Should You Build or Buy an AI Tool? A Simple Test
By Adam, AI Engineer · 31 October 2026 · 4 min read
Buy when the job is standard. Build when the workflow is what makes your company different, or when the number of users makes subscriptions more expensive than owning the software. We build custom software for a living, so we have a reason to favour building. Here is a test you can check for yourself.
Start with the risk of failure
Gartner predicted in 2025 that more than 40% of agentic AI projects will be cancelled by the end of 2027, because of rising costs, unclear business value and weak risk controls. It is a forecast, not a measurement, but those three causes apply whether you buy or build. Name the business result you expect before you compare anything.
Five questions that decide it
- Is this workflow what makes your company different?
- Does every ready-made product force a big compromise?
- Do you need deep connections to your own systems?
- Do data or compliance rules limit what you can hand to a vendor?
- Can you own, fund and maintain the software for years?
Three or more yes answers make building worth pricing. With fewer, start by buying. The last question matters most, because software that nobody owns slowly stops working.
Then compare total cost
These are example numbers, not real quotes, to show how the sums work over three years.
| Over 3 years | 25 users | 100 users |
|---|---|---|
| Buy: $40 per user per month, $2,000 setup | $38,000 | $146,000 |
| Build: $30,000, $400 a month running, 20% yearly maintenance | $62,400 | $62,400 |
With 25 users, buying is clearly cheaper. With 100 users, the answer reverses, assuming the running cost stays the same. In practice running costs grow with use, so check that too. You can test your own numbers in the build vs buy calculator.
Costs people forget
- When buying: price rises, usage limits, paying for add-ons, moving your data out later, and changing how your team works to fit the product.
- When building: security updates, AI usage bills, staff time for testing, and keeping knowledge in the team when people leave.
The safest route
For most companies, the safest order is to buy first, learn what you really need, then build only the part that a product cannot do. A one-month trial with real work teaches you more than a long requirements document, and if you do build later, you will know exactly what to ask for.