The Anatomy Of Agency AI Bills: Tokens, Seats And Who Foots It

Agency AI costs are climbing and metered model usage is becoming the norm, forcing holding companies to take a hard look at who pays the token bill. According to Ramp research, about 31% of companies now spend more than $10,000 a month on AI. Named voices across the piece describe three models: Markacy's Tucker Matheson treats AI as variable cost ("I still need people who can get on a phone with a CMO and be impressive"), Crispin's Freddy Dabaghi locks most AI deals into flat fixed fees with annual renewals, and Dept's Jonathan Whiteside runs a monthly AI Tool Committee to approve, test or drop tools. At the Goldman Sachs conference, Omnicom CFO Phil Angelastro said the holdco is building a model to manage and get reimbursed for token costs, "Ultimately, we're going to move towards a more outcome-driven model." At least one agency is experimenting with billing AI token costs to clients as materials, separate from human labour.

The crux

AI has moved from experimentation to operational cost, and the industry's answer to unpredictable token bills is a push toward outcome-based pricing and "materials" billing, which will rewire how agencies charge and, downstream, how they value publisher inventory.

Sell-side read

If agencies shift to outcome-based remuneration, publishers get pulled into the same logic: media owners will be asked to prove business outcomes, not just deliver impressions, and those who can tie inventory to performance will command the premium while pure reach is repriced down.

Watch

Whether 2027 planning cycles formalise token-cost reimbursement as a standard line item, which would mark AI cost as permanent agency infrastructure rather than a discretionary experiment.