AI as a line item — cost recovery for marketing agencies.

Of every industry making this category error, agencies are losing the most money. Agencies are AI-heavy users by nature — copywriting drafts, image generation, video tools, social content, market research, analytics summaries, brief drafting, deck production, A/B variant generation. The AI bill is bigger than it is in almost any other professional-services category, and most of it is sitting silently in agency overhead.

What's already in your SOW

Look at a typical agency master services agreement or statement of work. It already itemizes:

• Stock asset licensing — Shutterstock, Getty, Artgrid, AudioJungle, model releases.
• Third-party tool costs procured for the engagement — heat-mapping software, paid research panels, specialty SaaS seats stood up for the client's campaign.
• Production pass-throughs — print runs, premium hosting, ad-platform spend, talent fees, location rentals.
• Travel and out-of-pocket — shoots, client meetings, conferences attended on the client's behalf.

The contract framework is already there. Every one of these line items follows the same principle: direct costs incurred for the benefit of this client are billed back to this client, typically at cost or at cost-plus-markup. AI tokens used to draft a client's blog content, generate concept imagery for their campaign, or research their competitive landscape are exactly the same category of cost.

Why the agency leak is the biggest

A small law firm might spend $35 per matter per month on AI. A small construction GC might spend $250 per project. A small agency easily spends ten times what a law firm does, per client, because AI is doing more of the actual production work.

The result: agencies that aren't billing AI back are subsidizing client work out of their own margin in amounts that meaningfully affect the P&L. And it gets worse as AI adoption deepens, because the per-client AI spend keeps climbing while the agency's billing structure doesn't change.

The numbers for a small shop

Consider a small independent agency with about 8 active retainer clients in a typical month. Across content, creative, research, and analytics, AI usage averages roughly $400 per client per month:

• 8 clients × $400/client = $3,200/month
• $3,200 × 12 months = ~$38,400/year

That's nearly five times the per-firm leak of a comparable small law firm — because agencies use AI more and the AI is doing more of the deliverable production. $38,400 of margin per year, on a small shop, is the difference between hiring another junior and not. It is not a rounding error.

Pass-through, markup, or productized?

Agencies have more billing flexibility than most professional-services firms, which means there are three legitimate ways to handle AI cost recovery — each with different P&L implications:

• Pass-through at cost. Cleanest, most defensible. AI shows up as a reimbursable line on the monthly invoice at actual cost, like Shutterstock licenses.
• Pass-through with markup. Same line item, but billed at a fixed markup (commonly 15–25%) to cover the agency's overhead in managing the AI infrastructure. Standard practice with production pass-throughs already.
• Productized. AI capability is bundled into a named service tier ("Content+" or similar) and priced as a premium offering. The agency captures the upside and the client gets a predictable bill.

The right choice depends on the agency's positioning and client relationships. The wrong choice is the unspoken one — burying it in overhead.

The piece that was missing

Agencies have always tracked time to clients. The same posting discipline needs to extend to AI usage — every query tagged to the client and the campaign at the moment of use. Without that, the AI bill lands in firm overhead and the per-client P&L reports underestimate the cost of serving each account.

The accounting pattern is general — see the Cost Recovery overview for the full framework, or Interchange for the platform that makes per-client AI attribution operational.