PTQ Q4 2026 Issue

More documentation is not automatically better governance. The system should capture the decision as the work hap- pens and present the next shift with the minimum context needed to act safely (see Figure 1 ). Trends should come from the historian. Equipment status should come from the maintenance record. Active permits and isolations should come from control of work. The oper- ator should add judgement, not reproduce data that already exists elsewhere. This also matters for workforce transition. Refineries depend on knowledge built over years of starts and stops, feed changes, equipment failures, and near misses. Training transfers part of that knowledge. A gov- erned exception history includes another part: the conditions that prompted escalation, the chosen response, and whether it worked. Rather than creating a massive and rigid rule book, a good judgement that is easier to inherit should be made. Closure must improve the next response Many processes treat closure as the end. For exception man- agement, closure should also improve the next decision. If an AI-generated anomaly was rejected, the reason matters. The signal may have duplicated an existing alarm, lacked operat- ing context, or described a condition that was technically real but not actionable. If the recommendation was accepted, the refinery should know whether the intervention changed the outcome. Without that feedback, the model produces predictions but does not learn the operating value of those predictions. The same applies without AI. Repeated temporary repairs may indicate that a planning or capital decision is being deferred. Several off-specifica - tion cases with the same contributing condition may justify a procedure change or control review. Recurring alarm-related exceptions may send the site back to the alarm rationalisa- tion phase. A cluster of overdue inspection actions may expose a resource or decision-right problem rather than an inspec- tion problem. Good closure asks two questions. Is this case resolved? What should change because we have seen it? Compress the waiting, never the control The most useful measure in exception management is the exposure window: the time between detecting a material deviation and verifying a safe, effective response. The win- dow remains open for these reasons. Engineers need enough information to make a sound deci- sion. Inspection may require access and preparation. A lab- oratory confirmation takes the time required by the method. Permits and isolations must be correct. MOC must address the technical basis and consequences of a change. Windows may be open due to long wait times. A request sits in an inbox. A reviewer does not know that the matter is on a critical path. Operations and maintenance assess the same issue separately. A decision is pending until the next daily meeting, as no alternate authority has been defined. The night shift leaves context for the morning shift, but not an accountable case. Exception management compresses this waiting time. It must never compress the window by skipping a safety-critical

crosses several of them. Consider an off-specification prod - uct result. A resolution must involve the laboratory, opera- tions, process engineering, planning, and commercial teams. Each function can do its own part well. The delay appears between the parts. Who owns the overall clock? Which result is current? Has the affected material been placed on hold? What decision releases it? Has the next shift received the same operating context? People often carry this work manually. They copy readings into messages, chase reviewers, reconcile different versions, and rebuild the history when the matter is audited. The sys- tems hold the records, but the people carry the process. The inversion should be true. Systems should carry the work, not make people work for them. This means adding a governance layer across core systems. An exception opens from the source signal. It receives an owner. The relevant context travels with it. Required reviews occur in the correct order or in parallel where the operating rules allow. Missed response times escalate. Decisions return to the relevant systems of record. Closure requires evidence. The technology is useful because it reduces reconstruction and ambiguity. The operating rules remain in authority. Shift handover is where continuity is tested The quality of an exception process becomes visible at shift change. A well-run handover does more than transfer a list of open items. It transfers the state of each important decision. What was observed? What has been ruled out? Which safeguard is experiencing higher-than-usual demand? What work is authorised? Which condition requires the next shift to stop, escalate, or change the plan? Who owns the next action, and when is it due? Experienced operators carry this context naturally. They know which deviations are stable, which are deceptive, and which can deteriorate quickly when the operating mode changes. The risk arises when the context remains with the people, and they eventually retire. The answer is not to force operators to write longer notes. Figure 1 The integration of digital monitoring and data ana- lytics in modern energy infrastructure

PTQ Q4 2026