How Enterprise SaaS Startups Are Made
The origin story of enterprise SaaS companies can be abstracted to a quite repetitive narrative arch. It is a simplification, but it goes something like:
1. We are experts in this space
2. As a result, we learned everyone in this space has this problem
3. We came up with the best solution applying our expertise and technical ability
4. We can build and scale this solution
5. Companies will buy this solution to their problem
Building a successful SaaS company is by no means easy just because the story can be abstracted but many successful origins resemble the above.
How AI changes “Build or Buy”
Whether one should buy a solution to their problem, potentially from a startup, or build it themselves has always been a question for any enterprise. “Build or buy” is thus also a key consideration in the abbreviated startup ‘story’ told above. For instance, in my previous capacity meeting early stage founders as a venture investor, significant time was spent discussing a potential customers motivations. Even if a startup has built the best solution, it does not always mean the described potential customers will buy it. Founders and their investors put themselves in the shoes of CIOs and CISOs who must make the decision whether to solve a problem by purchasing a SaaS solution or building software internally. They are weighing customization and flexibility, cost and time, scalability and evolution, and their technical capabilities as an organization. It is not a simple equation and AI has changed it.
AI has and will continue to make companies more capability of solving problems for themselves, in cases where they might have previously purchased software. In general, and particularly during a contractionary period in a market, enterprises are highly motivated to solve their own problems, rather than buying software, if it is 1) the most economical approach and 2) within their capabilities. AI will increasingly prove both statement 1 and 2 to be more frequently true.
First, software development has been increasingly abstracted by large language models such that more people can “code” program logic through natural language, lowering the barrier to entry in terms of enterprise capabilities. Additionally, large language models increasingly represent the foundation of software capabilities and are often highly general purpose, meaning less software will do more tasks. Software development is not only becoming easier, the resulting software is also becoming more capable.
In summary, enterprises are more motivated and more capable of solving their own problems, ie “building rather than buying”, as a result of AI. Sounds like a recipe for trouble for SaaS startups, particularly as the public markets punish the likes of $PATH and $CRM for slowing growth and sales friction this quarter!
Fortunately, given I intend to pursue a career investing in the space, this is not the end of SaaS. Instead, as with any consequential technology evolution, it is a repositioning of the opportunity set.
I believe there are two key opportunity areas in SaaS that are often intersecting and of particular interest in this new build with AI or buy paradox:
1. The hard: complex problems faced industry-wide that have implications of immense breadth and depth on the customer.
2. The legacy: industries that lack software engineering capabilities internally to build solutions.
The Hard
As Akash Kumar of Matrix India put it in the worthwhile read AI is eating the software that is eating the world, “We are now nearing the end of an era in software where businesses solving the problem statement of “you don’t need to write code for non-core task XYZ, here is a SaaS”, shall see value erosion.”
As AI has increased the “build” capabilities of enterprises, simple and non-core tasks and problems will be solved “in-house” rather than paying for software. Additionally, when companies do pay for non-core simple software, there will be more pricing pressure placed on the providers of such software.
To be a high-value SaaS company with pricing power, the problem to be solved must be impactful on a dollar basis and in terms of a potential customers workflow.
Put simply, the bar has been raised and the problem you are solving must be hard.
The Legacy
While the eventual state is every company having software engineering capabilities powered by AI, there will be a long tail of laggard industries. Specifically, legacy industries where companies have not historically spent significant money on software development. This includes areas like restaurants, construction, life sciences, and industrial manufacturing. Today, whether it be Procore in construction or Veeva in life sciences, companies in this space have shown a strong preference for vertical SaaS offerings.
There are several potential reasons for this phenomenon and why it is likely to continue. Many purchasing decisions in this space are highly referral-based. You don’t see much Gartner or G2 coverage for construction software! Instead, there is CONEXPO-CON/AGG every 3 years in Las Vegas that features 2,000 exhibitors across 2.8 million square feet of exhibit space which construction companies make their technology purchases.
Additionally, these industries are often characterized by a culture of “buy over build”, specifically strong "Nobody Gets Fired For Buying IBM" effects. Many large players in legacy industries depend on consultants for technology initiatives and, when they are making their own buying decisions, the career calculus of making a bet on a big software initiative does not often point towards “build”. Instead, buying the gold standard has never put an executives’ head of the chopping block. Becoming this gold standard is likely unicorn status for any entrepreneur.
In the end...
AI does not spell the death of today’s software business, as some have declared. But it does, as with any important technology, fundamentally changing the opportunity set by shifting the key “build or buy” decision.