Sam Gerstenzang argues AI expands SaaS spend while splitting value into small opinionated point tools and large full-service “does the work” products—squeezing classic mid-market systems of record from both sides.

Software Goldilocks: the market for software is expanding but is the System of Record dead?

In the world of vertical software, the system of record was the gold standard for years. It was the place with all the core workflows and data for a business, a complete software solution.

AI has upended this. But I don't think SaaS is dead as many fear: I think more money will be spent on SaaS than ever before.

Yes, the big frontier labs are busy 'licking the cookie' claiming they'll eat everything. But horizontal platforms have never dominated the vertical businesses in the history of any software platform. There's always room to go deeper, more specialized, more opinionated in vertical software, and software is a relatively small expense for most businesses.

We are, however, going to see a shift in how software is priced and where the value is captured. I think of software in three buckets:

Small Software: point solutions. Often without a moat, covers a single set of workflows.
Medium Software: the system of record. This used to be the 'complete' solution. Think: Toast, ServiceTitan, etc.
Large Software: a new category where the software 'does the work', it competes not with the software budget, but with the cost of employment. Think: Moxie, Fora, etc.

What's new

First, the 'large software' bucket didn't used to exist. At Moxie, we've had a mixed software and services since 2022 and, as a result, captured an unusually high take rate. With AI, there are now many more full-service software businesses to be built, where software can do what only human employees could do before.

The second thing that changed is it's now possible to build 'small' software and we are about to see a massive proliferation of point solutions that might not have been viable businesses before. With lower moats and smaller TAMs, they might simply have been too expensive to build and market before.

For example, imagine a tool that collates, analyzes and predicts death data across the U.S. The TAM is tiny: maybe you can get 20% of the 20,000 funeral homes to pay $300/year—$1.2 million in revenue. Before AI, you might have needed an engineer, a designer, some data analysts. Now this all might cost $200k and you're taking home $1m.

There are going to be thousands of these, and they're going to chip away at the value of the System of Record by creating a graduation path from the spreadsheet-defined workflow that doesn't jump straight to site-wide licenses.

If we look at the value capture of software, most of the growth will go to those two categories: small, opinionated bootstrapped tools on one end, and full service venture funded businesses on the other.

It's nearly anathema to consider, but I wonder if we'll see a drop in the price of software for core systems of records. Systems of Record are going to be fighting on all sides: fighting the cheaper, less featured small tools on one end, and full service systems that give away the core software for free on the other.

Originally published on Sam Gerstenzang's Substack.