April 7, 2023
The Future of Finance AI: Winners and Returns
Interested in how PE firms are using generative AI to improve their investment processes *today*? In the final article of our "Return On AI" series, we give you a peak under the hood of how we think about solving efficiency challenges in the due diligence space, and how investors we talk to think about implementing similar systems.

AI is delivering generational levels of change and opportunity to all knowledge work industries, and the private finance sector is no exception. Two months ago, we kicked off this first article series for Capital AI, introducing the concept of “R-AI”: our view that early adopters of AI in finance have a unique potential for outsized returns. In just these last two months, Google, OpenAI, Meta and Anthropic all released game-changing models. Several top research and consulting firms released reports on the role of AI in the private investing space. And more readers than ever have reached out to learn about what we are working on at Keye.

If you haven’t already, we highly recommend skimming our last few articles for a deeper dive on the developments in the space, and some wisdom from top voices in the industry around how to leverage these new technologies for increased returns.

To recap, our view is that staying up-to-date on the impacts of generative AI is no longer optional – both investors and portfolio companies will see the effects unfold over the coming years, and 80% of PE firms expect to use it in some capacity. But in the race to leverage AI, how managers implement solutions from large language models to vertical applications could well be more critical than merely adopting them.

Our Perspective

To wrap up our first series, we will share our view – informed by hundreds of conversations with investors and technologists – on what will separate winners from laggards in the space, and what success looks like for those who get it right. How much value is generated matters, and according to McKinsey it could be up to $4.4 Trillion a year – with a “T”. But increasingly, investors are recognizing the rise of competition for AI’s value and developing an interest in capturing a larger share for themselves.

Increased efficiency will deliver a new wave of time and cost savings to PE firms, but over the long run, returns will be sustained by defensible moats in strategy. Recently, the trending view among industry analysts has been that supply ultimately becomes constrained as you go toward either end of the stack. In common parlance, there has been a boom of investment into LLMs with theoretically unlimited scale - until you hit compute constraints, that is. Going the other direction, LLMs will continue to provide tireless analytical capabilities in applications, but there are only so many user-facing companies and distribution channels to provide the right service level to a customer.

Because of this, PE managers need to consider the following factors to develop strategic fit with an AI provider:

  • Customization vs. Specialization: If LLMs and applications continue to advance in their ability to be easily customized, they could potentially supplant the need for both vertical applications and general-use knowledge engines by offering a "best of both worlds" scenario. However, this assumes that companies have the resources and expertise to effectively tailor these LLMs to their specific needs, which may not always be the case.
  • Integration and Ease of Use: Vertical applications offer the advantage of being ready to use with minimal setup, designed specifically for industry-specific tasks. For many firms, especially those without significant in-house AI expertise, this could remain a compelling reason to prefer specialized solutions.
  • Cost-effectiveness: As LLMs become more affordable, the cost advantage may shift towards customizable solutions. However, the total cost of ownership, including maintenance and updates, should also be considered, as specialized applications may offer more predictable long-term costs.
  • Depth of Insights: While general LLMs are becoming increasingly powerful, vertical applications may continue to offer deeper, more actionable insights in specific areas due to their specialized data models and integration with industry-specific data sources.

The Chase for Fast-Moving Tech

Not only does investment firms’ vision for AI need to be carefully crafted - implementation must be navigated with consideration for how the technology may unfold in the coming months and years.

We are constantly observing a few factors that continue to change quickly, with a lack of consensus around how they will shake out:

  • Transparency and Explainability: AI models need to provide superhuman speed and levels of insight to be helpful, but users are also demanding they be transparent and explainable to ensure trust and avoid potential biases. Today, growing context windows make experts optimistic that this problem is solvable, but others believe that we will never truly be able to explain AI’s “logic”.
  • Data Security and Privacy: Robust data security and privacy protocols are essential to protecting sensitive financial information. Technology for siloing and securing data is relatively straightforward, but evolving standards, regulatory landscapes and cyberthreats keep this area top of mind for many decision-makers.
  • Human-in-the-Loop Approach: While AI can automate tasks and provide valuable insights, human expertise and judgment remain crucial in the investment decision-making process. This leaves some investors hesitant to completely change their processes - which is understandable. Others are excited for just how transformative of a technology it will be, despite the need for human involvement.

Integrating AI with existing workflows and systems

Building on the idea of the “human-in-the-loop” approach, integrating AI isn't all about replacing the old with the new. It's about weaving new threads into existing decision-making processes to strengthen investment strategies and portfolio management more generally.

Let’s return to the example of AI-augmented due diligence at a private equity firm.

Typical low-tech diligence process at a mid-market PE firm; source(s): Keye, IBM

Today, it is not only possible, but sometimes the norm, that a deal team at a mid-market PE shop spend up to two weeks in intensive diligence on a deal that is ultimately nixed for having a volatile financial profile, or a risk-heavy IP pipeline. While these may sound like issues that should be apparent on the surface, it often takes weeks of research to dig beneath a polished investor memo and understand the company’s real profile in the context of the industry.

We are seeing promising results from our own product testing at Keye, as well as in the market, that generative AI, alongside RAG systems and the right data solutions, can vastly outperform human-only teams. Our experiments show that deal team + AI collaboration will be the norm in due diligence In the very near future, generating both operational efficiencies prized in the private equity industry, as well as strategic advantages for investors.

Today's potential for an AI-augmented & accelerated due diligence process. Source(s): Keye, IBM

With both LLMs and humans in the loop, AI-augmented due diligence can generate roughly the first 80% of the relevant analysis within just a few hours of organizing and uploading VDR files. Firms can then allocate more time to discussion and digging deeper, spending valuable time tackling the tougher questions around a deal. Alternatively, deal teams might be able to reject a larger number of potential deals much faster, allowing them to bid on more deals, or deals that would have previously seemed too murky or complex to diligence.

Looking Ahead: How To Maximize ROI

So how should investment firms, whose primary strengths are to make wise investment and management decisions, seek to develop the capabilities that will allow them to capture these gains? As we have discussed in our last article, some are entering into costly development races with LLM providers or venture firms to be the first to market among their peer set. Others are purchasing app after app off-the-shelf based on what is immediately available to test their way to the best solution.

While we don’t necessarily recommend against trying these approaches, we do suggest proceeding with caution. Both of these strategies push firms to invest without full understanding of their risks, including switching costs with cloud and application providers, as well as unfavorable cost structures (high upfront investment and maintenance costs).

Strategic partnerships are one way to maximize learning and internal capabilities, while also softening the risk of potentially over-investing in applications and avoiding lock-in with cloud providers. As we discussed in our article on the AI stack, specialized application providers who understand both the technology and their clients' industry offer the highest potential as potential co-developers. Consulting firms may assist with implementation or strategy, but for mid-market firms where prolonged engagements become prohibitively expensive, the startup landscape offers opportunities to work with high-energy, full-service teams at a fraction of the cost.

In-house development vs. design partnership. Source: Keye

Customized development on a scalable platform is the approach we have seen work most effectively, enabling investors to build highly usable solutions in a cost effective manner, with maximal optionality for future evolutions in the space. In a field where a nuanced investment edge can lead to outsized gains, generic solutions lead to generic results. Specialized AI partnerships are not an expense but an investment in precision-engineered success.

Top vertical generative AI opportunities. Source: CB Insights

A recent report by CB Insights agrees, noting financial analysis as a top 10 opportunity for sticky AI solutions. By embracing AI and strategically integrating its capabilities via partnerships, firms can catapult ahead of the pack when it comes to alpha generation. As the technology matures and the ecosystem evolves, the future of private finance promises to be one powered by intelligent augmentation, shaping the landscape for years to come.

What’s next?

The winners in this space will be those who view AI not as a one-time implementation but as a continuous journey of innovation and adaptation.

While this concludes our first article series on returns from adopting AI, we are looking forward to bringing you even more exciting and relevant content. As our design partnership program progresses at Keye, we are excited to share case studies from our early-stage partners, deep dives into the technology enabling AI-augmented finance, and more exciting use cases we are seeing PE leaders adopt.

If you are interested in learning more, please reach out to us here on Linkedin or at founders@keye.co. We would love to hear your ideas for content and research too, so don’t hesitate to leave a comment below. Thanks for following along so far, and stay tuned!

Recent Blogs posted from our team

See more from our latest series on AI in the private finance space.

Want to stay in the loop?

Enter your email and our team will provide regular updates.

Thank you!

Thank you!
Your submission has been received!
Something went wrong! Try again later