Pic via Getty Images
Blackbird ventures an opinion on AI outlook: There’s no sign of a bubble
Tech
share
Artificial intelligence is just getting started, with Australia’s biggest venture capital firm dismissing concerns about a bubble forming in the much-hyped technology after China’s DeepSeek began to puncture dizzying valuations.
Blackbird Ventures partner Tom Humphrey said 2025 would be “the year of enterprise AI” when the nation’s biggest companies begin deploying it at scale to jump-start flatling productivity.
Despite AI’s promises to lift output from workers, many companies have struggled to make the leap from pilot study to deploying it at scale. According to Cisco, 85 per cent of chief executives consider launching an AI strategy in the next 18 months a top priority but only 13 per cent believe they have the capabilities to do so.
Mr Humphrey said the tide was beginning to turn. Blackbird invested almost $300m in Australian and New Zealand start-ups last year across 20 new and 30 follow-on rounds, and is expecting to invest a similar amount this year. At the same time, it returned more than $1.1bn to its investors.
Mr Humphrey said companies such as Commonwealth Bank and Telstra were betting big on AI, showing smaller businesses that the technology could deliver a return on investment in a relative short time, spurring growth.
“The CTO (chief technology officer) of Commonwealth Bank came out and said that they had rolled out GitHub co-pilot across the engineering teams. After a few months they saw 30 per cent productivity improvements,” Mr Humphrey said. “That should get anyone excited, because, firstly, we’re talking about Commonwealth Bank, which is probably one of the slowest moving, archaic kind of enterprises in Australia.
“If they can kind of move that quickly to deploy an AI application to drive 30 per cent productivity improvements across a really large team, like that, (it) is real value.”
Telstra has formed a $700m joint venture with Accenture to accelerate AI adoption across the telco. “We’re going to see a lot more of that,” Mr Humphrey said. And it will start with customer support functions.
“For most organisations, support is both a huge cost centre and a significant driver of customer experience. At the same time, LLMs (large language models) are at a point where AI can handle a rapidly expanding list of text and voice-based customer needs with greater speed, lower cost, more control, and higher accuracy.”
Mr Humphrey said early adopters had seen “huge savings”, citing Klarna’s AI assistant – which it launched early last year.
“Within one month it was handling two thirds of its customer inquiries and saving an estimated $40m in support costs,” he said. “With savings like these, the market opportunity for customer support AI is enormous. AI competes for customer service/headcount budgets, which are generally over ten times the scale of support software budgets.”
And this is where Blackbird is eyeing investment opportunities. It led a $9m raise for Lorikeet, which builds AI agents designed to perform various customer services tasks and is now valued at $100m. Blackbird led another $5m raise for Noosa-based Springboards, an AI platform designed specifically for the advertising creative industry. It is now valued at more than $20m.
“Glaring opportunities like this don’t go unnoticed and there are a number of players fighting for a piece of the market pie,” Mr Humphrey said. “Larger incumbents have scrambled to launch their own solutions, and a handful of start-ups have begun to attract attention and funding.
“Our bet is that there will be a handful of big winners in this market and that Lorikeet has a shot at being one of them.”
Mr Humphrey said while Blackbird made 20 new investments last year, it met with about 1600 companies.
“The volume of companies we saw last year is probably reflective of that maturation of the ecosystem. I think we’re seeing more founders, founding companies, than ever before.
“I don’t think there’s hype here. I don’t think this is a bubble. And the reason for that is because there’s just so much substance under the surface.”
He wasn’t phased about the launch of China’s DeepSeek in January, which showed that AI models could be trained at a significantly lower cost.
“I’d say, at the foundational model level, two things are going on: the costs are coming down a lot and there’s also a lot more competition to choose from. And I’d probably actually say in addition to that, the quality of models is going up higher. It’s making the economics of building AI applications … much better.”
This article first appeared in The Australian.
Read MoreTech
