MANUFACTURERS have never had greater visibility into the condition of their assets. Predictive maintenance systems can detect changes in vibration, temperature and performance long before equipment reaches the point of failure, while connected machines generate a continuous stream of operational data that can be used to monitor asset health. 

Despite these advances, unplanned downtime continues to place significant pressure on manufacturers. Recent research by Fluke found that 68% of UK manufacturers experienced unplanned downtime during the previous year, with the total cost estimated at up to £736m every week. Lost production, overtime, expedited deliveries and wasted materials all contribute to the financial impact, demonstrating that seeing a problem coming is only part of the challenge. 

That gap between having the right data and making it usable is where many operational teams still lose time. Pulling together context from multiple systems quickly enough to make confident decisions and execute work at pace creates unnecessary friction. 

Operational friction: The real constraint

A maintenance alert may indicate that a critical pump is showing signs of deterioration, but it does not decide the best response. Engineers still need to review maintenance histories, check spare parts availability, assess production schedules, and work out whether the job fits an existing shutdown window. 

In theory, digital investment should make this easier. But the information needed to answer those questions often remains dispersed. Maintenance histories may sit within a CMMS, production schedules within an ERP platform, machine data within SCADA or MES systems, while technical documentation and engineering drawings are stored elsewhere. 

This issue shows up anywhere operational teams need to move from signal to execution, such as when carrying out root cause investigations or assembling compliance documentation. Collating data can seem like a small task when done once, but over time it steals availability from higher-value engineering work. 

Bringing operational context together 

Instead of simply retrieving information, an AI agent can coordinate context, systems and actions across a workflow, while keeping people firmly in control of decisions. The goal is to reduce the friction between insight and execution by bringing together what an engineer needs to decide and act on, then helping coordinate what needs to happen next. 

Consider a recurring seal failure on a production pump. Rather than manually searching maintenance records, reviewing condition monitoring trends, checking inventory levels and confirming upcoming production schedules, an AI agent could assemble that information automatically. It could surface related work orders, highlight recurring failure patterns, and present the operational context needed before an engineer starts chasing information across multiple systems.

Agentic AI can even streamline the entire process, from drafting a work pack, identifying the right procedures and drawings, and flagging dependencies that commonly slow execution. 

Human-agent collaboration that augments expertise 

For manufacturers facing skills shortages and an ageing workforce, improving access to operational knowledge is becoming increasingly important. Make UK has identified skills shortages as one of the biggest challenges facing UK manufacturers, while experienced engineers continue to retire with decades of practical knowledge. 

Much of that expertise already exists inside maintenance histories, engineering documentation and historical work orders. But it is not always accessible quickly enough when decisions need to be made. When human–agent collaboration is done well, engineers spend less time assembling context and more time applying expertise in evidence assessment and planning work with confidence. 

When operational decisions rely on a small number of experts, bottlenecks build fast. Better access to context helps spread capability across teams without lowering standards or delegating responsibility to a black box. 

Taking prediction into execution 

Able to make and execute decisions faster and to a higher standard, the AI-enabled companies of the future have been dubbed "Frontier Firms". Building on predictive maintenance practices which tell you something needs attention, agentic AI pulls together context at speed to help organisations decide, coordinate and act. 

None of this removes the need for reliable data, clear governance and well-defined processes. As with any digital initiative, organisations will get the best results by starting with a specific workflow where friction is obvious and measurable (for example, maintenance planning or root cause investigations), then expanding into broader operational use cases as confidence and controls mature. 

Predictive maintenance continues to transform manufacturers’ ability to identify developing equipment issues. The next stage of technology evolution is reducing the delay between detection and delivery. 

Most maintenance teams already have the data, but they’re unable to bring it together quickly enough to make confident decisions and keep work moving under pressure. The organisations that use agentic AI to cut friction and increase engineers’ control will be in a stronger position to improve reliability, minimise unnecessary downtime, and get more value from scarce engineering expertise. 

Ali Mustoe-Playfair is director of agentic operations at ANS

For more information:

www.ans.co.uk

Tel: 0800 458 4545