Johnny Rohrbach, co-founder of Silverside AI, emphasized that there is no single “silver bullet” model for AI, noting that different AI applications involve varied token economics. Marketing service companies face the challenge of deciding whether to absorb AI costs themselves or pass them on to clients, with no clear consensus emerging. Digiday interviewed a range of media, creative, and agency professionals to explore current approaches. At full-service agency Merge, CTO Kyle Smith explained that AI compute costs are billed to clients on a metered, case-by-case basis. Similarly, production company Big Spaceship treats token expenses like other production costs, as CEO Taryn Crouthers noted. Silverside AI employs a subscription model with “seat” pricing akin to Salesforce, according to Rohrbach. Conversely, at full-service agency RPA, SVP Lisa Herdman said they absorb token costs due to ongoing testing phases and uncertain client benefits, unwilling to charge clients for unproven services. Anomaly's global chief AI officer Chris Neff shared concerns that passing on AI fees might appear as a “money grab. ” Brandtech’s AI platform Pencil offers clients “generation credits” for AI outputs, with pricing tied to volume; CEO Will Hanschell highlighted that larger client commitments enable better deals with AI providers, lowering costs, and clients can top up tokens as needed. Pencil’s tiered pricing ensures fairness between large and small clients, with features like legal indemnification reserved for bigger users. Hanschell emphasized aligning incentives by charging based on usage while keeping pricing straightforward. Media agencies like Horizon Media approach the matter differently.

Horizon launched its gen AI platform Blu in December 2025, and EVP Krish Kuruppath noted token expenses mainly arise from onboarding and development, with costs becoming significant only at large user scales. They apply a “nominal” fee intended for cost recovery rather than profit, as Chief Product Officer Domenic Venuto explained, aiming to avoid client barriers while protecting margins. Kepler, another media agency, has not pursued bulk AI token deals; VP Peter Rice stated they embed AI capabilities within their retainer-based client work without passing on token costs, focusing instead on delivering impactful results rather than tracking token consumption. Similarly, Lerma/ agency secures bulk token agreements to include AI costs transparently within client pricing at cost, as Chief Data Officer Josh Archer described. If tokens are underestimated, surplus AI-generated assets are provided to clients, turning it into an opportunity to add value. The practice of agencies securing discounted bulk token purchases raises parallels to controversial principal media buying, where markups on media spend are common. Consultant James Londal argued that chasing token price arbitrage is less significant than labor savings AI enables by automating reporting and workflows. Hanschell agreed, emphasizing competition should be merit-based rather than purely on price. Jess Lewis, CTO at Crossmedia, warned that focusing on passing through token costs risks agencies prioritizing margins over delivering appropriate value and architecture suited for clients. Token pricing is expected to become part of agency negotiations and audits in the future; Ebiquity CEO Ruben Schreurs predicted token usage will be scrutinized similarly to media spend during contract reviews. Agency leaders recognize client CFOs will tightly control pass-through costs, so they prefer to highlight AI’s business benefits rather than reducing usage to cost savings alone. Schreurs advised focusing less on AI inputs like tokens and more on how AI drives incremental business outcomes for clients.