AI is Not the Death of Vertical SaaS
What the All-In Podcast gets wrong about vertical software
On this past week’s All-In Podcast, the “Besties” held their annual predictions episode where they each share predictions across a number of political and business categories for the year ahead. I am an avid listener of the podcast — I’ve listened to the last ~120 episodes and have referred dozens of friends and family members to start listening— and even traveled across the country to attend their All-In Summit in LA in 2023. I find myself almost equally agreeing and disagreeing with what the four hosts have to say, and come back to listening each week because of the informed discussion and multiple perspectives offered on most topics.
In their predictions episode released this past Friday, one of the categories toward the end (tune to 58:40 if you want to listen for yourself) was to share a prediction on the “worst-performing asset in 2024”. David Friedberg, a very successful founder-turned-investor/incubator-turned-founder/CEO-once-again, shared his prediction that vertical software companies will be the worst-performing asset in 2024. His rationale was that AI has enabled companies to build, with no-to-low code, their own internal software solutions at a much lower cost than the price of purchasing comparable software from a 3rd-party. As a result of this new ability to build internal tools more seamlessly, there will be increased churn and pricing compression resulting in deflationary SaaS pricing power.
Friedberg cited a personal example that either informed or reinforced this prediction. His company had been spending $5k/seat/year for a vertical-specific data management product. When one of his engineers learned the high cost of the product, they decided to build it themself, presumably utilizing some AI tools to streamline the development. As a result, Friedberg’s company will now churn from the product they were spending $5k/seat/year on and, in addition, his engineer will open-source the code for this product or sell the product for a very low price to others in the industry, thereby destroying the pricing power of the product they are churning from. By extension, Friedberg believes this will be a recurring phenomenon in many other companies and across all industries. (Friedberg did caveat that there exists a per-seat per-month cost threshold where it makes less sense to build internally vs. simply buying externally, but didn’t provide a precise cutoff.)
While I believe this argument/opinion has merit, I am firmly on the opposite end of this “short”, and our firm has certainly continued to put our money behind our “long” position in vertical SaaS businesses such as Carefeed, Endear, AireXpert, Parting Pro, etc. As I’ve written about previously, we believe that vertical software is an incredibly resilient category. Since I focused that post more so on the turbulent economic market environment and did not discuss vertical software’s resiliency to major technological shifts (such as AI), I figured this would be a great time to briefly share why I think vertical SaaS is very well-situated to withstand AI (i.e. why I think Friedberg’s prediction will be proved wrong).
First, let’s start with a counter to Friedberg’s core argument, which is that AI will enable teams to seamlessly build internal products that are close to parity with best-in-class third-party SaaS products. While AI has certainly streamlined software development and turned good engineers into “10x engineers” (and “10x engineers” into “100x engineers”), many of the end-markets where customers are purchasing vertical software don’t have technical teams and internal capabilities to leverage these AI tools for building software. David’s genetics company, Ohalo, is surely a team comprised of incredibly talented and technical people given the nature of their business; however, having an engineer who can take their own software development abilities, combine them with AI, and build a tool to replace an existing SaaS tool, is not likely to be the case for most of the companies operating in the end-markets that are targeted by vertical SaaS.
As a proxy, let’s look at the top 25 publicly-traded vertical SaaS businesses by market cap (as of 1/11/2024):
Many of these vertical SaaS end-markets contain businesses that do not have the technical talent, and therefore internal capabilities, to build internal software products: restaurants, general contractors, call centers, architecture firms, property managers, nonprofits, wealth managers, e-commerce brands, etc. Furthermore, in many of these industries like construction, the companies have historically never spent money on technology. If you want to see one simple example: walk onto a construction site and what you’ll see is that there are still post-it notes and whiteboards hanging around. I believe it will be difficult to go from literal pen and paper to building their own tools internally using AI. In reality, many of these industries are still so early in their tech adoption phase and will purchase off-the-shelf solutions tailor-made for their industry’s needs.
Second, even for those industries where you might have a technical person or team on staff — looking at the above list of publicly-traded vertical SaaS businesses, that might include chip manufacturing, banking, higher ed, and energy — the idea that companies within these industries will focus time and human capital on building internal software vs. spending that time/resources on their core competency seems unlikely. Even if these companies could spend a lot less money on internally-developed software than purchasing third-party software, cost cannot be the only factor in the equation: you must also factor in human psychology. As the old saying goes: “you don’t get fired for hiring IBM or McKinsey.” Well, if you’re a company operating in an industry like construction or life sciences, you also don’t get fired for purchasing Procore or Veeva. The same gold standard rationale applies within specific industries, and for an executive to make a decision to build software internally (and continually upkeep that software, including all the data privacy and other security and regulatory issues that come with maintaining software) vs. purchase software from the industry-standard vertical SaaS provider is a major career risk.
Lastly, if/when AI does make it easy to build internal products with fewer resources, the same efficiency benefits would apply to the 3rd-party vertical SaaS companies, but at an even larger scale. These vertical SaaS businesses have their own teams of engineers that will either (a) become significantly more productive, thereby being able to sufficiently launch new features/products and stay way ahead of any internal products being built by leaner teams, justifying their pricing, or (2) those teams of engineers will be reduced (since you can now build more with fewer people) allowing the 3rd-party vertical SaaS businesses to have flexibility in decreasing their prices while keeping their operating margins the same. Said a bit differently, the benefit of AI for vertical software will either be helping historically labor-constrained vertical SaaS companies build a lot better software through efficiency gains, or democratizing access to vertical software through a combination of bringing down pricing and layering in AI copilots that make the software even smarter and simpler to use. This argument is the closest to David’s deflation argument, however this should still lead to similar levels of profitability and higher volume of sales (given demand tends to go up when pricing comes down and tools become better and simpler to use/adopt), which would result in vertical SaaS being in a similar or better place to today (and far from the worst-performing asset in 2024).
I welcome any additional thoughts or feedback on the above. We spend a lot of time in vertical software and are always interested in hearing how founders and other investors are thinking about what’s ahead.
