Six consumer AI subscriptions at roughly $20 to $30 each comes to about $125 per person per month. That number is easy to compute, easy to put in a deck, and is not the reason fragmented AI access is expensive.
Procurement pays for it first
Every provider is a vendor. Every vendor needs a security review, a DPA, a subprocessor assessment, an owner in the vendor register, and a renewal date somebody watches. At the mid-market companies we talked to, onboarding one software vendor properly consumes somewhere between six and twenty hours across security, legal and finance.
Six providers is therefore not six times $20. It is six security reviews, six DPAs, and six renewals. Teams respond to this the way you would expect: they stop doing it properly. Which brings us to the second cost.
Policy that exists only on paper
When access is fragmented, the AI policy is unenforceable by construction. You cannot tell which employees hold which subscriptions, because most were bought on personal cards and expensed. You cannot tell what data went where, because each provider's console shows only its own traffic and most consumer tiers show nothing useful at all.
Nearly every security lead we spoke to described the same shape: a written policy about what may be pasted into an AI tool, and no mechanism whatsoever to observe compliance. The policy is a document, not a control.
Spend nobody can see
Expensed consumer subscriptions do not show up as a line item called "AI". They show up as seven small recurring charges across a dozen cards, categorised as software or, frequently, as nothing at all. Two finance teams told us they discovered their real AI spend was between three and five times what their budget line said, and found out during an unrelated audit.
The problem is not that the spend is large. It is that it is invisible, so nobody can decide whether it is worth it.
The switching tax
There is also a smaller, more human cost. When a person holds four subscriptions, they do not fluidly pick the best model per task. They develop a habit around one, use it for everything, and occasionally remember the others exist. The value of holding four subscriptions is largely unrealised because moving between them means re-pasting context into a different tab.
What consolidation actually changes
Putting access behind one login does not make the models better. What it changes is the surrounding accounting:
- One vendor to review, one DPA, one renewal, one entry in the register.
- One audit log covering every provider, which turns the policy into something observable.
- One invoice, which means the spend has a number and the number has an owner.
- One context, which means switching models is a click rather than a copy-paste, so people actually do it.
The seat price saving is real and it is the least interesting part.