AI spend is growing fast. Are organisations ready to manage the complexity?

By Geoffrey Boutin, Matthew Goodwin East, Alex Gooch, Brendan McGonigle, Maxim Sankey

Organisations are under pressure to scale AI quickly, but the spend can be much harder to predict. As adoption spreads across employees, developers, business processes, and enterprise software, Finance and technology leaders need to enable experimentation while gaining a much clearer view of how much AI is costing, what is driving the overall spend, and where it is heading. 

Many organisations do not yet have that visibility. AI spend can rise rapidly with consumption, vary significantly between users and use cases, and depend on the tools and models selected. At the same time, new AI costs are increasingly appearing within products and services organisations already buy, making the overall picture harder to see. 

Getting ahead of this challenge starts with understanding the different sources of AI spend and the factors driving them. With a clearer picture of demand, organisations can build more realistic forecasts, identify where intervention is needed, and make better-informed commercial decisions.

Unlike traditional software, where licences and user numbers often provide a relatively stable basis for budgeting, AI spend can be ‘uncapped’ and continue to change significantly after the initial purchase. 

Depending on the service, organisations may pay by seat, credits, token consumption, or some combination of these. Usage can also vary dramatically. A relatively small number of employees or teams can consume disproportionately more than others, while developers may have very different requirements from occasional users of a general-purpose assistant. Premium (or frontier) models and features can increase consumption further. 

Organisations may therefore set budgets and make supplier commitments before they have enough usage history to understand what normal demand looks like. As adoption develops, actual consumption can diverge and often significantly exceed these initial assumptions. 

The questions then quickly become: Why is spend changing? Who is driving it? What should we budget for next? Answering those questions requires organisations to look beyond the total AI bill and understand the different sources of demand behind it.