#StartupSpotlight is a series on Information Age shining a light on Australian start-ups disrupting the status quo.
The explosion of interest in artificial intelligence (AI) has seen data scientists snapped up in droves by cashed-up multinationals, but Richard Kimber believes the right combination of AI nous and business acumen will nonetheless help his Australian AI start-up make its mark in this fast-emerging industry.
Expertise is something the new venture, Daisee, has in plentiful supply. Since it was founded in August 2017, Daisee has grown to 25 staff – and Kimber, who as CEO has been directing the company’s full-court press onto the local market, wants this to reach 40 engineers this year and 100 in 2019.
Partnerships with universities – first Melbourne’s Deakin University, then Macquarie University and the University of NSW – have not only increased the company’s profile within the data-scientist world, but have given a leg up to the venture, which made its public debut in February this year.
Serial entrepreneur Kimber – an ex-ANZ and -Google executive who maintains advisory roles with the likes of AustralianSuper and Unlockd – previously guided payments provider OFX Group to a presence in six countries based on his understanding of both the money and the technology aspects of a successful start-up.
He believes AI is in a golden era and – despite the mountains of cash being poured into the technology by behemoths like Google, Microsoft, Amazon, Uber and others – sees considerable opportunity for a company that can help businesses apply AI to solve real-world business modelling and automation problems.
“In Australia there are very few specialist AI companies,” he explains. “The real advancement and development of AI has been predominantly in the academic world.”
“Our view is that if we can take the best of that academic world – and apply it to the commercial world – we get a code base that has been tested over decades. We can hit the ground running.”
Veins of opportunity
The company’s February launch was accompanied by a survey report, which it commissioned from analyst firm House of Brand, suggesting that Australian companies were behind world benchmarks in AI spend: 14 percent have adopted AI, compared with the global average of 23 percent.
However, the proportion of Australian companies expecting to invest more than $1m in AI was expected to rise from 6 percent today, to 13 percent in 2022.
Kimber and his team have identified three “deep veins of opportunity” for AI within the Australian market: voice and call-centre applications within industries such as financial services, embodied in Daisee’s Echo product; retail supply-chain optimisation and pricing, serviced through the company’s Harmony business group; and computer vision, developing under the Acuity brand.
The demand for computer vision – used to great effect by the likes of Carsales, which has built an engine for automatically recognising different perspectives of a car – had “really surprised us”, Kimber says, noting strong demand from medical companies that see the technology’s potential for speeding analysis of CT scans, X-rays, and other medical imaging.
Whereas many companies had only implemented AI as rudimentary website-based chatbots, those implementations “are not really that advanced,” Kimber says. “Our view is that the real AI opportunities lie elsewhere, where you can apply vast data sets and have algorithms that really do adapt and learn.”
Investors agree. In March, the company secured $8.8m in a Series A funding round led by Alium Capital – whose co-founder Michael Considine said Daisee’s “proposition of an experienced team leveraging world class research for commercial application sat head and shoulders above what other companies are seeking to achieve”.
AI as a service
As he guides Daisee’s growing data-engineering team from the start-up phase into what he believes will quickly become a B2B AI powerhouse, Kimber says the ability to help companies embrace a more mature view of AI will be a real differentiator.
Many companies have dabbled in the technology but most “are not set up to do AI,” he explains. “There’s a lot of overhead, and a lot of tuning that goes into getting AI to work well.”
“Ultimately, the actual teams with the skills to deploy and use AI are quite small and limited,” he adds. “We have spoken with some Internet companies in Australia that have data-science teams, but their backlog of requirements and other initiatives has crowded out their AI development.”
That inexorable force was trapping many Australian AI engineers, he added, “who found themselves “buried in corporates doing work that is more around business analysis and intelligence, and not really leveraging their full suite of skills.”
Such staff would leverage the opportunity to put their shoulder against more complex modelling tasks of AI, he believes: “For us, the opportunity is to liberate them – and get them more focused on value-added work and predictive applications.”
While many AI routines are being put into the public domain by industry players keen to progress the industry as a whole, Kimber sees delivery and industry-specific optimisations as key differentiators that will help Daisee fend off the deluge of better-funded, high-profile cloud-based AI platforms.
A key short-term goal is the delivery of AI-as-a-service, which Kimber believes will allow companies to quickly integrate customised AI capabilities into their business applications. This, coupled with the locally developed code and support of three strong research universities, is the core of the company’s value proposition “and is our view as to what success looks like”.
So, too, is what he calls the “explainability” of AI – the ability for humans to understand how AI engines arrive at their conclusions, so that automated decisions with surprising outcomes can be unpicked during forensic analysis and remediation.
“To the extent that AI is a black-box type solution, regulators are going to request information about the origins of decisions and how they are made,” Kimber explains.
“Typical software works in two or three dimensions, whereas AI literally works in n dimensions. That dimensionality is what makes it difficult to explain how the models work – and the real winners will be the ones that apply them effectively.”
