Jan 2026
Recently there has been an indiscriminate selloff in SaaS stocks, partially driven by fear that AI is advancing so quickly that anyone can build and replicate software. As a whole, it may be broadly justified. If you suddenly have 100s of new competitors flooding your space and now need to pay inference costs (to serve the AI product you need to build in order to keep up), your pricing power, margin profile and future growth potential have decreased.
But it's not completely clear that that assertion is definitely true (yet). A more accurate take may be that a competitor can get to 80% feature parity faster than ever before. Yet, getting that final 20% can often be the most painful. The reality of whether any particular SaaS incumbent is impaired or thrives very much requires a case-by-case consideration.
Nevertheless, here are some high level thoughts on how incumbents and upstarts will be affected by the AI avalanche.
How incumbents will fare
An interesting observation from my vantage point is that many incumbents are not asleep at the wheel. Several “legacy” incumbents (e.g. Epic) have added AI functionality (AI scribes) and others, like ServiceTitan, have been swift to block startups from accessing their core system-of-record data. Many other systems of record are still proving to be sticky: famously, OpenAI and Anthropic both run on Workday.
This isn't to minimize the clear risks that incumbents face:
- Obviously, seat-based pricing is a threatened business model in an AI age where you can get closer to delivering end outcomes directly.
- Several incumbents (Salesforce, Adobe, Workday) still price on seats, but many have smartly moved towards usage-based pricing (e.g. Crowdstrike, Datadog)
- When agents perform most of the work, less time is spent in the core SOR and could be spent in the AI platform atop the SOR. It's an open question whether this minimizes the importance of the SOR
- Incumbents are prone to adding AI incrementally as it's difficult to rapidly change the cultural and financial direction of a big ship. But since AI technology can be so impactful (often a 10x better UX), incrementality can leave you at risk of getting outflanked by an upstart that's solving the problem in a way that is an order of magnitude better (e.g. way cheaper, way faster)
These are all the factors that need to be weighed when considering any specific company’s longer term prospects.
Late stage private SaaS
What’s interesting is that the same risk factors that apply to public SaaS can often also apply to late-stage private companies whose architectures and cultures, generally speaking, predate AI.
For example, late-stage companies like Clay.com have complex interfaces that take customers substantial time investment to learn. Many Clay experts famously spend weeks or months learning how to fully use the functionality. A lot of that functionality can now be abstracted away with natural language. And a lot of GTM engineering (the category that Clay plays in) is being triggered directly from the command line, through tools like Claude Code. The thing that worked in Clay’s favor could very plausibly become a huge achilles heel.
Similarly, an incumbent's core product could have been built for a problem that's much less relevant in the age of AI (e.g. SEO optimization). The larger the customer and revenue base, the bigger the challenge to burn the boats and start afresh.
There is also a whole cohort of PE-backed companies who are most definitely “pre AI” and will need to tell a very different AI story in order to go public. We’ve already started to see a lot of bolt-on acquisitions here to try and offset that stigma.
Without steely leadership, it can be quite difficult to completely rewire your offering and business model. This change can be even harder when you're VC backed (i.e. not yet profitable), since funding for late-stage companies has completely dried up unless you’re growing very quickly.
AI-first companies
So what does all of this mean for AI startups?
First, it’s not all bad news. Vertical AI has the potential to permeate every single services sector of the economy. In many cases, new vertical AI upstarts are going up against services incumbents who have no or little ability to build software. So there is more opportunity than ever to find a problem to solve and to solve it with AI in a way that is 10x better than it was done before.
The main thing is that founders need to think about defensibility from the early days. In an environment where you have 100 competitors - often including incumbents and the foundational model companies - defensibility becomes something to build from the early days.
Below are some ways you can do this:
- Incorporate a network effect into your business offering where possible
- Data network effects are now real (for the first time!) -- does your application generate proprietary and valuable data? If so, use it
- Differentiate on value proposition.
- One differentiated value proposition is selling the end outcome, not just a tool (which may require bundling software and services)
- Another may be to have several 10x better solutions. Being AI native allows you to imagine radically better ways of solving a problem for customers
- All-in-one offering can be hugely valuable -- as it broadens the options for wedge products and allows you to undercut competitors on a specific product line
- Distribution matters more than ever. Innovating on distribution (e.g. shoring up proprietary partnerships) can add defensibility
- Embrace customization: it used to be the case that you would reject customers who required custom offerings. But if AI can be used to rapidly produce custom instances for customers, it is now something to lean into
- GTM know-how
- If you are selling to SMBs, there is real know-how in certain distribution tactics
- Shoring up proprietary partnerships
- Signing long-term contracts