When Inovia Capital partner Hugues Lalancette sat down recently to build a financial model for his fund, he decided to experiment with a new approach. Rather than tackle the job manually, he opened Anthropic’s Claude, gave it some prompts to perform a task in Excel, and walked away. By the time Lalancette returned from his coffee break, the assignment was complete. The AI had crunched an hours-long job into minutes. 

For Lalancette, the moment drove home how quickly AI has taken over tasks many software companies were built to handle. “It’s a game changer,” he said. 

It’s the same realization that’s agitated public-market investors since the start of this year, after Anthropic released new Claude Code features that can build and run software nearly autonomously. The updates, which showed sudden leaps in what AI can do, have wiped billions of dollars off the valuations of legacy software firms. The iShares main exchange traded fund for the sector has plunged nearly 23 per cent since the start of the year, as of the end of trading Monday. Blue-chip giants Salesforce, IBM and Thomson Reuters shed 27.2 per cent, 19.2 per cent and 25.1 per cent, respectively. 

Away from the public scrutiny of the stock market, venture capital investors and the droves of software startups they back are quietly grappling with a similar, and potentially deeper, crisis. 

VCs and analysts told The Logic new AI tools are forcing investors to scan their portfolios and confront which companies actually own something that AI can’t easily replicate. “If you can explain the problem in plain English, Claude Code can generate the software for you in five minutes,” said Allen Lau, operating partner at Two Small Fish Ventures. 

Venture investors have been wary of AI’s threat to software startups since ChatGPT arrived in late 2022. Fund managers responded by imploring their portfolio companies to use the technology to save them time and money. The sense of urgency, however, has quickly ramped up with the latest Claude releases and the market’s reaction to them. “This is almost a once-per-generation disruption,” said Lau.

Canada’s venture capital portfolios are heavily weighted toward startups that sell software. The information-and-communications-technology sector—which is mostly software companies—captured more than half of all VC dollars invested in 2022, according to the Canadian Venture Capital and Private Equity Association.  

What once made software startups attractive—their code and the engineers who built it—no longer carry as much value. That’s forcing investors to reassess what they own. “Any enterprise [software] company that was built over the last five to 10 years can be rebuilt in six months or less,” said Sophie Forest, a partner at Brightspark Ventures. 

Her firm is pushing founders to shift their operations to AI wherever possible, and then take stock of what value they have beyond code, such as long-standing customer relationships, and use that to their advantage. The pressure from AI means every company needs to ramp up their speed and business volume, she said.  

Lalancette said the enterprise clients buying this software now expect more from their vendors. “Demand is shifting,” he said. “You need an AI story to tap into enterprise budgets.” 

Many of those customers are experimenting on their own, testing tools like Claude to do tasks they might have otherwise paid a software company to do, said Patrick Lor, managing partner at Panache Ventures. That’s raised the bar for what vendors must deliver. At this point, he said, the productivity gains don’t have to be massive. They just need to be better than what the customer can achieve on their own. Clients may pay to save just 10 minutes on a task, said Lor. Increase those savings to 30 minutes, he said, and “they’ll pay real money for it.”

Investors who spoke to The Logic said not every company can adjust, and many will inevitably fail. “I think most of them won’t be able to adapt,” said Forest. 

Ray Wyand, CEO of Verified Metrics, which does due diligence on companies for venture capital investors, estimates that about a third of venture-backed businesses can “genuinely reposition themselves” to thrive in an AI-dominated market. Some may not need to, he said, and many others will struggle unsuccessfully. 

The expectation that many software startups won’t survive the shift is already influencing where venture capital money is going. A recent Inovia report found that AI-native startups received 40 per cent of all venture investment in Canada last year, up 14 percentage points from 2024 and double the share from 2015. 

The result is a strange feedback loop, said Wyand, in which venture capitalists are now financing the very technology that could erase much of their portfolio value. The impact won’t be evenly distributed, he argued. Large global funds with stakes in firms like OpenAI and Anthropic are poised to offset their software losses with their AI bets. Smaller funds that are mostly full of conventional software—of which there are many in Canada—have more to lose and less to gain, Wyand said. 

Public markets swing fast, and the recent software sell-off may reflect shifting expectations more than immediate disruption, Wyand said. Private markets tend to follow eventually, he added, and the corrections can be even more dramatic. 

Not all software is equally vulnerable, investors told The Logic. Products built around deep expertise, proprietary information or in regulated environments remain harder to replicate. Lor pointed to industrial engineering and infrastructure planning tools, where legal liability and confidential data limit what generalized AI systems can replace. “Where we’ve been looking is in areas where there are barriers beyond the code,” he said. 

One thing investors agreed on was there’s no more time to wait and see how their software companies will weather the AI boom. “The worst thing we could do,” Forest said, “is put our heads in the sand and pretend it’s not happening.”