Your next results announcement will be read by machines before most humans open it. They’ll summarise what's material, score the sentiment, and in some cases trade off it, all before the investor call begins.
While this may seem too inhuman to be real, it's the environment investor relations (IR) professionals are navigating right now. The people reviewing your announcement may have run it through an AI tool before picking up the phone. The broker report in your inbox may have been drafted by one. The share price movement that followed your release may have been driven entirely by algorithms, with no human ever reading a word of it.
"It's not people reacting, it's machines reacting," one IR professional told us. "Because it's too quick to be a human."
For IR professionals, this changes how disclosure works in ways that are still being defined.
At Atticus, we have a front row seat to the changes. IR teams use Atticus to verify disclosure documents before they go to market – and the IR leaders we speak to are telling us how those documents are increasingly being read by machines. We recently spoke to four of Australia's largest listed companies and heard the same things repeatedly: how thoroughly AI tools are now reading disclosure documents, and how automatically machines are trading off them.
We heard that thoroughness and speed are increasing the pressure on IR teams – to move faster, to cover more, and to be as accurate as possible, all while errors are easier than ever to find. For us this poses a critical question: in a market where speed is becoming the main competitive advantage, how do you move faster without sacrificing accuracy?
What IR professionals are seeing
The four leaders we spoke to were each navigating the shift differently.
One IR leader described receiving a broker report scoring her company's disclosures for negative sentiment across three document types. The pattern was clear: the more scripted the communication, the lower the negative sentiment score. The written announcement attracted around 4 negative words per thousand. Prepared commentary came in at around 7. The Q&A, unscripted and spontaneous, scored above 13, more than three times the written announcement. A high score means the document is more likely to trigger a negative automated reaction before any investor has read a word of it. Her team runs every announcement through AI before release to bring the scores down as much as possible. The result, in her words, is language that is "quite boring. No adjectives. Very vanilla" – but language that scores well, which is now part of the job.
For others, the more pressing concern is what happens on the trading side. Quant funds and algorithmic systems don't wait for a human to form a view, they process disclosures automatically and act on them. "All trading between 10.30 and 3pm is quant-driven," one IR leader told us. "It's only at the start and end of the day that you're seeing real investors."
For one professional with two decades in the industry, the more pressing question is what the role looks like in five years. "What does the future of investor relations look like, given the rise of quant funds, index funds, passive money?" he asked. "The pool of active capital is getting smaller. Does the role change? Am I now more focused on governance?" Nobody we spoke to thought the IR function was going away. But the question of what the role looks like is one that only the future will decide.
The skill you can't automate
As optimising for sentiment becomes standard practice, a different problem emerges. When every IR team is running announcements through AI to strip out negative language, everything starts to sound the same.
The skill that gets lost is what one senior IR leader called the "intentionality of messaging" based on critical thinking and judgement. Not choosing words to score well, but choosing words because you've thought carefully about what you need them to do. "I don't want to outsource that development of skill to the machine and lose it," she said
The more productive use of AI, in her view, isn't to produce the message, it's to stress-test the thinking behind it. "AI should spar and make you better," she told us. "Test my logic and framework. Challenge me." Used that way, AI sharpens the judgement that makes IR teams valuable. Used as a shortcut to the finished product, it erodes it.
Confidence is the outcome
Machines are reading disclosures more thoroughly than ever. Markets are moving faster, with algorithmic trading driving decisions without human intervention. We all want speed, until it affects accuracy. How sure can IR teams be that they can release faster with total confidence that every risk is managed?
The IR leaders navigating this well have resolved that tension the same way: by building a verification process thorough enough to catch everything, and embedded enough to move fast. When every statement has been checked, every number has someone accountable for it, and the evidence is organised into a complete audit trail, the question of whether to release stops being a source of anxiety.
One IR leader at a major Australian bank described his measure of success as nobody calling after a release to say a number is wrong.
The goal a well-run IR function works toward is being able to tell the board or the executive that the document has been verified: that every statement has been checked for accuracy, backed by evidence, given the tick of approval, and organised into a single, searchable audit trail.
At large listed companies, the personal and reputational risk carried by the board and the executive is real and individual. A documented verification process is what makes that risk manageable. When a question arises about a number – at 9pm the night before a release, or in a regulatory context – the record shows who verified it, when, and against what evidence. The recent Nuix judgment, a case in which ASIC challenged a listed company's disclosure process and lost, turned in part on exactly this point: the process was rigorous, it was documented, and it was followed.
In a market where disclosures are scored and interrogated before the investor call begins, confidence in a disclosure comes from having a rigorous process, running it consistently and efficiently, and leaning on human judgement to provide the final decision. At least one thing remains clear: we're not quite ready for machines to carry that much responsibility.
.avif)
