AI can make P2P lending safer and more fruitful for investors

Peer-to-peer (P2P) lending was born from the idea that cutting out the middle man – i.e. the bank – could make finance cheaper and more accessible. But in recent years, the lack of an intermediary has arguably undermined a key part of the sector, namely the secondary market for trading in existing P2P loans.

Could artificial intelligence (AI) change that situation and inject fresh energy – and useful oversight – into the market?

Trajectory of the P2P market

Starting in 2008, a number of P2P platforms began to launch these secondary markets, which allowed lenders to make an early exit from some of their loans by selling them on to others.

Part of the motivation was to address a fall in liquidity levels on the platforms. A rising number of defaults by borrowers had started to make lenders more picky. The hope was that they might become less risk averse if they knew they could offload any unwanted loans ahead of the maturity date.

The first to offer this was California-based Lending Club, which launched its secondary market in 2008. For a while it seemed to work well and others followed, but over the past decade many of these secondary markets have shut down.

Prosper closed its one in October 2016, and Lending Club shuttered its offering in August 2020 (and then exited the P2P lending market entirely a few months later).

A key problem was that the secondary markets started to suffer from the same liquidity problems that had prompted their launch in the first place. The platforms found there simply wasn’t enough interest among lenders to make the markets viable – perhaps because lenders viewed loans in the secondary market with greater suspicion, given they could not know why the existing lender wanted to offload them.

In fact, in one aspect there is more information about a loan in a secondary market as there will be a track record of repayments (or defaults) by the borrower which is not available in the primary market.

The situation was then exacerbated by the Covid-19 pandemic, which squeezed the market from both sides. The collapse in economic activity harmed the creditworthiness of many borrowers and, at the same time, led to a rise in the number of investors that wanted to sell their P2P loans early.

Regulatory issues have also affected the market. This year the Reserve Bank of India, the country’s central bank, caused shockwaves in the P2P industry when it issued new guidelines that restricted secondary market activity for platforms there.

The trajectory of secondary market activity over the past decade is rather curious though, given how well such mechanisms operate in other sectors. In the car market, for example, there is a broad and deep second-hand market. Stock markets around the world enable trading in shares in listed companies, while over-the-counter mechanisms do the same for sovereign debts.

The P2P market has proved to be different though, probably because of the lack of intermediaries who can re-evaluate existing loans, bundle them up with other debts and then syndicate them as a package. In traditional debt markets, that securitisation work is done by banks and asset management firms, but there has been no-one to take on that job when it comes to the P2P secondary markets.

AI tools in P2P platforms

This is one area in which AI tools could step in, by effectively becoming that missing intermediary.

AI could package up loans into a portfolio which matches the needs of a lender, making sure that it meets the level of risk they are willing to bear. If done effectively, it might even increase the profits they make.

Of course, this is not something that just works for the secondary markets. The same tools can be used for the primary market too, or indeed a combination of the two. AI could optimise at speed a pool of perhaps hundreds or thousands of loans, both from the primary market and also from the secondary market, helping lenders construct a portfolio which balances their risk and returns.

For an AI engine this is a fairly straightforward optimisation task. A lender could set their desired goals and let the AI engine work out how best to achieve them. Some lenders by nature are quite happy to take on higher levels of risk in return for greater potential returns, while others are looking to minimise their risk and will accept lower gains. In either case, AI could help to match a lender’s requirements with the opportunities in the marketplace.

There are potential benefits for borrowers too. Just as an AI engine could help to match lenders with the most appropriate borrowers, so it could also evaluate the financial requirements of a borrower, their ability to make repayments and, combined with other factors, then match them with the most appropriate lenders.

There are a number of thorny questions that remain to be answered though. For example, should an AI engine be able to consider information about a borrower sourced from outside the platform?

A borrower's likelihood of defaulting could perhaps be better predicted by an AI engine that was able to incorporate information about their social media posts, career history, transaction history and other past behaviours which might be correlated with repayment performance. But such an approach could throw up ethical issues.

A related consideration is to what extent a borrower should be penalised due to a previous default – a decision which should not necessarily be outsourced to an AI model.

The reality is that AI is not always the best answer to a problem and some human engagement will still be necessary. Even though AI can offer useful suggestions, lenders and borrowers still ought to exercise their own judgment too.

Just as you would be unwise to unthinkingly copy and paste any information you might garner from the likes of ChatGPT or Claude, so you should not have blind faith in what an AI model might be able to do on a P2P platform. The AI can give you advice, but some financial nous will still be needed to validate and assess that advice.

From the perspective of the P2P platforms, there are few obvious downsides to integrating AI services. If it leads to improved outcomes for investors, then it should attract more lenders; and if it attracts more investors, it should in turn attract more borrowers too, helping to ensure liquidity in the future.