The vendor quote is usually the smallest number you'll pay.
AI programs rarely blow their budget on the thing they bought. They blow it on the four costs that never made it onto the quote — integration, maintenance, the people to run it, and the price of getting back out. Here's how to see all five before you commit.
By Joe Perone, Senior Principal Last updated
What we talked through
Total cost of ownership for an AI program spreads across five buckets. The one on the invoice is almost never the biggest:
- Build or buy. The license, subscription, or development cost. Real, but usually the smallest line and the easiest to see.
- Integration. Wiring it into your actual data and systems. Routinely underestimated because the demo runs on clean, isolated data and your world isn't clean or isolated.
- Maintenance. Models drift, data changes, edge cases pile up. Someone owns keeping it accurate — every year, not just at launch.
- Talent. The people to operate, monitor, and improve it. Often the largest line and the one least likely to be in the original budget.
- Exit cost. What it takes to switch vendors or shut it down — data extraction, retraining, migration. Nobody prices it, and it's exactly what quietly locks you in.
For smaller organizations, the biggest overrun risk usually isn't compute at all — it's scope creep ("while we're in there, can it also do X?") and subscription sprawl, where several teams independently pay for overlapping AI tools nobody tracks centrally. It's worth auditing; most orgs that do are surprised. This is exactly why we run fixed-scope, fixed-price engagements: open-ended "AI transformation" budgets are nearly impossible to forecast, and a firm boundary is the cheapest cost control there is.
Interactive · ~60 seconds
Total Cost of AI Ownership calculator
Enter a rough first-year estimate for each of the five categories. The chart shows the relative weight of each — usually the surprise is how small "build/buy" looks next to the rest.
Where the money goes
First-year total
$135,000
Directional only — this is a thinking tool, not a quote. Your real numbers depend on scope, data, and team.
Take it with you
The five costs of AI ownership
| Cost category | What hides here |
|---|---|
| Build / Buy | The visible line. Compare against the four below before you treat it as "the cost." |
| Integration | Data plumbing, auth, edge cases, and the gap between the demo environment and yours. |
| Maintenance | Drift, retraining, monitoring, and support — every year, not just at launch. |
| Talent | The person or fraction of a person who actually owns it. Frequently the biggest line. |
| Exit | Data extraction, migration, retraining elsewhere — the cost that quietly creates lock-in. |
Want the full worksheet?
Get the TCO worksheet with prompts for every line item
A printable version of this calculator with the questions to ask a vendor for each of the five costs — so nothing hides in the quote. We'll email it now.
No spam, no drip sequence. We use your email to send what you asked for and, at most, one follow-up. Prefer to talk? Book a 30-minute strategy call — no deck, no junior gatekeeper.