The invoice is the small number. The real cost of vendor lock is three numbers underneath it, and they compound. This post does the arithmetic in public on a 3-year scenario.
The scenario, held constant
A mid-sized team is offered a 36-month AI services contract: $11,000 a month, 7% annual escalator, standard "platform" terms. The competing engagement ships artifacts the client owns at the end of every quarter. Same problem, same team, same calendar.
We score both on three buckets a CFO can defend in a board room: direct, indirect, optionality (Class C — modeled from disclosed contract clauses, not a worst case).
Direct cost — the invoice line
Vendor-lock. $11,000 × 12 = $132,000 in year one. The 7% escalator makes year two $141,240 and year three $151,127. Three-year direct cost: $424,367.
Exit-ready. A scoped engagement that ships the same workflows as owned artifacts lands in the $260,000–$310,000 range over three years, depending on whether the team renews quarterly scopes or runs the runbook themselves after year one. Midpoint: $285,000 (Class C — drawn from our quoted engagements, observable in signed orders).
Direct delta on the invoice alone: about $139,000 for exit-ready. That is not the headline number. The headline numbers are the next two.
Indirect cost — the work the invoice hides
Lock-in is paid for in staff hours that never appear on the vendor's bill. Three lines, all conservative.
Integration glue. SSO bridges, Slack hooks, CRM connectors, the eval harness against the platform's API. A two-engineer team spending one day a week on vendor-shaped glue is 100 days a year. At a fully-loaded $1,200 per engineer-day, that is $240,000 over three years of work that produces nothing the team owns (Class C — staffing pattern observable in three of our last five intakes).
Vendor-shaped workflows. Prompts, fine-tunes, dashboards, and chat histories live inside the platform. Migration off, when it happens, is a re-implementation paid once at termination.
Audit and renewal overhead. Procurement, legal, and security review at every renewal. Three cycles over three years: about $25,000 in loaded internal cost.
Exit-ready indirect: the same engineers tend a runbook the team owns, integration time cut by two-thirds, no migration project, one procurement cycle. Net near $80,000 over three years.
Indirect-cost delta: about $185,000 in favor of exit-ready — bigger than the direct delta, and almost never on the spreadsheet at signing.
Optionality cost — the option the team gives up
This is the hardest to put on a CFO's worksheet, and the one that matters most.
A team locked to a platform cannot accept a better offer when it appears. A model half the cost ships in month nine of year two; the team cannot adopt it without breaking the contract. A new hire is excellent at a different stack; the team cannot deploy the skill. A regulatory change requires data residency the platform does not support; the team cannot move.
We score optionality cost as the expected value of the switch the team was not free to make. At 2026's rate of substrate change, model one switch missed per year, at $25,000–$40,000 saved per switch. Midpoint over three years: $97,500.
Exit-ready: artifacts ship quarterly, so substrate switches happen at quarter boundaries at the cost of a one-week migration. Optionality cost: about $15,000 for friction.
Optionality-cost delta: about $82,500 in favor of exit-ready (Class C — modeled, falsifier below).
The total, on one line
Three-year direct + indirect + optionality, vendor lock: roughly $786,867. Three-year direct + indirect + optionality, exit-ready: roughly $380,000.
The delta is about $406,000. The invoice told you about $139,000 of it. The other two-thirds was in the work the invoice did not show.
Falsifier. If a real audit of three signed lock-in contracts at this scale shows indirect + optionality cost under $100,000 combined across the term, this model is wrong and we will retract the post in writing on the ledger.
Why the market still picks the trap
The invoice is one number; the real cost is three numbers that show up later. The hiring market is shaped by the same pressure — a team built on one vendor's UI is paid for the UI, not the underlying skill. Jay Kumar Chimata, who runs a 22,000-user AI talent platform, did an interview on what 2026 employers actually hire for versus what candidates study: Meet Jay Kumar Chimata: JobFirst.ai and the Real AI Job Market (Class E — primary-source interview with a hiring-marketplace operator).
Our read, in our words: hiring managers pay for the ability to ship and operate an owned artifact end-to-end. A team trapped inside a vendor's UI produces the discounted version of that skill on the company's dime.
What to ask before you sign
Three questions cover most of it.
- Show the three-year total with the escalator, including migration cost at termination.
- List the staff time to keep integrations running and renewals processed.
- If a better substrate ships in month 18, what does it cost to move?
Concrete answers are fine. Evasive answers are the trap. The intake call is on /workshop — same call whether we end up scoping an engagement or telling you the contract on your desk is fine.