What generative AI means for private equity

Just how are different firms approaching the threats and opportunities posed by ChatGPT?

In November 2022, Open AI led a tech revolution that pushed generative AI out of the lab and into the hands of the public, with the launch of ChatGPT. In the year that followed, businesses everywhere have been grappling with how to leverage the nascent technology. Private equity is no exception.

“Some firms are actively looking at use cases to generate cost savings or drive sales. Others are more cautious and are trying to assess whether this is a bubble or something genuinely transformative,” says Chris Gobby, managing director in Alvarez & Marsal’s private equity performance improvement team. “But while there was lots of noise around generative AI in 2023, I think 2024 will be the year when we really start to see these solutions being implemented at scale.”

Blackstone is one of those firms that is on the front foot. “We believe that generative AI can enhance private equity dealmaking by making it easier for analysts to navigate through reams of data when generating an investment thesis,” says John Stecher, Blackstone’s chief technology officer. “This would significantly reduce the amount of time we spend evaluating multiple deals and allow us to focus more on the deals we want to pursue while eliminating risk.

“Our perspective is that as generative AI becomes more commonplace it has the potential to make our employees more productive, enable them to delve deeper into data and help them make better informed investment decisions for our clients.”

Surveying codebases

Blackstone is already starting to leverage generative AI in a number of ways. It is using a product called Cody, from Sourcegraph, that surveys the firm’s extensive and fragmented codebases and then writes code, creating new software. On the investment side, meanwhile, Blackstone is building in-house tools that have the potential to intelligently augment investment processes using generative AI, by enabling teams to generate deep company and sector analysis. It is also exploring the roll-out of various co-pilot tools from Microsoft, Zoom and Salesforce to help employees be more productive.

“We believe that generative AI can enhance private equity dealmaking”

John Stecher,

Triton Partners is also already using generative AI for content summary in all aspects of deal sourcing. “We deal with vast quantities of information in our origination process, and summarising the salient points is a great use case for generative AI,” says Lyndon Arnold, head of technology at Triton. “We are already running various proofs of concept with different technologies and different vendors, to see how far we can go.”

Alongside new solutions, Triton is another firm that is exploring opportunities to exploit AI offerings from existing vendors such as Microsoft’s Co-Pilot product and with Neptune, Triton’s in-house low-code platform for workflow and automation.

Partners Group, meanwhile, was an early mover with its launch of Primera GPT in the first half of 2023. The firm is currently focusing on efficiency and insights-driven areas, according to the firm’s chief information security officer, Patrik Bless. “Everything that involves text writing including LP reporting and RFPs can be supported by generative AI.”

Crowd pleaser

EQT’s approach to generating use cases has been to crowdsource. “GPs invest so much in their people, so it is extremely empowering to bring people’s collective intelligence together to address this challenge,” says EQT’s global chief information and technology officer, Lisa Gunolf.

The firm has launched a cross-functional working group of almost 400 people around the globe to test and innovative around the use of ChatGPT. In addition to tech professionals and others from across the firm, there are more than 10 investment partners involved in assessing the potential of the technology.

Among the use cases that EQT is exploring is the ability to summarize data from lengthy documents, including those provided to LPs. “Helping investors navigate the wealth of information we provide them with is one of the most effective uses of generative AI to save time that we have seen so far,” Gunolf explains.

“AI can also be used to enhance day-to-day communications – summarizing key messages in lengthy emails, for example. In addition, the likes of Google and Microsoft are developing products that can help create first drafts of PowerPoint presentations. Chat GPT can be used to create first drafts of legal documentation too. All of these things help drive efficiency.”

Identifying the risks

Despite the undoubted excitement that generative AI is creating, private equity firms are approaching the technology with pronounced caution. In particular, there is a clear understanding that human intervention is still very much required. “You always have to have human supervision,” says Arnold.

“Generative AI is not a replacement for human judgement,” adds Gunolf. The risks associated with implementing generative AI are not being underestimated, including the threat to data protection.

“Data privacy and protection is top of mind for us,” says Stecher. “Prior to evaluating and starting to leverage generative AI platforms, we spent, and continue to spend, a significant amount of time seeking to ensure that our data won’t be leaked to the base model and is handled in a proprietary manner.”

Bless agrees: “The biggest risks that we see is the risk of losing data, and inaccurate outcomes. Uncontrolled usage of AI could mean confidential information ends up in the training sets of public models. That is something that absolutely has to be avoided as a responsible player in the field.”

“We are already running various proof of concepts with different technologies and different vendors” 

Lyndon Arnold,
Triton Partners

EQT believes it was the first private equity firm to use ChatGPT Enterprise in order to ensure the integrity of its data. However, Gobby says this isn’t a foolproof option. “Enterprise ChatGPT is good, but it isn’t fully enterprise ready.

“We talk with them on a regular basis, and they are bringing out lots of new measures to try and meet customer requirements. But the challenge is simply the speed that this is all moving. I don’t think anyone expected the use of OpenAI to explode in quite the way that it has and that means we don’t yet have the level of operational security or functionality that you might have working inside Microsoft, for example.”

Cyber-risk is another dominant concern. “Generative AI has provided a great boost for cyber-criminals, and it is vital to make sure that you have the right defenses in place,” says Bless.

Gunolf points in particular to the potential for cyber-criminals to use generative AI to enhance phishing threats by imitating the voice of known individuals in a firm.

Other challenges include the biasing of foundation models, says Gobby. “There is the risk, put simply, of getting the wrong results. Private equity firms also have to make sure they are not just generating generic outputs that any other private equity firm could get.”

“It’s a question of accuracy,” adds Arnold. “The way that generative AI works means it is open to hallucinations. Any use of the technology for intelligent work must be checked and double checked. ChatGPT works on large language models, drawing on billions of pieces of data from the internet, and the results are not always what we would want them to be. That is precisely why it should only ever be used as an aid to the individual, and not a substitute.”

Investors are inevitably asking a lot of questions about the advent of generative AI, encompassing both the threats and opportunities. “We see a lot of interest from LPs in terms of what is possible,” says Bless, “but we always frame that in the context of risk management, to ensure investors know we are focused on using AI responsibly.”

“Our approach,” says Arnold, “is designed to make sure our LPs recognize that we are alive to the opportunities that generative AI presents, while emphasizing that it is something we are exploring in a secure and tested way, without any compromise to data.”

Gunolf adds: “Our LPs are very excited about AI, like the rest of the world, but the integrity of data is paramount to them. We are very clear about our compliance guidelines and restrictive approach.”

Establishing protocols

A combination of the ubiquitous buzz around generative AI and these significant risks means private equity firms are having to accelerate the roll-out of governance protocols to manage its use.

For Bless, the starting point needs to be understanding what is already taking place within the firm. There has been a huge amount of experimentation going on in individual divisions in businesses everywhere, but this needs to be harnessed and institutionalized if rewards are to be maximized, and risks mitigated.

“If you don’t have an inventory of activity and the experiments that are going on, then you can’t properly govern them, so that is the first step,” Bless says. “Once that is in place, you need to carry out risk assessments on each individual case. That will either lead to a decision to press pause if the analysis shows it is too risky at this stage, or it will give you conviction to proceed.”

EQT, meanwhile, established an ethical compliance forum specifically targeting the use of AI even before it launched its pilot program. “We wanted to make sure we were doing things right,” says Gunolf.

“Even though we are using ChatGPT Enterprise, we are still restrictive in what we allow. For example, the entry of personal data, such as performance reviews, is forbidden, which is essential from a GDPR compliance perspective. The same is true for client data. And while we are strict about what can be inputted, we are also very specific in terms of how outputs can be used.”

Triton does not allow the inputting of any firm information into OpenAI ChatGPT. “It can be used as a tool to query and analyze external information, but we have strict rules about not loading any Triton data into a public tool,” says Arnold. “We allow more flexibility when testing against our own internal data because we know the appropriate controls are in place.”

Looking forward, however, Triton plans to develop a full and in-depth AI policy governing what the technology can be used for and, more importantly, what it can’t. “We are developing training protocols for staff so that they understand the appropriate usages,” Arnold explains. “We use it as an aid and nothing more. We will, of course, continue to follow the appropriate regulations and guidance from industry bodies. In addition, I keep a close relationship with my fellow private equity CTOs to keep abreast of new developments.

“Finally, like all other CTOs, it is my responsibility to ensure that the right senior leadership committees and boards are kept informed, not only of the possibilities the technology presents, but also the drawbacks and risks.”