AI finds role in startup funding

The use of artificial intelligence to assess credit risk is starting to take root in Asia, with one Singapore-based fund management company leveraging the technology to weigh debt financing decisions for startups.
Mars Growth Capital, established through a partnership between Japanese lender Mitsubishi UFJ Financial Group and Israeli fintech Liquidity Group in 2020, has four funds managing a combined US$1bn under its non-dilutive debt financing strategy – a unicorn fund for late-stage technology companies, a growth fund for companies in Asia, a Europe fund, and a Japan fund.
Mars, which started with an initial capital commitment of US$80m, has provided funding to several Asian startups, including food delivery company HungryPanda and SoftBank-backed mobile advertising provider InMobi.
Osamu Abe, MUFG’s chief of staff for Asia Pacific, who serves on the board of the Mars Equity Dragon Fund, said the bank entered the partnership with a view to address a vacuum in funding for startups, as fast-growing companies often needed cash quickly but founders did not want to dilute their stakes. With Mars' debt financing, the client's source of repayment comes from its next equity round, and while Mars may sometimes ask for collateral, that is rare since digital startups typically do not have meaningful assets.
“We are always seeking the future of finance – what we can do, what we can try out, knowing that nobody is sure how long this current way of banking continues,” said Abe. “This was a great way for us to jump into a new lending business.”
Leveraging MUFG’s financial strengths and Liquidity’s proprietary machine learning technology, Mars Growth says it can provide termsheets of up to US$120m in non-dilutive capital within 72 hours.
After entering into a non-disclosure agreement, Mars gets access to real-time financial and accounting data from client bank accounts, accounting systems, customer relationship management databases and any other useful internal resources. The data are fed into Liquidity’s machine learning and decision-making algorithm to forecast earnings and cashflow and predict how much runway a firm has until it depletes its current funding and needs another round.
This AI-based credit scoring model “in the end accounts for 60% to 70% of our decision. The rest is taken by a human, or a 'human in-the-loop' as we call it," Abe said, adding that this involves more traditional, nuanced aspects of decision-making, such as talking to management.
“The result speaks for itself. We have had no defaults for the last four years,” he said.
The Mars Equity Dragon Fund, launched in 2023, will also use the AI model to help make equity funding decisions for middle to late-stage startups.
Private market potential
The next possible use of this technology, or something like it, could be in private credit.
“Obviously, that is the next large pool of market wallet that is out there and I’m sure some of our peers are doing it already," Abe said.
“There is so much interest from other institutions. So far, we haven’t joined hands per se, but there is much interest," he said, adding that while the same technology could be applicable, it would be likely to use different parameters, or different data points for private credit.
Experts agree that while AI has use cases in credit risk assessment in private credit, it is not an alternative to human involvement.
Sumit Bhandari, lead portfolio manager for Asia private credit at Allianz Global Investors in Singapore, noted that while there is a part of credit risk analysis that can be done with AI, "the entire process of making private investments is much more involved".
“AI can be particularly useful in terms of standardising certain things, such as taking a look at operational metrics, checking for the same things so that in the heat of the moment or the speed of the deal you don’t miss out on some things that you shouldn’t be missing,” Bhandari said.
However, “in private credit, the willingness to pay comes through years of relationship building, years of knowing a company, not just knowing a company but also their ecosystem. A lot of that knowledge needs to be kept up to date and synthesised. It is not a point in time analysis; it is a period of time analysis. This is where having the right portfolio managers who are generating the right deals is really critical and half the alpha in a private credit strategy in Asia,” he said.
The role of AI in credit risk assessment is growing globally. For example, investment firm Muzinich & Co’s parallel lending team, which has established relationships with 55 banks across Europe, already uses AI and machine learning to assist investment decisions.
“We use AI, machine learning to increase the accuracy of analysis and improve investment decision-making. The system employs powerful statistical tools to identify patterns and/or predict whether a company may default, which are helpful when capital preservation is the key objective. We also have our typical due diligence that we do from a human analytics perspective," said Andrew Tan, Muzinich's Asia Pacific CEO and head of Asia Pacific private debt.
“Even if a bank says yes to a deal, but say we run the deal, numbers, deal parameters through AI and it comes out and says the likelihood of something unpleasant happening is high – we pass on the transaction," he said.