What AI is already costing you

The first question in most AI conversations is what to invest. A more useful first question is what is already being spent, because the answer is rarely zero and it is almost never visible in one place.

1. What is bundled into licences you already buy

Major vendors have moved AI capability into base tiers. You are paying for it whether or not anyone uses it, and a proportion of your staff now have AI in front of them without a decision being taken.

Two implications. The capability may already cover a use case somebody is proposing to buy separately. And the policy question — what may be put into these tools — has been answered by default unless you answer it deliberately.

2. Consumption you have not been tracking

Agent and assistant usage billed by the unit rather than the seat. It can be started by anyone who can build a flow, it grows with adoption, and it does not trigger a procurement conversation.

The first month is when habits form and nobody is watching. By the time the number is large enough to notice, several teams depend on the thing generating it.

3. Tools bought locally

Writing assistants, meeting transcription, research tools, design tools. Each below the threshold that would route it through procurement, each on a departmental card, each with terms nobody has read about what happens to the content put into it.

Expense claims and identity sign-in logs will find most of this in an afternoon.

4. Staff time

The largest and least measurable. People are using these tools on work, learning by trial, redoing outputs that were not good enough, and sharing techniques informally. Some of this is genuinely productive. Some is a slower path to the same result, done with more enthusiasm.

You cannot price this precisely and it is worth asking about rather than ignoring, because it tells you where the demand actually is.

Why finding these matters before you fund anything

It changes what you need to buy. Often a proposed purchase duplicates something already inside a licence you hold.

It shows where the real demand is. The tools people bought themselves are evidence of a problem worth solving, and they are a better guide than a workshop.

It sizes the exposure. Company information is already going into tools with varying terms. That is a present fact, not a future risk.

It makes the business case honest. A case comparing a new investment against zero is comparing against the wrong number.

What to do with the total

Publish it, once, to the people who will decide what happens next. A single number for what AI already costs across those four categories reframes the conversation from whether to start to what to do about what is already happening — which is a more useful place to begin.