AI for Energy Teams, Starting Nowhere Near a Safety Case

Nothing here touches a control system or a safety case. That is deliberate. Prove the method where the stakes are administrative, then earn the right to move closer to operations. ## 1. Read your own maintenance history Take three years of work orders and failure records for one class of equipment. Look at what actually failed, what the precursor readings were, and how often a planned intervention prevented an unplanned one. Most organisations doing this properly for the first time discover two things. A small number of failure modes account for most of the downtime, and the warning signs were usually present in data somebody already had. You do not need a platform for this. You need the analysis done once, honestly, on your own records. ## 2. Compliance and reporting documentation Regulatory submissions, Nigerian content reporting, permit renewals, tender responses. Large volumes of structured writing that follows known formats and consumes skilled people. Draft generated, expert reviewed, never submitted unread. The time saved is immediate and no operational risk is involved at any point. ## 3. Consumption anomalies Fuel and power consumption that does not match output is a detectable pattern, whether the cause is a failing asset, a metering problem or a loss. Put consumption against production for one site over twelve months and look at where the relationship breaks. This is arithmetic before it is machine learning. Do the arithmetic first. ## Before any of them Write down which AI tools are approved and what data may go into them, and apply it to the commercial floor as well as the operational estate. Your engineering side already lives under change control. Your commercial side, in most companies in this sector, has no rule at all, and it is holding reserve information, partner terms and pre disclosure production data. That is BROAI, and in this sector the exposure is market sensitive rather than merely awkward. This is the Scan and Score phase of SWITCH, the framework I use with clients. Pick one. Give it an owner and a number. Thirty days, then keep it or stop. If you want this worked through with your own assets, data and regulatory obligations rather than as a generic session, the [speaking and training page](/speaking) sets out the formats, or reach me at me@martins.com.ng or through the [work with me page](/work-with-me).