Correction, 2026-10-02: we re-checked the Forrester and 4A's figures on Forrester's own pages; they stand. Several statements about what agencies do were not backed by any data: that "the agencies that solved this" replaced the hourly model, that "some agencies" raised hourly rates dramatically, that selling AI as a discount is "the most common error", that value pricing is "the oldest pricing model in consulting", and that some agencies "often find" it is the only model that fits. They are now advice or removed. The recap reel in the cost table was wrong: twelve 5-second clips make a 60-second reel, not a 90-second one, so the row now shows a 60-second reel at 765 credits (it said 846 for 90 seconds). We also added that Kling v3 needs a start image, that a twelve-clip reel with music is assembled on the Video Editor timeline (Montage joins at most 10 clips and its output has no sound), Forrester's proposed replacement for FTE-based fees, and a pointer to the ad disclosure rules.
Correction, 2026-10-01: the first version of this page said agency production costs fell "roughly 70%" (50% to 85% by task) "across agencies we've observed", that a 30-variant campaign fell from $2,000 to $15, and gave typical rates ($800 an hour, $150 to $350 per deliverable set, $2,500 to $5,000 hourly caps) and client reactions as observed facts. We have no data behind any of these, so they are removed or now shown as illustrations. Generation costs are now computed from 8frame's credit prices (30 five-second clips are 2,580 credits, not $15), the "16-model platform" line is gone, and the industry figures come from Forrester and the 4A's.
Agency AI pricing is under pressure because generative AI changes how long creative work takes, not just how it is made. Forrester calls the result an "AI cost center crisis": AI helps agencies deliver more, "But they are not being compensated for the value creation." When a 30-variant social set costs a few thousand credits of generation and an afternoon of review, billing by the hour pays less for the same output, and an agency that doesn't update its pricing structure risks leaving margin on the table or cutting its rates in ways it can't walk back. Most haven't repriced yet: in Forrester and the 4A's report The State Of AI Inside US Marketing Agencies, 2026, nine in 10 US agencies use generative AI, but 61% still classify AI as a "cost of business" with limited direct monetization. This article covers four pricing models that can replace hourly billing, what the generation line actually costs, and what agencies should protect.
TL;DR
- Hourly billing is under pressure because AI cuts the hours a deliverable takes while output volume goes up. Billing on time penalizes that efficiency.
- Four models can replace it: outcome-based, retainer-plus-AI-volume, hybrid hourly, and value pricing. You can run two of them in parallel, for example hybrid hourly for one client and retainer-plus-volume for another.
- Most agencies are absorbing AI, not pricing it. Forrester and the 4A's found 61% of US agencies treat AI as a "cost of business"; a year earlier Forrester reported 75% funding their generative AI capabilities without passing the cost to clients. Pricing proactively is how you keep the efficiency gain.
Why the Hourly Rate Model Stops Working
The hourly rate worked when time and output were roughly correlated. Say a junior designer spent 8 hours building 20 social variants. You billed 8 hours. The client got 20 variants. The rate made sense.
AI breaks that correlation. A creative director can now generate 30 variants on Kling v3 (from a product still) or Seedance 2.0 in a single session, review them, and select the top 12 for client delivery. The output is higher volume, and the time is a fraction of what it was. If you bill on time, you collect a session's worth of rate for what used to be a full day of work.
Raising the hourly rate to compensate is tempting, but it is not a sustainable fix. Clients do the math when a short session is billed at a rate several times what they used to pay, and it invites negotiation, skepticism, and scope creep.
The more durable fix is to change what you bill for. Forrester argues the same way: it advocates a remuneration model it calls the "human/technology equivalent" to replace the current full-time-equivalent (FTE)-based services model, with the cost of AI technologies "permanently coupled with the hourly costs of the experts who wield them," so that agency relationships focus on "value, outcomes, and growth, rather than scopes, FTEs, and billable hours."
4 Pricing Models That Can Replace Hourly Billing
1. Outcome-Based Pricing
The cleanest break from hourly billing. The agency prices based on the business outcome it's contracted to influence, not the creative work produced. A social agency might charge a monthly fee tied to a CPM or ROAS target. The creative production volume is internal to the agency; the client pays for results.
This works when the agency has enough historical data to price the risk. It needs more trust, and usually a longer client relationship, to start. But it aligns the agency's incentives with the client's actual goals.
AI makes this model easier to run. When you can generate and test 30 variants in a session instead of 2, you have more chances to find a high performer and hit the target.
2. Retainer Plus AI Volume
A fixed monthly retainer covers strategy, creative direction, client management, and a baseline output. Above that baseline, there's a separate AI volume rate per deliverable set. An illustration, not a benchmark: a monthly base fee for the retained work, plus a fixed price per additional variant package above an agreed number of pieces.
This model is honest about what the cost structure actually looks like and lets clients self-select into more volume when they want it. It's also predictable on both sides. The agency isn't surprised by a month where a client suddenly wants 80 variants for a launch; the client isn't surprised by a bill that seems disconnected from time spent.
It is a natural fit for performance marketing clients who already buy on volume metrics, because they understand the variant logic.
3. Hybrid Hourly
Not every task is AI-eligible, and not every client is ready for outcome-based pricing. Hybrid hourly is the transitional model. AI-eligible tasks (variant generation, mood board production, reference image creation) are priced on a per-deliverable basis. Strategic tasks (brief development, client strategy sessions, creative concept work, performance analysis) stay on a capped hourly basis.
Price AI-generated outputs per deliverable set, not per asset: a set of 20 social variants is one line item, not 20. Put a monthly cap on the strategic hours, sized to the client.
This model can be easier to sell to clients who've had hourly retainers for years, because it keeps the hourly rate for the work they already understand and introduces deliverable pricing only for the new AI-generated category.
4. Value-Based Pricing
The fee reflects the value the work creates for the client, not the cost of producing it. A campaign that drives six figures of attributed revenue isn't worth a small fee just because the production was efficient.
Value-based pricing requires the clearest articulation of contribution, the most confidence on the agency's part, and the most trust from the client. It's also where the ceiling is highest. For outcome-driven creative work where attribution is clear, it may be the model that best reflects what you're actually providing.
AI doesn't change value-based pricing fundamentally. It just removes the "but our costs are high so the price makes sense" fallback. You're pricing on contribution, full stop.
What the Cost-of-Goods Reduction Actually Looks Like
There is no reliable industry-wide percentage for how much AI cuts agency production cost; it depends on the task mix and on how much review and editing the work needs. Forrester's Jay Pattisall writes that some agencies "reduce production costs between 40% and 50%," which is a useful order of magnitude, not a benchmark. What you can compute exactly is the generation line. On 8frame, in credits (a credit is about $0.01 at pack rate):
| Task | Generation on 8frame | Human time to plan for |
|---|---|---|
| 30 social video variants (5 s, Kling v3 Standard with sound, each from a start image) | 30 x 86 = 2,580 credits (about $25.80) | review and selection |
| Mood board, 16 images (Seedream 5 Lite) | 16 x 6 = 96 credits (about $0.96) | curation |
| Pitch deck visuals, 14 images (Nano Banana 2 at 1K) | 14 x 11 = 154 credits (about $1.54) | curation |
| Client recap reel, 60 s (twelve silent 5 s Kling v3 Standard clips plus a 60 s ElevenLabs Music track) | 12 x 57 + 81 = 765 credits (about $7.65) | edit on the Video Editor timeline |
Kling v3 animates a start image, so each video variant needs a still: a product shot or a generated frame. Rejected takes cost the same as kept ones, so budget for re-runs. The generation line is small; the review and edit time is where the cost now sits, and it is real and non-negotiable; see the "3 mistakes" section below. Video costs more per piece than stills in credits (57 to 86 for a 5 s Kling v3 Standard clip, against 6 to 11 for the images above), and a clip takes longer to review than a still. For the recap reel, put the clips and the music track together on the Video Editor timeline: Montage joins at most 10 clips, and its output has no sound.
What to Keep for Yourself
The cost-of-goods math above only tells half the story. When you reprice, be clear about which parts of the work AI doesn't touch. Don't discount those parts to compensate for cheaper production. Lean into them.
Strategy and brief development. A model can draft a brief, but it can't own one. The creative strategy, the audience insight, the campaign concept, the brief that shapes every downstream output: that's still human work, and it's where most of the client relationship lives.
Brand voice and judgment. AI generates options. It doesn't know whether the third variant is off-brand in a way that only becomes obvious when you've worked with a client for two years. That judgment is the agency's. It's what the review pass is for. It's also what justifies the fee.
Performance analysis and iteration logic. Running 30 variants is easy. Knowing which metrics to optimize for, how to read the test results, and what to iterate on in round two requires understanding the client's business. AI accelerates the production loop; humans still drive the decision loop.
Client management and trust. The relationship is the product, as much as any deliverable. Automating client communication along with production is a fast way to make clients notice. The strategic check-ins, the presentations, the difficult conversations about what isn't working: these are where the agency's value is most visible.
3 Mistakes in Agency AI Pricing
Selling AI as a discount. The agency's production costs dropped, so it cuts the retainer rate to pass the savings to the client. It looks generous. What it actually does is reset the client's price expectations downward, signal that the prior rate was inflated, and remove the margin that was supposed to fund the agency's capabilities. Absorbing the cost is already widespread: in Forrester's 2025 research, 75% of agencies funded their generative AI capabilities themselves. The better move is to hold rates or increase them while delivering more output. You're not cheaper. You're higher-capacity.
Hiding AI use. Treating AI generation as a back-office efficiency and never disclosing it is a risk: if clients discover it, the conversation becomes adversarial, and the trust damage is hard to repair. The better framing is proactive: you now run a multi-model generation workflow that lets you deliver more variants, faster mood boards, and higher-volume testing. The creative strategy and quality review are still yours. Clients who still object after that conversation are a different kind of problem. The ads themselves may need a disclosure too: New York has required ads to disclose AI-generated synthetic performers since June 9, 2026, California will from January 1, 2027, and TikTok's ad policy asks for a label on completely AI-generated content. See do you have to disclose AI-generated ads for the conditions; it is general information, not legal advice.
Skipping margin protection. AI workflows are fast, and speed creates pressure to skip the review step. Outputs then hit clients with subtle problems: a video variant where the brand color is off, a mood board image that looks fine at thumbnail size and wrong at full-screen, a style that drifts from the client's guidelines in a way that accumulates across a month of content. Budget fixed review time for every deliverable set; 20 to 30 minutes is a sensible starting point to adjust from. That's not a cost; it's the quality floor that makes the deliverable worth billing.
FAQ
How should agencies price work when AI cuts production costs?
Price on output value and strategic contribution, not on production cost. A campaign that generates 30 tested variants and surfaces the highest-performing hook is worth more to the client than a campaign with 2 variants, regardless of how long it took to produce. If you've been billing on hours, this is the forcing function to move to deliverable-based or value-based pricing.
Do clients care that AI was used in their creative work?
Assume they do, and don't leave them to find out. Disclose proactively and present it as a capability upgrade: more variants, faster iterations, more references. Focus on what hasn't changed: the strategy, the direction, the quality review. The damaging case is a client discovering AI use they weren't told about.
What's the right retainer rate for an AI-enabled agency?
There is no universal rate, but the floor should reflect that you're now delivering more than you used to. Cutting the retainer to "pass on" AI savings resets the client's price expectations downward. Holding the rate and adding deliverables, or moving the volume onto a separate line, keeps the margin that funds your AI capability.
How many agencies are actually charging for AI?
Most still don't. In Forrester and the 4A's The State Of AI Inside US Marketing Agencies, 2026, nine in 10 agencies use generative AI and half use agentic AI, but 61% classify AI as a "cost of business" with limited direct monetization; 31% plan to monetize agentic AI within the next 24 months. In June 2025 Forrester reported that 75% of agencies funded their generative AI capabilities without passing the cost to clients.
Sources
- Forrester and 4A's, The State Of AI Inside US Marketing Agencies, 2026: Forrester press release, June 24, 2026, checked 2026-10-02
- 75% of agencies bear generative AI costs; some agencies cutting production costs 40% to 50%; the "human/technology equivalent" model: Jay Pattisall, Forrester, "The AI Cost Center Crisis", June 16, 2025, checked 2026-10-02
- New York synthetic performer disclosure law: Governor's announcement of June 9, 2026, checked 2026-10-02
- California SB 1050 (chapter 246 of 2026): bill text, bill history; effective date under the California Constitution, Article IV, Section 8: leginfo, checked 2026-10-02
- TikTok advertising policy, "Misleading and false content" (last updated April 2026): TikTok Business Help Center, checked 2026-10-02
- 8frame prices (Kling v3 Standard 86 per 5 s with sound, 57 silent; Seedream 5 Lite 6; Nano Banana 2 at 1K 11; ElevenLabs Music 81 per started minute) and Montage limits: the 8frame tool registry and render code, checked 2026-10-02
For the workflows behind this repricing shift, read AI for marketing agencies, which walks through five production workflows with model routing. Ready-made templates are in 8frame workflows.
Try it without signing up: the 8frame canvas opens on a real board. The canvas is free and unlimited; generation is paid from $19/month.