GenAI investing: the potential and pitfalls for Private Equity investors
I recently had the privilege of attending a lunch discussion about GenAI in London with Arthur Mensch, CEO and co-founder of Mistral AI, and Jack Hidary, CEO of SandboxAQ. There were about 20 top US investors on hand — all asset managers or PE investors — to listen to a robust debate about LLMs and the subsequent applications.
It is perfectly normal that investors want to grapple with all the essential technical details when an explosive innovation emerges. This is a period of huge change that may well go beyond a platform shift, so it’s essential to be deeply engaged. At such an early stage, there are many unknowns, such as which version of the technology — open source or closed — or model size — large or small — will ultimately win.
For Private Equity investors, the opportunities in GenAI are not the companies building the infrastructure (i.e., the models). This remains better suited for venture capital, big tech, and large asset managers because the costs (and risks) are still too high. Instead, PE should focus on the companies deploying it in their business. PE managers must own the companies integrating GenAI into their products and services and avoid the ones that are not. As in any platform shift, the first investment move is to move away from the “losers,” i.e., those companies that face rapid technological obsolescence because they cannot transform themselves to remain competitive in the new reality.
GenAI will create massive opportunities in two ways. First, it will drive dramatic improvements in productivity and efficiency that will lead to significant EBITDA improvement. That initial wave will be followed by an even more radical transformation of existing companies, allowing them to build entirely new business models. At the same time, this new technology comes with new costs that can be hard to identify.
Let’s break down those three categories for evaluating GenAI investments for PE portfolio companies:
- Productivity gains
- Business model transformation
- Hidden costs
Productivity Gains
Gen AI will lead to substantial productivity gains. Think of this as Stage 1 of GenAI’s impact. Let’s take the example of three markets that are particularly ripe to benefit from GenAI and on which we have completed extensive work at the firm: Online education, SaaS, and Digital Marketing.
GenAI will allow companies in these sectors to improve their P&L without fundamentally changing their product service. A BCG study with several leading universities found that GenAI could drive productivity gains ranging from 22% to 42% across different departments.
We expect it to have a powerful impact in three areas:
- R&D: GenAI will replace many entry-level (and not only) software coding jobs. Companies may see as much as 30% productivity gains thanks to such efficient automation.
- Marketing: GenAI can produce articles, marketing messages, and emails. This drives greater efficiency and reach while significantly lowering costs. An MIT study last year found that ChatGPT reduced by 40% the time for writers to complete a project while output quality increased by 18%.
- G&A: Software that already runs on AI will get even faster and more powerful. So any company that uses tools like Saleforce or Google docs will become even more productive as those services bake GenAI into their platforms.
Key GenAI takeaway: Rethink productivity expectations.
Business model transformation
This is where GenAI becomes a black swan event for PE. Businesses that leverage GenAI to reinvent the core business will thrive and achieve substantial competitive advantages. Companies that don’t will disappear.
PE managers must urgently grasp this reality. GenAI will create additional value for customers and clients by harnessing data. This will be critical over FY24 because the tech is here now. This will be the year of deployments.
Let’s see how this looks across three use cases.
Online education: massive personalization
Chegg, an American education services company, built its business by providing a physical textbook rental service and grew into a digital learning platform. Last April, the company launched CheggMate, a new GenAI-powered study companion built on GPT-4. In this case, GenAI enables personalized tests and other lessons tailored to a student’s learning experience, including enhanced conversational features. Despite this early introduction of the technology, Chegg’s has seen a decline in its revenues and user base — and has been walloped in the stock market as a result — highlighting the threat that GenAI can pose to a business.
Meanwhile, language learning app Duolingo released a new subscription tier last year called Duolingo Max. The advanced service offers personalized features such as Explain My Answer — a chat feature that explains why a response was correct or incorrect — and Roleplay — a responsive conversational tool that adapts to the user’s replies. The company charges $14 per month for Duolingo Max versus $6.99 per month for its classic subscription.
SaaS: Delivering more value
That same kind of personalization will reshape software applications that are embedded in workflows. They will do this by leveraging propriety data and the inherent network effects in SaaS platforms. The result will be a vastly improved customer experience.
Consider Salesforce, which has become a prominent GenAI evangelist. Last year, Salesforce released Einstein GPT, which creates personalized content for every customer using Salesforce. The super-charged platform generates customized emails to send customers, responds to customer queries, and auto-generates code for developers. The company has bet big on this technology with Salesforce Ventures’ $250 million Generative AI Fund to back GenAI startups.
Digital Marketing
Salesforce is also deploying the power of GenAI to help its customers reinvent marketing. Not long after releasing Einstein GPT, the company unveiled Marketing GPT. The latter included features that let marketers generate personalized messages — text and visual — and content to reach customers across all platforms, including email, web, and mobile. Marketing GPT automates the process of creating different audience segments for more precise targeting. Natural language prompts and AI-driven recommendations help marketers refine their messaging.
GenAI threatens generalist digital agencies, as the industry has known, and requires leveling up the game either through specialization or through broadening the offering spectrum beyond marketing to broader consulting, for example.
These GenAI features create a massive competitive advantage for companies in these three markets — online education, SaaS, and digital marketing. Conversely, a company that has not done the same has virtually no chance of staying competitive in the medium-long term. Such businesses could become irrelevant in months.
As Salesforce CEO Marc Benioff said on the company’s Q2 ’24 earnings call: “The reality is every company will undergo an AI transformation with the customer at the center because every AI transformation begins and ends with the customer.”
Key GenAI takeaway: This will make some businesses irrelevant. Now is the time to pivot.
Hidden Costs
Private Equity investors who grasp the first two topics will now understand the urgency and opportunity that GenAI has created. But it’s equally important that they recognize the possibility of unpleasant surprises in the form of undisclosed costs.
Many companies adopting GenAI need to be more transparent about what they are spending. This is becoming an essential issue for the financial industry, and public and private companies must make more disclosures. If the costs aren’t clear, then it’s impossible to know whether the gains from new business models powered by GenAI deliver net positives to the bottom line, i.e., the EBITDA margin.
These hidden costs can come in many forms. For instance, a company may need to overhaul its legacy infrastructure to have the right tech stack and data-sharing capabilities to optimize for GenAI. Such investments may be unavoidable due to the computationally intensive process of training and deploying GenAI. These new models may also require data storage and sharing re-architecting to meet any privacy or regulatory obligations. On top of all this, there could be costs for MLOps, data security, and other unexpected items.
Then there is the question of which foundational model — large or small — a company uses. Depending on the choice, the inference costs — the cost of running a query by calling the LLM — can vary dramatically by factors of hundreds. There are ways to reduce the inference cost by using a smaller model or open source, but the costs should be clear.
What compounds this issue for investors is that companies are tucking these GenAI-related expenses into places like R&D budgets. That can be misleading. Some of these expenses should be included as direct costs, ideally with a separate reporting line. Investors must demand clarity on this spending and ensure they have sufficient visibility.
Key GenAI takeaway: Hidden costs could negate the benefits.
****
Key questions
When analyzing the benefits and costs of GenAI, it’s not just a matter of productivity or revenues. The fundamental question is this: How do you unleash value?
Here are key questions a PE investor should ask when considering a new investment or as part of the value creation plan of its portfolio company:
- How does GenAI impact my assets? Not just from a cost perspective but from a broader perspective.
- What are the costs in the short term? What is the opex/capex structure that I need to have? What ROI am I expecting?
- Do I have the right team to analyze and execute a GenAI strategy?
Technologies like GenAI that create such a fundamental paradigm shift can make it difficult for investors to scrutinize and quantify the impact because there is almost no historical data for benchmarking. But there are still ways to evaluate the financial aspects of such disruptive innovations. Understanding how to do that with GenAI should be a major priority for PE investors in 2024, a pivotal year for digital transformation.
