Comment: Decision delegation
By Hamish Paget-Brown, associate at ECI Partners.
The explosion of ChatGPT onto the public scene in November 2022 has reignited the question of whether AI can replace or even perform better than humans when it comes to making decisions on investments. Some clearly think so. At the same time, it is hard to ignore the significant risks, and indeed legal complications, associated with handing over the responsibility for such significant decisions to technology. And for the private equity industry in particular, how might AI shape its future?
Current use cases
There are several areas where AI is already being used within private equity. Sourcing potential targets has been the obvious place to start for a lot of companies, because at this stage you are looking at high-level statistics and not drilling down into the details of a potential investment.
For example, at ECI Partners we developed our own AI tool, Amplifind, initially focused on sourcing proprietary deals. Before we launched Amplifind in 2019, we were looking at about 1,000 businesses every month, and pulling five or so that were interesting based on an initial assessment of the company’s financials and characteristics. With the help of our AI sourcing tool, we have seen an eightfold increase in lead-sourcing efficiency, both through higher quality businesses being presented to us and providing more information on those businesses to drive decision making. While the AI cannot say “this is a good investment”, it can certainly say: “Based on what you have invested in before, this is likely to be a good investment – so why don’t you take a further look?”
AI is also currently proving to be useful in pipeline management. Pipelines are on average increasing in size as the lead time to new investments gets longer. Integrating AI solutions into the pipeline management process can help firms prioritise opportunities and automate processes such as news monitoring, business updates and communications with the potential investment.
The growth of generative AI could also really impact how private equity firms use AI. At ECI, we have recently integrated OpenAI’s Davinci into Amplifind to provide users with automated business descriptions, reducing the need to leave the app to find additional context. In a similar line of thinking, we are also currently looking at developing a bespoke language-based similarity model in order to source and originate deals by looking at unstructured data, rather than largely relying on financials or simple keyword tags. This would be used to very quickly map out an entire market and find companies of interest in an incredibly time- and cost-efficient manner.
Would human decision making be replaced?
Private equity is a people industry, with relationships at the heart of a lot of the decisions that are made. For example, feeling that you would be able to work with a management team for the foreseeable future is an important factor in deciding whether to invest, and not something that AI will be able to decide for you.
Added to this is the fact that when making a decision about an investment, a firm and its executives will do a full suite of due diligence (DD). Forming a view and answering the key questions on an opportunity based on the DD requires interpretation and a complex contextual understanding of what is often a highly nuanced company, market and sale process, which is informed by a verbal overlay and countless meetings and discussions. Generative AI may be able to help summarise a 200-page CDD report, and even help draft the more descriptive elements of an IC paper, but this layer of human interaction and understanding that private equity executives can transpose onto the top of the more quantitative data would be difficult for an AI model to replicate.
Finally, we should consider the fact that the decisions that private equity executives make are often subjective. As an example, they may have to make a call on the future trends in a market, which may or may not be the same as the result of statistical probability based on previous trends. Intuition and experience both play a role in their decisions and while AI may accurately be able to replicate the latter, it will have a hard time replacing the former.
The risks of relying on AI
While the possibilities created by the use of AI in private equity are varied and exciting, there are also significant risks associated with relying on this new technology too heavily.
It is worth noting that for a firm like ECI, we generally have a maximum of 15 companies in a fund and invest in about three or four companies a year. This means that every investment decision we make, either positive or negative, is extremely important and will have a considerable impact on the firm’s performance, and the success of the fund as a whole. While AI might tell you to follow a certain course of action, the reasoning behind that decision will likely be unobtainable. This lack of accountability and understanding is a huge risk when it comes to making such consequential decisions.
There is also the concern that, because models are trained on historic data by necessity, they might not be able to keep up with emerging trends. Large language models (LLMs), especially, require a very long training period due to the vastness of the dataset they are learning from. ChatGPT for example, generally lacks knowledge of events that have happened after its data cut-off in September 2021, greatly limiting LLMs’ ability to analyse recent or upcoming trends and new markets.
And of course, as with most new technologies, there are significant compliance risks associated with the use of generative AI, in particular when it comes to data security, GDPR and copyright. Especially given it is a technology that has not been fully regulated yet, private equity firms will need to be careful that they consider the potential legal implications of its use, both now and in the future.
This is not to say that AI will not play an important role in private equity in the coming years – it most certainly will. Origination and sourcing, pipeline management and prioritisation, document summarisation and drafting, and various other workflows we haven’t touched on can all benefit significantly from AI-enabled processes, driving increased efficiency. However, there are also aspects to decision making that we think it is unlikely AI will be able to replicate in the near future. Private equity is primarily a people industry, and the ability of humans to build relationships and develop a deep, contextual understanding of a potential investment is essential to the investment process. Time will tell if AI will develop these capabilities in the future, but it is not there yet.