The Year AI Budgets Get Audited

AI spending has now been running long enough at most large organisations to be examined rather than assumed. Budget holders who received approval on potential are being asked to demonstrate outcome.

A significant number cannot, and the reason is usually that nobody agreed what success would look like before the money was committed.

Usage was measured; outcomes were not

The most common reporting we see counts licences deployed, users active, queries processed. None of these is a business result.

A tool can be used enthusiastically by a large number of people and change nothing measurable about cost, cycle time or quality. Where usage is the only available metric, the honest position is that the effect is unknown.

Pilots that never ended

A recurring pattern is a portfolio of initiatives, each individually modest, none formally concluded. They continue because stopping requires a decision and continuing does not.

The remedy is a hard review point. Every initiative gets an explicit decision to scale, stop or continue with a defined end date. Most portfolios shrink considerably under this discipline, and the freed budget usually funds the two or three that were genuinely working.

The savings landed somewhere else

A frequent complication is that efficiency gains are real but never materialise financially. A process becomes faster and the team stays the same size, so cost is unchanged.

That is a legitimate outcome if the capacity was redeployed to something valuable, and a failure if it was simply absorbed. The distinction has to be made explicitly, because it will not appear in any report by itself.

Vendor costs have not been stable

Pricing across the sector has moved repeatedly, in both directions and with changing structures. Organisations that budgeted on early pricing have found their assumptions unreliable.

This is a good moment to establish what you are actually committed to, what a substantial price change would do to your business case, and how difficult substitution would be.

What a defensible position looks like

For each initiative you should be able to state the business metric it was intended to move, its value before and after, the fully loaded cost, and whether the benefit was banked or absorbed.

Where you cannot, that is not necessarily a reason to stop. It is a reason to establish a baseline now and set a review date. The organisations coming out of this well are not the ones that spent least. They are the ones that can tell you what happened.