One in three ASX companies make unvalidated AI mentions

To nobody’s surprise, AI has climbed the agenda in corporate Australia over the past five years. For some companies, this new technology is doing real work. For others that are afraid to be left behind, it is mostly talk. The data history is short, but at Plato we have been trying to tell the two apart, and to keep an eye on the companies we believe sit in the second group.

AI talk on the ASX

Our natural language processing models read each call, separate management from analysts, and extract the performance metrics a company presents. Against that we flag AI references from a curated term list, with rules that stop “AI” being mistaken for “Air” or a company name, adjusting for call length.

As you’d expect, AI mentions in earnings calls have climbed steadily since 2022, and among the companies that raise it, the topic takes up more airtime.

The chart above shows the share of ASX companies making an AI mention we cannot validate. We call a mention “unvalidated” when a company mentions AI but neither reports an AI-linked metric nor draws a single analyst question on the topic. Today around 1 in 3 ASX companies (31.7%) make unvalidated mentions of AI, up from 1 in 7 in 2022. This is not proof of “AI washing”, since a company might raise AI in passing or when discussing external trends, but it tells us where to look.

Managed Fund

Plato Global Alpha Fund

Global Shares

View

ETF

Plato Global Alpha Complex ETF (PGA1)

Global Shares

View

We use three measures to rank this “AI overstatement”:

1. Talk-to-substance ratio. We weigh how much a company talks about AI against how many hard, AI-specific figures it presents (“AI spend”, “copilot seats”, “GPU utilisation”, “% bookings with AI”). Companies earning from AI usually have a figure to point to. We notice when there isn’t one.
2. Presentation vs Q&A asymmetry. If AI is concentrated in prepared remarks, but analysts don’t pick it up, that’s a strong tell. For example, AI might come up 20 times in prepared remarks without drawing a single analyst question.
3. New story, vanishing metrics. Companies that ramp up the AI narrative in the very same result where they quietly drop figures they used to highlight. A classic shifting the goalposts to distract investors.

A foreign case study: Presto

Presto Automation was a US-listed restaurant technology company selling AI for drive-thru ordering. On its November 2023 earnings call, management returned to AI repeatedly, almost always without a figure attached. Its chairman told investors:

“I’ve Said It Before And I’ll Say It Again, I Believe Presto Is One Of The Most Immediately Actionable And Scalable Vertical Applications Of AI Today, Even Though I Know It Will Take Some Time To Reflect In The Numbers.”

A grand claim, with the admission that it was not yet in the numbers. That is exactly what our first measure flags: plenty of talk, nothing an investor could check.

The SEC later found Presto had mislead investors, telling the market its technology would eliminate human order taking, when in practice humans were still handling around 70% of orders. The measure does not detect fraud, but it was early. Presto was delisted from Nasdaq in August 2024, and its lenders forced a sale of the business that December.

The AI program nobody asked about

We combine the three measures into a single score to flag companies we believe are overstating their use of AI. The history is short, so we make no claim about performance. Breville’s (ASX: BRG) recent earnings call caught our attention.

Full disclosure: Plato has held a short position in Breville since March 2026, well before the company appeared in this screen.

Breville makes premium kitchen appliances. If you own an espresso machine that cost more than your first car, there is a fair chance it has their name on it. It is not, by any definition, a technology company.

Which is what makes its FY26 results presentation interesting.

Breville had a reasonable year and a difficult one at the same time. Revenue hit a record $1.81 billion, up 6.7%. But profit barely moved and gross margin fell 60 basis points, squeezed by US tariffs and the cost of shifting manufacturing out of China. The company declined to give guidance for the year ahead. The shares fell on the day.

So there was plenty for management to talk about. Instead, during the CEO’s speech, which ran for 15 and a half minutes, over nine minutes were spent talking about an “AI transformation program.”

The chief executive returned to an “innovation flywheel”, where spending on new products drives revenue, which funds expansion, which drives more revenue. Historically, he said, that flywheel ran on two things: people and money. Not anymore.

“For 2026 Forward, We Now Have An Incremental Lever, AI Tokens. We Can Now Drive The Flywheel With Head Count, Tokens, And Money.”

A coffee machine company describing AI tokens as a factor of production, alongside labour and capital, is a reasonable place to start paying attention.

Not one dollar of spend was disclosed. No headcount, no named technology partner, and nothing on what the program had earned. The company said only that “our total spend on the program is deliberate and controlled.”

Breville did offer numbers, but they were operational rather than financial, and the company generated them itself. Over 50% of employees are now “amplified,” defined as confidently using AI as a tool, measured internally with no methodology given. A recipe application delivered a “6.3x improvement,” calculated against an estimate of how long the old process would have taken. Useful signs of progress, but not figures an investor could check, or put in a model.

Then the presentation ended and the analysts asked their questions.

Twelve of them were on the call, from UBS, Macquarie, Goldman Sachs, Barrenjoey, RBC, Jarden and six other firms. Between them they did not mention AI once. Not the transformation program, not the applications, not the tokens, nothing.

That is the gap our screen is built to find. 

A company chose to spend a meaningful part of its results presentation on an AI story, in a year when its actual profit challenge was trade policy and margin. It attached no figures to that story that an investor could check. And the twelve professionals paid to scrutinise the company decided it was not worth a single question.

None of this means Breville is doing anything improper, and it is not a prediction about the share price. But when talk about AI runs well ahead of anything measurable, and the market does not engage with it at all, we think that gap is worth watching. It is one of the clearest examples we found this reporting season.