Use this whenSomeone wants a new tool, whether built or bought, or a software renewal is coming up.
Why it exists
With AI, someone can build a working tool in an afternoon, and the demo makes buying software look like a waste of money. But the afternoon is the cheapest part. Comparing a quick build with a vendor’s license price ignores most of what it costs to keep a tool running and get people to use it. And both options skip a question that should come first: does this work need a tool at all?
The framework
Build
What does it take to make it work?
The vibe-coding time, plus the integrations, security review, and testing needed before anyone else relies on it. Usually the smallest number on the page.
Maintenance
Who keeps it working next year?
Bugs, model and API changes, data changes, and the person who gets the message when it breaks. If nobody owns it, it stops working quietly.
Rollout
What does it take to put it in front of everyone who needs it?
Access, permissions, documentation, support, and connecting it to the systems people already use. This cost applies to bought software too.
Change management
What will it take for people to actually work differently?
Training, new habits, retiring the old way of doing things, and leaders who reinforce the change. Often the largest cost, whether you build or buy.
How to use it
- Start with the third option. Can the work be removed, or can an agent be given access to the data and actions instead of adding another interface? Price that first.
- If the work has to stay, estimate all four costs for building, and the same four for buying. A vendor still needs rollout and change management.
- Name who will own maintenance. If you can’t, the build isn’t ready to be shared.
- Compare over two or three years, not the first month.
- Decide, and write down what would make you revisit it.
Revisit the decision when
- The vendor ships the feature you built, or stops shipping the one you bought.
- The owner leaves and nobody picks it up.
- Usage drops, which usually means the work moved somewhere else.
- The underlying model or API changes in a way that breaks it.
- The cost assumptions turn out to have been wrong.
In practice
Part of our results at Rivian came from simplifying the technology stack and reducing tooling and licensing spend. Some of that came from replacing software with tools teams built and owned themselves. Just as often, the honest answer was that the build only looked cheaper because maintenance and rollout weren’t being counted.
Where it breaks down
Estimates for maintenance and change management are rough, and that’s fine. The point is to put them on the page rather than treat them as zero. A personal tool used by one person doesn’t need this analysis at all.
Try it with AI
Help me decide whether to build, buy, or remove. Start by asking whether the work could be removed, or whether an agent with access to the data and actions could replace a new interface. If the work has to stay, estimate four costs over three years for both building and buying: 1. Build: making it work, including integrations, security, and testing (or, for buying, license and setup). 2. Maintenance: who keeps it working, and what that takes. 3. Rollout: access, support, documentation, and connecting it to existing systems. 4. Change management: what it will take for people to actually work differently. Ask me one question at a time. Then compare the options, flag any cost I've treated as zero, and tell me what would make us revisit the decision.
Build, buy, or remove cost sheet