operations.
Systems that do more of the contextualising and analysis, with people deciding what happens next – that’s what it means to have humans in the lead.
REFERENCES
Elser, B et al, Accenture: Cracking the labor productivity conundrum in the chemical industry.
Elser et al; Accenture: Where’s the money in sustainability for chemical companies?
Shook, E & Daugherty, P; Accenture: Work, workforce, workers
Mori, L et al, McKinsey: How AI enables new possibilites in chemicals
The Harris Poll & Google Workspace: New research from Google Workspace and The Harris Poll shows rising leaders
Apotheker, J et al (BCG): The Widening AI Value Gap are embracing AI to drive impact at work (Press release)
Singla A et al (McKinsey): The State of AI in 2025
Estrada, S (Fortune): MIT report: 95% of generative AI pilots at companies are failing
Yankovitz, D (Deloitte): 2026 Chemical Industry Outlook
Schuman, E (CIO): 88% of AI pilots fail to reach production — but that’s not all on IT
