The latest forecast
At its Orlando symposium on 21 October, Gartner predicted that by 2029, 60% of organisations deploying AI would establish a dedicated function to map total AI cost to value or profit. The firm also forecast that by 2030, 80% of organisations with public-facing AI would have experienced attacks designed to exhaust their spending budgets. Both figures are forecasts, not measurements of current adoption or attacks. [1]
Context: usage is not the outcome
An organisation can count prompts, accounts and model calls without knowing whether a service improved. A more useful assessment links a specific task to its cost, the quality of the output and the work required to check it. This makes evaluation harder than a simple adoption chart, but also makes it more relevant to a decision about whether the system should continue.
Why budgets become a security question
A public service that charges for each model request needs rules for who can use it and how much work each request can trigger. In practical terms, a team should ask how it would notice abnormal use and how it could stop spending without disabling an essential service. These are assessment questions arising from the forecast; they do not establish that a named company has suffered such an attack.
What to watch next
The test for buyers is whether suppliers and internal teams can explain cost controls in terms that finance and security staff both understand. A demonstration of a capable assistant leaves that operational question unanswered. Gartner’s projections can help frame planning, but their dates and percentages should remain visible as assumptions about the future. They should not be presented as inevitable outcomes or as evidence that every AI investment is failing.
