AI In Private Equity: A Framework For Evaluation
It's all the buzz but, how does it work?
Happy Sunday and welcome to Investing in AI. Be sure to check out our AI Innovators Podcast for interesting discussions with the people building the future of AI.
Today I want to talk about applying AI to the private equity model. Everyone seems to be talking about this but, almost no one seems to actually know how to do it. I first wrote about this in 2017, pointing out that with the increasing importance of data for AI, there were probably opportunities to buy companies for valuable data sets they didn’t realize were so valuable. Since then, I’ve had the chance to lead or be part of 4 acquisitions that took place through an AI lens. The lessons learned could fill a book but I will start with just this post to layout the framework I’ve used to think through these acquisitions.
The reason this model is so interesting is that it has the potential to give you venture style upside with private equity style downside. AI can be a massive accelerant to growth leading to huge outcomes, but by buying companies with existing cash flows, the downside risk isn’t as likely to be a zero as it is in traditional venture capital.
There are three levels of AI that can be applied to existing companies when you buy them out. Level 1 is Common AI, the type of stuff everyone should be doing. Level 2 is Proprietary AI, the things that are unique to that specific business. Level 3 is Future AI, the idea that a company has assets that will increase in value in the future.
Common AI - This is what most people look for when they look at buying and “AI-ifying” a company. These are the simple things everyone should be doing like using generative AI for marketing personalization or installing a customer support chatbot. This about using best of breed AI tools for your operations. It can make a difference but also, lots of people are going to do this so its unclear if it can be a long term competitive advantage. If you are a first mover in an area where there is a significant amount of cost tied up in some line item that AI can now tackle, that may be a good place to look. The most common example is people buying an applying AI to call centers. It will definitely happen but, I’m not yet convinced that whoever does it will have any long term defensibility.
Proprietary AI - This is the stuff that is unique to the business you acquire. For example, one of the businesses we bought had products that if you uploaded a picture of your pet, they would print out your pets face on socks or pajamas or other pieces of clothing. To do so, they had humans trace out the face of the pet in the uploaded pictures so they were removed and separated from the background. This was an expense that cost about $1M per year. We built an AI model to do this work. It found and traced the pet face in uploaded pictures, and saved the company an estimated $800K per year.
Future AI - This is stuff that will be more valuable in the future based on where AI is going. I wrote about this for our buyout of FeedbackNow. One big part of the investment thesis was that AI models will need more real-time data in the future, and FeedbackNow has interesting and proprietary real time data.
We also evaluate companies for acquisition on the non-AI things as well. Seeing opportunities for traditional areas of business improvement helps lower the risk of the acquisition as well, if the AI transformations don’t pan out. The primary difference with traditional private equity is that we are looking more at innovation opportunities rather than just cost cutting opportunities.
There still aren’t very many people who can look at an income statement and company operations and figure out where to apply AI to improve EBITDA. It requires a broad set of experiences buying companies, selling companies, running companies, and applying AI to workflows. The team here has done quite a bit of it, compared to where we are in the market. So if you have a business that you think could benefit from an AI transformation, and want an investment partner who can help with that, please reach out.
Thanks for reading.
Hey Rob! Interesting ideas, especially how AI in PE can offer venture-like upside but with lower downside risk. This week, I'm posting about a venture firm using AI to eliminate human bias in deal selection. Wondering if you think these AI-driven approaches will trickle down to even earlier stages of investing?
wanted to share some insights on AI adoption patterns I've observed while working with various private equity firms and their portfolio companies.
While cost reduction remains a key driver, the implementation landscape is more nuanced. Effective AI integration typically requires top-down direction, yet I've found that many portfolio company CEOs aren't fully immersed in the technical possibilities or challenges. This creates adoption hurdles that go beyond simply identifying problems and deploying solutions.
An interesting dynamic I've noticed in the last 9-12 months is the emergence of "shadow AI" - employees using their own AI tools independently without cross-company coordination or visibility. This fragmented approach, while showing initiative, often prevents organizations from achieving strategic, enterprise-wide benefits. Sometimes you end up with legacy technology debt driven totally by AI.
Well, over 75% of companies have AI initiatives in place. But whether you ask McKenzie or BCG or anthropic, a lot of those efforts are noisy with no real ROI yet.
I’m definitely seeing strong movement in innovating, but it’s about 25% of the companies where the adoption is happening. In those cases there is strong governance and a strong understanding, but driving the innovation purely from the PE firm doesn’t optimize collaboration with their portfolio companies.