Build vs Buy Software in 2026: The Math Changed and Most Companies Haven't Redone It

I spent thirteen years at Comcast Labs watching custom builds turn into maintenance bills. The first version was never the problem. The tenth was. Somebody had to keep it running, patch it, and change it every time the business changed, and renting made that somebody else's problem. For most of my career that was the right trade.

I do not give that answer anymore. Here is why, and how to tell if it applies to you.

The pattern we keep seeing

Buyers are walking into first calls asking to own the thing. Not license it. Own it.

AI did not make custom software free. It made the part that used to kill you, the maintenance and the changes, cheap enough that ownership is back on the table. Meanwhile renting got worse. Seat pricing punishes you for growing. The roadmap goes wherever the vendor's biggest customer wants it to go. Both sides moved at once and most companies are still doing math from five years ago.

That is why the question keeps coming up. Buyers did not get braver. The math moved and they noticed before the vendors did.

Try this tomorrow: the five-minute renewal test

Pull up your next renewal. Three questions.

What is the job? Not the features. One sentence. "Takes customer messages and turns them into tickets." If you cannot say it in one sentence, renew it. The job is too big to rebuild.

What would it take to walk away? Can you get your data out, and does anything else break if this tool disappears? If leaving is easy, you have leverage either way. If leaving is hard, that is why the price keeps going up.

When did we last question this line? If the answer is never, the math was done under conditions that do not exist anymore.

One-sentence job, easy to leave, never questioned. That is the one worth pricing an alternative for. Everything else, renew and move on. You are not looking for a reason to cancel everything. You are looking for the one tool where the math flipped and nobody noticed.

What I would renew every time

Since I am telling people to build, here is where I will not.

• Anything with a compliance stamp. Payroll, accounting, tax, anything an auditor expects to see a known name on. The vendor's liability is part of what you are paying for.
• Anything where the network is the product. Payment processors, marketplaces, scheduling tools your customers already have on their phones. You cannot build the other side of the network.
• Anything your whole industry runs on. If every peer, partner and new hire already knows the tool, the switching cost is not the software. It is the people.

If your resented renewal is on this list, resent it and renew it. The build case is for the narrow tools around these.

"We own it" means three different things

There is the code, the logic that makes it yours. There is the data, everything it learns from and everything it produces. And there is the model, the AI underneath doing the reading and writing.

You can own the first two and rent the third. Most companies should. The model keeps getting cheaper and better every few months, and renting it means you can swap it when something better ships. Owning the code and the data is what makes that swap yours to make instead of a vendor's. That is the real reason the math moved.

We support ownership models and the details get settled at scoping. The own-cloud deployment guide shows what the three layers look like in an actual architecture, and the integration-cost guide shows what a build like this costs against a subscription at your real volume. Know the three layers and you are a sharper buyer no matter who you build with.

The short version

If a renewal is coming and you cannot say what the tool does in one sentence, renew it. If you can, and you resent the bill, it is worth a conversation. That is what I do all day.

Frequently asked questions

Has AI changed the build vs buy decision for software?

Yes, in a specific way. It made ongoing maintenance and changes, the part that historically made custom software expensive, cheap enough that owning narrow tools is viable again. At the same time, subscription pricing and vendor roadmaps got worse for buyers.

What software should a business still buy rather than build?

Anything with a compliance stamp (payroll, accounting, tax), anything where the network is the product (payments, marketplaces, consumer scheduling), and anything the whole industry runs on. The build case is for the narrow tools around those.

What does owning your software actually mean?

Three layers: the code, the data, and the model. Own the code and the data, rent the model. The model keeps improving and getting cheaper, and owning the other two layers is what lets you swap it on your terms.

How do I decide which subscription to replace?

Run the five-minute test on your next renewal: can you state the job in one sentence, how hard is it to leave, and when did you last question the line? One-sentence job, easy to leave, never questioned is the one to price an alternative for.

First published in The Humming Agent Brief, Shawn's twice-monthly newsletter for business owners. Subscribe there to get the next issue first.