Vertical software in the age of GenAI.
Our current point of view on what Generative AI is changing, and what it isn't, for mission-critical vertical software companies. We're excited about where GenAI is taking us, and we look forward to working through this generational change with our partner companies.
We spend a lot of our time debating this, with each other and with the teams we work with. Here's where our thinking has landed.
GenAI is moving faster than any platform shift we've worked through, and the picture keeps changing under us. So we hold these views with conviction, but we also expect to revise them over time.
GenAI is the biggest change to how software gets built, tested, deployed, and used since the move to the cloud began fifteen years ago, and it will change how people interact with software more than anything since the graphical user interface replaced the command line. The consequence we think about most is a new kind of user. Agents, deployed by your customers and increasingly built into your own product, now operate software on behalf of the humans who used to click every button. We can no longer build software assuming that a person is on the other end.
A close second is how much a focused team can now get done. The most ambitious version of the roadmap is increasingly within reach of the team already in the room.
What sets a software company up to win in the GenAI era?
Every software company is grappling right now with whether GenAI will disrupt or compound its value proposition. We don't think the answer is random, and for most companies we don't think it will be all-or-nothing. In our view, three characteristics put an incumbent in a structurally better starting position.
Deep vertical focus.
Software built around the specific workflows, edge cases, and unwritten rules of an industry sits on knowledge that took years to accumulate. That knowledge keeps compounding because customers trust these vendors enough to keep shaping the product as the industry changes. And much of what separates 'world class' from 'good enough' in a vertical doesn't exist anywhere a model could have been trained on it. That comes from being in the trenches with customers every single day.
Critical operations.
Software that runs a company's critical operations is trusted with work that has to be right every time. Many of those operations carry regulatory exposure too, and the consequences land on the customer rather than the vendor: fines, a failed audit, a suspended license. Rules are frequently ambiguous, change often, and sometimes contradict each other, so software trusted to interpret them and document the reasoning is carrying real risk on the customer's behalf. That trust takes years to earn.
Trusted system of record.
The companies that own the important data, the trusted store of what's true, are the ones other systems plug into. In a world of agents, we expect that ownership to decide whether an agent reaches for a system or routes around it. None of this is new to GenAI: owning the record mattered through every earlier wave of machine learning too. But GenAI raises the stakes.
Most companies have one of these. Some have two. The ones with all three have the profile that gets us most excited.
But Andy Grove is still correct that only the paranoid survive.
None of this is permission to sit still. Every software company, well-positioned or not, has work to do: invest in how the product gets built, get the team using these tools every day, and often rethink what the product even is, from its interface to its core capabilities. We remember how desktop-to-cloud went. The companies that kept their heads in the sand for the first few years spent the next ten paying for it. The encouraging difference this time is how much earlier, and how much more seriously, the industry is taking GenAI than it took SaaS.
This is the work we like best: partnering with mission-critical software companies that fit this profile and doing the work alongside the teams. This page has been mostly about product strategy, but that's not where the opportunity stops. We see just as much of it in how you sell, how you work with customers, and how you run the company day to day, from finance to HR. Where this gets specific (which companies, which workflows, where the real leverage is) is a conversation, not a webpage. If any of this matches what you're seeing, we'd like to compare notes.
What our team is thinking about.
Should you cap your team's AI spending?
Some companies now reward employees for burning tokens; others slap on hard monthly caps after blowing through their forecasts. We do neither: fund thoughtful use, refuse to reward raw consumption, and leave room for the people who find real leverage.
Ben Feldman Investors who operate, or investors who are 'operationally inclined'?
Nearly every firm below $100M of ARR brands itself "operationally focused" — gray-haired advisors and an ops team a phone call away. It works, but it's imperfect. Everyone on our team has been both an investor and a CEO, which is why we send Functional Operating Executives into every company we back.
Garrett Levey Saying 'no' is hard. When is it the right answer?
Our largest customer offered us over a million dollars a year to build a new module. We ran the diligence, found the category had defeated everyone who tried it, and said no while keeping them as our biggest advocate.
Bill Kloza Want to learn more? Start a conversation.
We'd love to meet, trade notes on what we're seeing as the software world goes through a generational evolution, and share more about our philosophy for building great software companies.