Imagine you walk into your finance job one morning, and, when you open your laptop, there are no Microsoft apps on it. No Outlook, no Powerpoint, and…no Excel. It’s hard to imagine, isn’t it? Microsoft Office isn’t just a bundle of applications – it forms the substrate for how work gets done in “serious” industries like finance and consulting, and has for a long time.

In 2006, Google Docs came out and placed pressure on Microsoft to offer cheaper, cloud-based alternatives full of ‘smart’ features like “Recommended Charts” and “Autofill Column”. Many were initially skeptical of Satya Nadella’s push to the cloud in the early 2010s, and whether enterprises would really leave behind manual versioning and single device access for real-time collaboration and cloud drives.

While Google and Microsoft were building competing products which would have looked very similar to any tech industry outsider, a closer look would tell you that the two productivity suites were actually aimed at rather different use cases and user types. While Google’s Workspace integrated with mass market products like Drive, Search and Gmail, Office sought to retain the power features that made Excel a mainstay of finance work.

As Ben Thompson recently wrote about in Stratechery, this isn’t an accident or even a deliberate tactic on Microsoft’s part. It’s really a reflection of their product philosophy as a whole. In his words:

“In Google’s view, computers help you get things done — and save you time — by doing things for you…In [Microsoft’s] philosophy, the expectation is not that the computer does your work for you, but rather that the computer enables you to do your work better and more efficiently.”

That brings us back to our experience speaking with technology users in the private equity space. Overwhelmingly, the most common piece of feedback we heard from the market for due diligence tools like Keye, was that users (investors) were not ready to upend their existing workflows to use AI. In other words, they wanted the “Microsoft Office” of AI (i.e., the toolkit) and not the “Google Workspace” (the work assistant).

Again, this resonates with Thompson’s words:

It’s one thing to charge $30/month on a per-seat basis to make an amazing new way to work available to all of your employees; it’s another thing entirely — a much more difficult thing — to get all of your employees to change the way they work in order to benefit from your investment, and to make Copilot Pages the “new artifact for the AI age”, in line with the spreadsheet in the personal computer age.

Still, if technologists’ prediction that AI will be the next transformational platform like computing or cloud is true, then Microsoft and Google will need to be a big part of that shift in order to maintain their positions as the leading enterprise work tools.

For AI to be successful in fields like private equity and investment banking, the industry likely needs to move away from being anchored on the “copilot paradigm”, where LLMs are a tool that make analysts and dealmakers ‘smarter’. They also need to get to a point where they are being embedded in investors’ existing workflows, and even replacing parts of those workflows.

Instead, it may be other firms building specific end user applications that will enable the most immediate productivity gains by plugging into existing processes, like Udu for deal sourcing, Keye for due diligence, and Visible.vc for portco monitoring.

Evaluating Microsoft AI for PE

Given the dominance of Office in the financial industry, it is hard to dismiss the case outright that Microsoft will offer compelling AI solutions for its core user base. While it is easier to say that Google’s Gemini will likely play a role elsewhere, Microsoft does seem intent on pushing AI adoption among its traditional customers like investment and consulting firms. In addition, we are already seeing the most adoption among private finance firms of OpenAI and Microsoft AI products.

Certainly, backend solutions like Azure and GPT APIs will continue to be popular components for in-house AI solutions (for the firms that build them). But to what extent will off-the-shelf ChatGPT and Co-Pilot be meaningful additions to the private equity workflow?

AI skeptics like Goldman Sachs’ senior tech analyst Jim Covello cautions that the cost-benefit of AI has to be made more clear-cut. Covello, who has scrutinized AI’s long-term potential in finance, recently pointed to a project within Goldman that leveraged generative AI to automate financial reporting—a task that, while saving 20 minutes per report, came with a six-fold increase in costs. For gains to be made in productivity and ROI, and hence, adoption at financial firms, AI outputs have to be more transformative, cutting process times by over 50% or providing entirely new insights that teams otherwise would have missed.

A large reason for Google and Microsoft’s lead in AI application adoption so far is simply their strategy of embedding it as a feature rather than reimagining it from the ground up. Ease of access, integration and distribution to teams still plays a big role in enterprise adoption. Copilot integrates seamlessly with Microsoft 365 applications, leveraging familiar platforms like Excel and PowerPoint to introduce AI-powered efficiencies, though it may leave users wondering what the best use case is.

Second, these solutions might be seen as “safe bets” - both internally, where security and privacy may be big concerns when it comes to AI, and externally, from a reputational and governance perspective. As PE firms handle sensitive financial data, the chosen AI solution must not only scale with the firm's growing data needs but also meet stringent security standards to protect investment information. As the saying goes, “nobody has ever been fired for buying Microsoft”.

Lastly, there is a technical component that is still somewhat speculative. While utilization of AI features within applications like Excel seems to be low today, it is hard to predict the extent to which Microsoft will be able to improve the underlying LLMs and productize their capabilities. In theory, utilized within Excel, Copilot can automate financial forecasts and scenario analysis, providing PE analysts with robust tools to support investment decisions. In practice, nobody knows whether this can, or will, come to fruition.

The Risk of Sticking with a “Safe Bet”

The largest concern we hear about using big name AI features like Copilot on a more regular basis is simply that it is not yet finding strong use cases among experienced financial professionals within their existing workflows. This essentially makes LLM-based chatbots and tool upgrades solutions in search of a problem, rather than a product designed for investment workflows from the ground up.

While one might argue that the technology is still early, the possibility exists that they might not ever get the level of accuracy, reliability and cost-effectiveness to be truly agentic. Despite advances, the level of precision required in finance often exceeds what a broad-based copilot style solution can deliver consistently, leading to potential financial repercussions due to minor errors.

One alternative some firms consider is building custom solutions on a private cloud such as Microsoft’s Azure AI solution. The cost of implementing and maintaining in-house AI solutions that build on general foundations rather than the right architecture needed for the job, especially in compliance-heavy sectors like finance, can quickly negate any cost-effectiveness, especially as these systems require ongoing updates and security measures to stay relevant and secure.

The other elephant in the room here is existing LLMs’ capabilities in the math and quantitative analysis space. While these computational abilities are improving as we discussed in our last post, the only real way to hedge against the risk of hallucination and failure mode is to keep the actual analysis in the hands of financial professionals for now, and give them the tools to accelerate, verify and enrich their analyses with more tools and data.

Going back to Thompson’s points above, Microsoft seems to be pursuing a path of work automation uncharacteristically to its past product philosophy. Instead of “let me give you some tools and information to help you with that analysis”, the AI’s approach is “let me give this analysis a try by myself”. While the difference may seem subtle, it makes a big difference to investors, who are loath to give up control and visibility into models and investment decisions.

Ultimately, we think this means that advanced, specialized use cases like due diligence, investment decisioning and scenario modeling will continue to remain in the realm of specialized solutions, for now. While we certainly would not discourage anyone who is already using Microsoft from trying Copilot, the reality is that it is unlikely to be transformative in a way that gives deal teams a real advantage moving forward.

For that, we are always available to discuss why we think Keye might be the right fit. If you are ready to learn more, don’t hesitate to reach out at founders@keye.co