Definition
Operator Manual Competency Retention Index, or OMCRI, is the change in operators'' and engineers'' demonstrated ability to execute critical manual procedures over time as AI and automation handle more routine operations.
The Measurement Gap
Camila is the operations training coordinator at a Gulf Coast refinery that has been on its current generation of advanced process control for almost a decade. The site has invested heavily in AI-augmented operator advisory systems over the last four years, a layer above the APC that recommends operator actions, automates routine envelope adjustments, and provides natural-language explanations of process state to the operators on shift. The system has been popular with operations. Console workload is down, the rate of routine adjustments operators have to make per shift has dropped substantially, and the kind of constant fine-touching the units used to need has largely been delegated to the AI. The unit operators on the recent technology generation came up through the AI-augmented era, and they are skilled at supervising the AI''s recommendations and at responding to its alerts.
The drill was a tabletop exercise on a fictional but realistic upset scenario: a complete failure of the FCC''s advanced control package, with the unit reverting to basic regulatory control, during a period of unusual feed slate flexibility. The board operator on duty for the drill was a competent, experienced operator with twelve years on the unit, eight of them under progressively more capable AI advisory. The drill ran for forty minutes. The board operator handled the initial alarm cascade well (the regulatory loops were familiar) but began to slow as the drill''s scenario required the operator to manually balance reactor severity, catalyst circulation, and product cut points without the advisory system''s recommendations. The decisions the operator made were defensible but conservative. Margin would have been lost. By the end of the drill, the unit was stable but operating well below its potential output, and the operator was visibly working harder than the routine pace of an AI-augmented shift would have produced.
The post-drill review was instructive. The senior operator who facilitated the drill (a thirty-year veteran who pre-dated APC and had managed many cold starts and emergency shutdowns from the manual era) pointed out that the drill scenario was exactly the kind of scenario the unit''s most experienced operators would have handled in their sleep fifteen years ago. The board operator was capable; the muscle memory was not. The decisions he had made were grounded in textbook knowledge that he had been taught and never practiced. The instinct that comes from running the unit yourself, day after day, was not in him because he had never run the unit himself for any extended period. The AI had run it for him.
Camila pulled the operator workforce demographics. About forty percent of the current board operators had less than five years on the unit and had only ever known the AI-augmented environment. About thirty percent had five-to-twelve years and had seen the AI introduced during their careers but had operated pre-AI for at least a few years. About thirty percent were senior operators with substantial pre-AI experience. The drill performance correlated cleanly with the cohort split. Senior operators handled the manual scenario competently if a little rusty. Mid-career operators struggled more but recovered. The newest cohort showed the largest gap between their textbook competence and their operational ability to execute manual procedures under time pressure.
Legacy operator competency metrics do not catch this. Training completion is at one hundred percent. Annual qualification testing is at one hundred percent. Safety performance is excellent: the AI is doing its job. The unit''s normal operations are smooth. The skill the workforce is losing is the skill that surfaces only when the AI cannot do its job, which is by design a rare event. The competency drift is silent until the event that requires it arrives.
What is missing is a metric that asks whether the workforce can still execute the manual procedures the safety case assumes they can execute when the AI is unavailable.
Formula
Operator Manual Competency Retention Index, or OMCRI, is the change in operators'' and engineers'' demonstrated ability to execute critical manual procedures over time as AI and automation handle more routine operations. The KPI exists to answer one decision: is the workforce''s ability to handle emergencies, novel situations, and manual fallback operations being eroded by the AI''s presence in ways the operator''s safety case has not accounted for?
The structural form is a longitudinal performance measurement against a defined baseline. For each critical manual procedure (emergency shutdown sequences, manual well control responses, cold startup procedures on process units, manual pipeline valve operations, hand control of artificial lift during automation failures) the workforce is assessed periodically through simulator exercises, field drills, or structured tabletop exercises. The assessment produces a performance score against established benchmarks. OMCRI is the percentage change in average score from a baseline period to the current period. The baseline is the first assessment under the program or, where the data exists, a pre-AI-introduction benchmark. Negative OMCRI is atrophy; positive OMCRI is improvement.
The assessment method is the methodological core. A paper-based test of procedural knowledge will produce a score, but the score does not measure the operational competency the metric exists to track. The score has to come from an assessment that approximates the actual cognitive and procedural load of executing the procedure under realistic conditions. Simulator-based assessments are the strongest option where simulators exist. Field drills with the unit''s safety case in the loop are the second-best option for units where simulators are not available. Structured tabletop exercises with senior facilitators are workable for procedures that cannot reasonably be drilled in either form. The assessment design matters more than the metric design, a sloppy assessment will produce a confident-looking OMCRI that measures nothing useful.
The critical-procedure inventory is the second part of the metric''s honesty. Not every procedure the operations workforce can execute belongs in OMCRI. The right list is the procedures whose execution is required for the safety case to hold during AI failure or unavailability. Emergency shutdown sequences, well control responses to a kick, manual cold-startup procedures, emergency manual valve operations on critical relief paths, manual overrides of artificial lift, and similar procedures are typical. The inventory should be reviewed by process safety and operations leadership together, with each procedure''s inclusion defended against the safety case.
Tenure segmentation is essential. The workforce is not uniform; experienced operators who worked extensively pre-AI carry skills the AI-era cohort never developed. OMCRI computed as a workforce average will hide the cohort where the atrophy is most severe. The metric should be reported separately for at least three cohorts: senior operators with substantial pre-AI experience, mid-career operators who saw the AI introduced during their tenure, and post-AI-era operators who have only known the AI-augmented environment. The cohort that is growing (the post-AI cohort, as senior operators retire and new hires come in) is the cohort whose OMCRI trajectory governs the long-run workforce capability.
Proposed formula, pending validation pilots.