'Buy, don't build' was good advice. It needs a footnote now.

For about a decade, the safe answer to any software question was: buy the SaaS. Building was expensive, staffing was hard, and the vendor would keep improving the product on their dime. If the tool wasn't your core differentiator, you rented it.

That advice wasn't wrong. It's just aging fast.

What changed under the hood

Retool's 2026 Build vs. Buy report surveyed engineering and IT leaders and found a clear pattern: AI-assisted development is meaningfully lowering the cost and time to build internal tools, and teams are replacing SaaS with in-house builds more often than they were two years ago. The report also notes the return of shadow IT, this time in a new form. It's not sales ops buying an unapproved subscription. It's a small team spinning up a working internal app in a week because they finally can.

Bain's 2025 technology report frames the same shift from the incumbent's side. In Will Agentic AI Disrupt SaaS?, the analysts argue that agentic AI is starting to unbundle workflows that used to be locked inside systems of record. The moat wasn't the database. It was the fact that rebuilding the surface around it was too expensive to bother. That's the assumption AI is chipping at.

Janus Henderson's investor-side analysis reads the tape the same way: SaaS multiples are being re-rated, margins are under pressure, and the market is starting to separate vendors with real defensibility from ones that were selling a UI wrapped around a workflow.

The footnote

So here's the update to the old rule.

Buy, don't build, unless the build is small, boring, and specific to how you actually work. That last clause used to be rare. Now it's most of the interesting cases.

A custom quoting tool that matches your pricing logic. A dashboard that pulls three systems together the way your ops lead actually thinks about the business. An intake form that writes to your CRM, your accounting tool, and Slack in one motion. Two years ago those were $80k projects. Now they're smaller, and they don't come with a per-seat bill that grows every time you hire.

That doesn't mean rip out everything. Your accounting system, your email, your payments processor. Keep buying those. Nobody's building a better Stripe on the weekend.

What's worth reconsidering are the middle-tier tools. The $30k-a-year SaaS that does 40% of what you need and forces the other 60% into spreadsheets and Zapier. That's where the math has moved.

The implication

The question at renewal time used to be: is there a cheaper vendor? The question now is: would we rather own this?

For a lot of tools, the honest answer is starting to be yes.