WA organisations have more asset data than ever, and the question has moved from how much they hold to how much of it they are willing to act on. Engineers re-check dashboards against source reports before they sign off, and meetings often open by settling whose number is right. In one roundtable, 60% named fragmented systems as the single biggest barrier to trusted asset information. The constraint is less about technology than about having one version of the truth people believe.

Maintenance systems are very good at sending work out to the field and less good at learning from what comes back. Crews write the same correction on the same work pack and watch it return unchanged, and some stop writing because nothing visibly changes. Closing that loop is what separates a maintenance plan that improves every cycle from one that repeats itself.

Artificial intelligence has moved from conference speculation to operational reality, but most organisations remain caught between aspiration and execution. Nearly 45% have not yet started with AI, a further 44% are in pilot or early implementation, and 76% of projects fail to deliver expected returns. The most widely adopted application is generative AI for everyday knowledge and productivity tasks, with 52.5 % of respondents already using these tools. Predictive maintenance ranks second at 37.5 %. Workforce AI skills score lowest of any readiness dimension at 2.69 out of 5, while leadership ambition sits higher at 3.34 – a gap that needs closing. 85% express concern about younger engineers over-relying on AI outputs. The real opportunity, however, is substantial: rather than optimising existing processes for marginal gains, AI enables organisations to reinvent how maintenance works entirely – dynamic, responsive, and tied to actual operating conditions. Getting there requires investing in clean data, integrated systems, and workforce capability first.

Strong technical cases lose rooms every week, not because the engineering is weak but because the case arrives in a language the decision makers do not use. Practitioners described translating asset risk three ways — safety exposure, financial impact and production impact — and finding that asset health on its own wins very little attention. The same translation problem runs downwards, where a strategy written for auditors rarely reaches the people doing the work.

Self-service tools have put maintenance analysis in everyone's hands, and governance has not kept pace. Teams build their own views from the same raw tables and arrive at different answers, while crews keep recording data that no decision uses. The practical fix practitioners pointed to is ownership: who owns master data, who may publish a metric, and which version the business runs on.

AI is already inside WA maintenance teams, often informally — summarising overnight breakdowns, finding a specification in a manual, drafting the first version of a report. It earns its place in retrieval and summary, where a person can check the output quickly, and it is weakest when asked to judge data nobody has governed. The question practitioners are working through is how to take the gains without thinning the fault-finding skills of the people who will check the machine.

The knowledge people most fear losing is not the procedure. It is the reasoning behind decisions: why a strategy was set this way, what was tried before, and how to weigh a condition report against a production target at two in the morning. Procedures can be written down. Judgement has to be built, and it is built in the time that busy operations rarely protect.

Supervisors are where the maintenance system meets the day, and most get there because they were the best tradesperson on the crew. The skills the job then asks for are different: setting expectations, coaching, and holding one standard across crews who rarely overlap on a FIFO roster. Admin and reporting pull them away from the field, which is where that standard is actually held.

Maintenance leaders in WA now change roughly every fifteen months, which is shorter than most improvement programs take to show a result. Programs carried by one champion stall when that person leaves, and decisions get reversed by people who never saw the reasoning. What outlasts a leader is what they modelled: the fundamentals they held to, the behaviour they accepted, and how safe their people felt to speak up.

Record tonnes and record harvests are being celebrated across the state, and assets are being run hard to deliver them. Practitioners described capital flowing to throughput while renewal waits, and maintenance debt that stays comfortably invisible while prices are strong. What they are asking for is a way to put the cost of deferral beside the production number, while there is still time to act on it.

In WA fixed plant, a shutdown is a rhythm rather than an event: every sixteen to eighteen weeks, a thousand work orders, and hundreds of contractors drawn from the same labour pool every other operator is drawing from. Scope that reopens the moment the equipment does, and lessons captured on paper once the event is over, make the shutdown the most expensive place for small gaps in discipline to appear.

For some WA operators the transition is no longer a strategy document, it is equipment in the workshop. New fleets and grid-scale storage arrive without failure histories, without proven maintenance strategies and without a labour market that already knows them, while the legacy fleet still has to be kept safe and reliable until the day it retires. Nationally, maintenance teams rate their own understanding of what decarbonisation means day to day at 3.26 out of 5. The gap is readiness rather than willingness.

Everyone knows prevention is worth more than recovery, and most of the recognition still goes to the recovery. The crew that brings the plant back at two in the morning gets the thanks; the one whose equipment never failed produces no story at all. Changing what gets celebrated is one of the cheapest levers a maintenance leader has.

The industry has made safety training systematic. It has not yet done the same for technical competence. Practitioners described people who hold the ticket but still need time, coaching and real fault-finding experience to become capable, and smaller operators choosing to grow their own because they cannot outbid the majors for people who already are.

When crews do not trust the supply chain, they quietly build one of their own. Hidden stores, parts expensed to jobs and never returned, and a "hand on part" habit double the stock on site and hollow out the data that reliability decisions depend on. Practitioners were clear that it is a symptom rather than the problem: make the official supply chain trustworthy and the private ones empty out.