Why I'm Writing About Build or Buy

Two years ago, I looked at my company’s software stack and counted 305 applications.

We were paying for all of them. We were using a fraction of them. And every quarter, someone proposed adding another one.

This isn’t unique. Organizations in Zylo’s 2026 SaaS Management Index averaged 305 applications, with 54% license utilization. Those are averages from Zylo’s customer dataset. Unused seats deserve a look, but their share is not the same as the share of spending we can recover.

I’m not writing this from the outside. I’m CMTO at a mid-market manufacturer. I run both marketing and technology. Every week, I make decisions about what to build in-house, what to buy from vendors, and what to cut entirely.

What changed

In 2023, I tried building a custom internal tool with AI assistance. It failed. The models couldn’t hold enough context, the code was brittle, and I went back to buying.

In 2025, I tried the same project. It worked in a weekend.

That gap, from a failed attempt to a useful first version, is the story of this moment. The five reasons organizations have always bought software (skill, cost, speed, maintenance, expertise) are eroding. Not all of them. Not for everything. But enough to change the calculus.

The framework

I needed a better way to make these decisions than gut feel or vendor demos. So I built one.

The B4 Framework scores software across four quadrants: Build, Buy, Bridge, or Beware. Each quadrant has clear criteria. Each tool in your stack can be scored against five dimensions. The output is a starting recommendation to test against your needs and operating capacity.

I’m writing a book about it. I’m publishing a directory mapping 1,600+ software categories to these quadrants. And every two weeks, I’m sending out The Build Report with fresh analysis.

This is where all of that lives.

If you’re a mid-market leader staring at a bloated stack and wondering what AI actually changes for your business, subscribe to The Build Report and let’s figure it out together.

Updated September 19, 2026: Clarified the benchmark population and the distinction between unused licenses and recoverable spending.