Daniel Petre has invested millions of dollars in software companies over the years. Now, with markets worried that rapid advances in artificial intelligence will all but wipe out their growth, the veteran venture capitalist is the first to warn that there will be plenty of casualties.
Or, as he describes it, “roadkill”. With every new announcement outlining new tasks that AI platforms can automate – the latest was Anthropic’s Claude, which can apparently replicate the work that specialist legal software does – technology stocks look deeper into the abyss.
That development, overnight on Tuesday, pushed the Nasdaq-listed shares of LegalZoom, which provides services to small legal practices, down 19.7 per cent. Thomson Reuters dropped 17 per cent. Atlassian, the Sydney-headquartered, New York-listed workplace software developer, saw its shares down 8.7 per cent. On the ASX, accounting tool Xero and logistics management platform WiseTech Global both took big hits.
“For some big software companies there are big switching costs, or they may have inputs from lots of changing private datasets that are difficult for an AI platform to match, but the key driver in the market today is that there will be lots of [software] businesses that aren’t that strong,” said Petre, the co-founder of Airtree Ventures and a former Microsoft executive.
Beyond the public markets, Petre is heavily invested in some of the country’s biggest and most promising private software businesses, from graphic design platform Canva to workplace tracking group SafetyCulture.
“Investing has now got a lot harder than it has been for the last 10 years and if a venture capital firm is not worried about this, then you have to question their capabilities in understanding what is happening,” he said.
Markets have been stunned by how quickly AI platforms like Claude seem to be able to offer tools that are comparable with software firms, allowing so-called vibe coders to build models without possessing specialist skills.
Alex Pollak, the chief investment officer at Loftus Peak, which has more than $1.1 billion under management, said the sell-off in software stocks was “probably overdone” and that products released by companies like Anthropic were unlikely to match those produced by traditional players, who can also provide ongoing support and frequent updates.
On Tuesday, Xero chief executive Sukhinder Singh Cassidy made a similar argument as she tried to convince investors that Xero’s accounting software was AI-proof. Atlassian chief executive Mike Cannon-Brookes will be pressed to do the same at his company’s quarterly earnings call on Friday.
Cloud and software names account for about 19 per cent of the Loftus Peak Global Disruption Fund, with about $700 million under management, and Pollak said it hasn’t sold these names yet. But he said the AI revolution “was creating this schizoid stockmarket” as investors fret over disruption, spending on infrastructure and how to monetise the new technology.
“This is the most disruptive moment for business and society since the introduction of the internet in 1999. We’ve been poring over this for months now, we can’t find a single industry that’s not affected by AI,” Pollak said.
“It is going to work. It’s already commercial, it’s deployed widely enough. It’s got enough acolytes, power users and corporations using it. We’re not going back. The question then becomes, is it being priced properly?”
Defensive moats
Two start-up founders who recently raised big money to build AI that will automate professional tasks also claim to be unperturbed by Claude.
Thomas Kelly, the chief executive and co-founder of Heidi Health, a medical scribe platform last valued at $US465 million ($662 million), and Phil Thurner, whose legal practice management software developer Nexl is now worth $100 million, both said their products were too embedded in the way that their customers worked to be easily replaced by AI-led replacements.
“The key question is where value actually accrues. In industries with low regulatory barriers and high interoperability, large general-purpose AI platforms will put real pressure on specialist software by abstracting away systems of record. We are already seeing early signs of that,” Kelly said.
“Healthcare is different. Value does not accrue solely to the model, it sits in regulatory compliance, deep clinical integration, trusted workflows and longstanding partnerships. You cannot simply plug in an API and operate safely at scale, and consumer AI platforms are not compliant with healthcare regulation in Australia, the US or Europe.”
An API, or application programming interface, is used to let different digital services and applications communicate and work together.
Thurner said Anthropic was an API business rather than a software rival, making its money from token usage, with more interest in pulling users away from OpenAI than replacing legal technology software vendors.
He said software like Nexl had a “defensive moat” against the disruption through its own data models, its grounding in firm-specific proprietary data and hard-earned knowledge of those businesses’ workflows. Anthropic and OpenAI were more of a threat to applications with limited features, he said.
“For serious vertical platforms Anthropic is more of an enabler,” Thurner added. “I’d frame this less as Anthropic eating legal tech and more as Anthropic competing to be the backbone for legal and professional-services AI. The real question for vertical software players is whether they’ve built real depth, or just a thin user interface on an AI model.”
Ahead of its earnings call, Canaccord Genuity analyst David Hynes said there was little Atlassian could do to turn negative sentiment around in the short term. He said Atlassian’s own AI efforts were going well, and its margins were generally growing across its applications.
“For an analyst who’s spent nearly 20 years covering a space that’s generally been loved, it’s wildly frustrating,” Hynes wrote. “It’s a fear of the unknown, the disprovable AI-bear thesis, and for that, there’s no magic elixir.”
Petre’s Airtree Ventures co-founder and a current partner at the firm Craig Blair said there was more variation in the impact of AI on software than was currently being displayed by a general market sell-off.
“The market is right that AI will have profound effects across software, services and investment businesses, but it’s not yet very discerning about who will benefit versus who will face headwinds,” he said.
“The future isn’t cleanly utopian or dystopian – outcomes will be shaped by new business models, products and pricing regimes so the future is nuanced and less predicable than simple headlines suggest.”