For the better part of two years, almost everything sold to Australian businesses has carried an artificial intelligence label. Software vendors, hardware makers and energy retailers have all reached for the same adjective, and nowhere has the trend been louder than in the solar and clean-energy market, where “smart” inverters, “intelligent” monitoring and “AI-optimised” storage have become standard sales language. A fresh warning aimed squarely at Australian operators argues that a good deal of this is marketing dressed up as machine learning, and that the companies buying it are paying a premium for intelligence that often is not there.
The caution, reported this week under the headline The Great AI Solar Con, lands on a tension that runs well beyond rooftops and battery cabinets. The phrase “artificial intelligence” now sells, and that commercial reality has blurred the line between genuinely adaptive systems and ordinary automation wearing a fashionable badge.
What is actually being sold
The distinction matters because the two things are not the same. A rules-based controller that switches a battery to charge when a timer or a price signal tells it to is useful, but it is not intelligent in any meaningful sense. It follows instructions a human wrote in advance. A genuine machine-learning system, by contrast, builds its own model of how a site behaves, learns from weather patterns, occupancy and tariff movements, and improves its decisions over time without being reprogrammed. Both can be valuable. Only one deserves the word that is driving the price up.
The warning to Australian businesses is essentially that the gap between the two is being papered over. Dashboards that simply graph yesterday’s generation are presented as predictive. Threshold alarms are described as anomaly detection. Scheduling logic that has existed in building management for decades is rebranded as optimisation driven by artificial intelligence. For a procurement manager comparing quotes, the labels offer no reliable way to tell which product actually earns its keep.
Two ways to read the hype
There are two defensible readings of all this. The sceptical view, and the one the source leans towards, is that “AI washing” has become a quiet con. Buyers are being charged for capability that does not exist, the real-world savings fall short of the pitch, and the term has been so thoroughly diluted that it no longer communicates anything. On this account, the honest response is to strip the branding back and judge a system on measured performance: how much it lifts self-consumption, how much it trims a demand charge, how quickly it pays for itself.
The more charitable reading, which many vendors would offer, is that the industry is mid-transition rather than mid-fraud. Plenty of energy platforms genuinely are moving from fixed rules toward forecasting and learning, and the marketing has simply run ahead of the engineering. On this view the problem is overclaiming rather than outright deception, and the fix is clearer disclosure about what a product does and does not do, rather than abandoning the technology. Both positions agree on the practical point that matters to a buyer: the label is doing too much work and the evidence is doing too little.
Why this bites harder in Australia
Australia is an unusually exposed market for this debate. The country has one of the highest rates of rooftop solar penetration in the world, battery uptake is climbing fast off the back of federal and state incentives, and commercial operators from cold-storage warehouses to shopping centres are under real pressure to tame energy bills that have swung violently in recent years. That combination creates enormous demand for anything promising smarter energy management, and demand of that kind is exactly where inflated claims flourish.
It also runs straight into a regulator that has made misleading environmental and technology claims a priority. The Australian Competition and Consumer Commission has spent the past two years pursuing greenwashing, taking action over sustainability statements it considered unsubstantiated, and it has signalled that unsupported claims about what a product can do fall within the same consumer-law framework. An “AI-powered” badge that cannot be backed by evidence sits uncomfortably close to that line. For Australian businesses, the risk is not only buying something that underperforms, but also, for those reselling or installing these systems, repeating claims they cannot stand behind.
There is a workforce dimension too. Small and medium operators rarely have an in-house data scientist to interrogate a vendor’s claims, so they rely on the sales deck. That asymmetry is precisely what the warning targets. When the person signing the contract cannot tell a neural network from a timer, the word “intelligent” becomes a substitute for scrutiny rather than a description of it.
How to buy without being conned
The practical advice flowing from the critique is refreshingly unglamorous. Ask the vendor to explain, in plain terms, what the system learns and how. Ask whether its decisions change over time or stay fixed once installed. Request measured results from comparable sites rather than modelled projections, and insist that any savings claim be tied to a baseline you can verify after the fact. Treat the letters “AI” in a product name as a prompt for questions, not as an answer to them.
None of this is hostile to genuine innovation. The energy sector is one of the areas where machine learning has a credible, concrete role, from forecasting generation to orchestrating fleets of batteries into virtual power plants. The argument is simply that the real thing can withstand scrutiny, and the marketing should not be allowed to coast on the credibility that the real thing has earned.
What is next
Expect the pressure to come from two directions. Regulators are likely to keep testing whether technology claims, like environmental ones, can be substantiated, and a single enforcement action over an “AI” claim would sharpen minds across the industry quickly. At the same time, buyers are becoming more literate, and the novelty premium attached to the term is already thinning. The vendors most exposed are those whose products cannot survive the plain-English question at the heart of this warning: what does your system actually learn, and can you prove it. For Australian businesses weighing their next solar or storage upgrade, that question is now the most valuable tool in the room.
Sources: Tech Business News.
