TL;DR
- Agencies accounted for 20.7% of 2025 marketing spend, and 39% of CMOs planned to cut agency budgets, according to the Gartner 2025 CMO Spend Survey.
- 22% of CMOs in that survey say generative AI has already reduced their reliance on external agencies for creativity and strategy.
- Halving agency spend is arithmetic, not magic: it means moving high-volume, template-driven production in-house while keeping strategy and brand-defining work external.
- Gartner found early generative AI adopters reported average cost savings of 15.2% and productivity gains of 22.6% — a 50% cut in agency line items requires reallocating work, not just adding tools.
How much of marketing budget goes to agencies?
Agencies take roughly a fifth of marketing spend. The Gartner 2025 CMO Spend Survey — 402 marketing leaders across North America, the UK, and Europe — put agencies at 20.7% of 2025 marketing spend, with overall budgets flat at 7.7% of company revenue for a second year.
That combination is why agency lines are under pressure. When the total doesn't grow, the only way to fund anything new is to reallocate. Gartner found 39% of CMOs planned to cut agency budgets, with the most common tactics being eliminating unproductive relationships and streamlining agency rosters, followed by renegotiating contracts and scopes of work.
Which work can realistically move in-house?
The work that moves in-house successfully is high-volume, template-driven, and brand-defined. The work that shouldn't is strategy, brand architecture, and the small number of hero executions that set the creative standard everything else follows.
The ANA's in-house agency research found 82% of marketers now have in-house agencies, handling 58% of total marketing workload. The categories most often internalised are telling: collateral and promotional materials (83%), email (80%), social media (78%), and brand identity (72%).
Notice what those have in common. They are repeatable, they are high-frequency, and the creative direction is already set. That is precisely where AI-assisted production changes the unit economics — and precisely where agency retainers are least efficient, because you are paying senior rates for adaptation work.
Where do the savings actually come from?
Savings come from four places, and only one of them is the tool. Volume is the reason it adds up: Adobe research found 62% of marketers say content demand has risen at least fivefold in two years, and seven in ten organisations now produce 1,000+ assets annually.
1. Adaptation and resizing. Turning one approved concept into 40 format, market, and language variants is the single largest volume category in most agency scopes, and the most mechanical.
2. Iteration cycles. Each agency round has coordination cost attached. Adobe found that for nearly half of marketers, moving one asset through creation, review, approval, and activation involves 51 to 200 people. Internal iteration collapses that loop.
3. Roster consolidation. Gartner's respondents named eliminating unproductive agency relationships as their top cost lever — often worth more than any per-asset efficiency.
4. Scope renegotiation. Once volume production moves internally, the remaining agency scope is smaller and more senior. You are buying thinking, not throughput, and that contract should look different.
Is a 50% reduction realistic?
A 50% cut to agency line items is achievable, but not through tooling alone. Gartner's survey of 822 business leaders found early generative AI adopters reported average cost savings of 15.2% and productivity improvements of 22.6% — real, but well short of half.
Getting to 50% means combining that efficiency with structural change: consolidating the roster, moving adaptation volume internally, and renegotiating what remains. In the Gartner CMO survey, 22% of CMOs said generative AI had already reduced their reliance on external agencies for creativity and strategy development.
Be honest about the offsetting costs. Digiday has reported on the hidden costs marketers encounter with generative AI in ad campaigns — rights clearance, review overhead, rework, and platform fees that don't appear in the initial business case. And Gartner has predicted 30% of generative AI projects would be abandoned after proof of concept by end of 2025, citing poor data quality and unclear business value.
How do you start without breaking anything?
Start with one campaign type, not the whole roster. Pick a high-volume, low-risk category — social adaptations or email creative — and run it internally for a quarter alongside your existing agency work.
Measure three things: cost per asset, cycle time from brief to live, and a brand compliance score on the output. If all three hold or improve, expand the category. If compliance drops, the constraint is your templates and guidelines, not the tool.
Then renegotiate. The mistake is cutting the retainer before proving internal capacity — you end up paying rush rates to fix the gap and conclude that in-housing doesn't work.
FAQ
What percentage of marketing budget should go to agencies?
There is no universal benchmark, but Gartner's 2025 CMO Spend Survey found agencies averaged 20.7% of marketing spend across 402 organisations in North America and Europe. Your appropriate share depends on internal capability, output volume, and how much brand-defining creative work you commission.
Can AI replace a creative agency?
No. AI-assisted production replaces adaptation and volume work — resizing, localising, versioning against an approved concept. Strategy, brand positioning, and originating creative ideas remain human work, and ANA data shows organisations keep those with agencies even at high in-housing rates.
How much can AI reduce creative production costs?
Gartner's survey of 822 business leaders found early generative AI adopters reported average cost savings of 15.2%. Larger reductions in agency line items typically come from combining that efficiency with roster consolidation and scope renegotiation rather than from tooling alone.
What are the hidden costs of in-housing creative production?
The main ones are internal labour, tooling and licensing, asset rights clearance, template and guideline development, and review overhead. Digiday has reported that marketers using generative AI in ad campaigns frequently encounter costs that were absent from the original business case.
Should I cut my agency retainer first or build internal capacity first?
Build capacity first. Run one high-volume category internally for a quarter while the agency relationship is still in place, measure cost per asset, cycle time, and brand compliance, then renegotiate scope based on what you proved rather than what you projected.
Sources
- Gartner — 2025 CMO Spend Survey (2025)
- ANA — In-House Agencies No Longer A Trend, They're Here To Stay
- Adobe — Content demand set to grow 5x by 2027
- Digiday — Marketers are keen to use generative AI in ad campaigns, but hidden costs lurk
- Gartner — 30% of Generative AI Projects Will Be Abandoned After Proof of Concept (2024)