AI Was Supposed to Cut Marketing Costs. 42% of Teams Say It Made Them Spend More.
The sales pitch for marketing AI has been consistent for two years: automate the grunt work, do more with less, reduce headcount, shrink your tech stack, watch the savings roll in. Clean narrative. Easy to model in a budget proposal.
The actual data tells a different story. A Semrush survey released this week found that 42% of marketers say their budgets increased because of AI. Only 16% report decreases. Headcount is growing, not shrinking: roughly a third of teams saw significant headcount growth, with another quarter reporting smaller increases. And despite nearly half of respondents saying they replaced "lots of tools" with AI, their overall tech stacks are expanding, not contracting.
AI is not a cost-cutting tool for most marketing organizations. It is a reallocation engine that creates new expenses at least as fast as it eliminates old ones.
The Budget Paradox: Flat Overall, Expanding Inside
Gartner's 2025 CMO Spend Survey pegged marketing budgets at 7.7% of overall company revenue, essentially flat. That number has not moved meaningfully in years. So when 42% of marketers say AI increased their budgets, they are not talking about new money appearing from corporate. They are talking about existing line items being reallocated internally.
Where the money comes from is the interesting part. The same Gartner data shows 39% of CMOs plan to cut agency budgets, and 22% say generative AI has already reduced their reliance on external agencies. The flow is clear: money moves from agency retainers into AI tools, internal headcount to manage those tools, and the training required to use them effectively.
On paper, this looks like efficiency. In practice, it is swapping one cost structure for another, often a more complex one, without the overall budget envelope growing to accommodate the transition costs.
59% of CMOs in the Gartner survey say their current budget is insufficient. That number predates the AI reallocation wave. I suspect it is higher now, though nobody is saying so publicly because admitting your AI initiative costs more than the thing it replaced is politically difficult in most organizations.
The $2-to-$3 Training Multiplier Nobody Budgets For
This number comes from a CMO panel at SXSW 2026, and once you hear it you start seeing it everywhere: for every $1 spent on AI tools, expect to spend $2 to $3 on training and change management. A retail CMO cited a 2.4x ratio. A B2B leader cited 2.8x.
Think about what that means for a team that just bought a $50,000 annual AI platform license. The tool itself is the smallest part of the cost. You need someone to configure it, someone to validate its outputs, training for the team members who will use it daily, process documentation, and ongoing oversight to catch when it produces something wrong. That is $100,000 to $150,000 in hidden costs on top of the license.
Most AI tool evaluations compare the license cost to the agency retainer or FTE time the tool replaces. Very few include the training multiplier. So the business case looks positive at the point of purchase and starts eroding almost immediately once implementation begins.
We covered E.l.f. Beauty's approach to this recently, and what made their case interesting was the honesty about using AI to replace work rather than pretending it was a cost-free augmentation. Most organizations are not being that candid, which means the hidden costs accumulate invisibly.
Stacks Are Growing. That Was Not the Plan.
The Semrush data on tech stacks is probably the most counterintuitive finding. Nearly 50% of respondents said they replaced "lots of tools" with AI. Another third replaced "at least a few." You would expect overall stack size to shrink as a result. It did not. About a third saw slight stack increases, and roughly a quarter saw significant growth.
The explanation is boring but important: AI tools replace point solutions but introduce new capabilities that require additional tools, workflows, and integration layers. You eliminate a social scheduling tool because your AI platform handles posting, but then you add an AI content review tool, a prompt management system, an analytics integration to measure AI-generated content performance, and a compliance layer to review outputs before they go live.
The net result is more tools, not fewer. More integration complexity. More vendor management overhead. More things to break. And because AI tools tend to have steeper learning curves than the simple point solutions they replaced, the operational burden on the team goes up even as the theoretical capability goes up.
Teams with larger budgets ($500K+) are significantly more likely to have replaced more tools with AI, which suggests that scale accelerates this dynamic. Bigger organizations replace more things and add more things in return.
The CMO Blind Spot That Worries Me Most
Gartner's February 2026 survey found a gap that I keep thinking about: 65% of CMOs expect AI to dramatically change their role, but only 32% say significant skill changes are needed. That 33-point gap between "this will change everything" and "but I do not need to learn much" is what Gartner calls the AI blind spot.
If you believe the transformation is coming but do not believe you need new skills to navigate it, you are functionally planning to manage a revolution with your existing toolbox. That is the kind of assumption that works right up until it does not.
The broader workforce data reinforces this: only 17% of marketing professionals report having comprehensive AI training, while 58% cite skills gaps as their top challenge. The gap between available skills and required skills is not closing, and the people in charge of closing it do not seem to think it is very wide.
I do not know if this is denial, information asymmetry, or genuine miscalibration. But the combination of flat budgets, expanding stacks, growing headcount, hidden training costs, and leadership that does not believe it needs significant new skills is a recipe for a very uncomfortable budget cycle in Q3 or Q4 of this year.
The Agency Reckoning That Is Already Underway
According to eMarketer data, 60% of senior marketing leaders spent less on agencies in 2025 due to AI. The Omnicom-IPG merger eliminated approximately 4,000 positions, targeting $750 million in cost savings. Those are not subtle signals. The agency model is being restructured in real time, and AI is the primary catalyst.
But here is the part that does not get enough attention: the work that agencies did does not disappear when the retainer gets cut. It transfers internally, often to teams that are not staffed or skilled for it. A brand that drops a $200K creative agency retainer and buys a $40K AI content tool has "saved" $160K on the line item and acquired a capability gap that will manifest as lower-quality output, slower turnaround, or both. The savings are real and measurable. The quality erosion is real and harder to measure, which means it gets ignored until it becomes visible in campaign performance or brand perception metrics.
We wrote about Microsoft's own terms classifying Copilot as entertainment-grade recently. If the AI tools replacing agency work are not designed for mission-critical marketing applications, the assumption that they can absorb agency-level work is, at minimum, worth stress-testing before you cut the retainer.
What This Looks Like in a Budget Conversation
If you are a marketing leader planning your Q3 or 2027 budget, the honest framing is this: AI is not going to reduce your total cost. It is going to change where the money goes. Tool costs down, training costs up. Agency costs down, internal complexity up. Stack consolidation in one area, stack expansion in another.
The teams that are going to navigate this well are the ones who budget for the transition costs, not just the tool costs. Include the 2-3x training multiplier in every AI tool evaluation. Staff for the oversight and quality control that AI outputs require. And do not assume that cutting an agency retainer means the work goes away. It means the work goes somewhere else, and you need to decide where before someone else decides for you.
The narrative that AI makes marketing cheaper is going to age poorly. The more accurate version: AI makes marketing different, and "different" costs money too. It just costs it in less visible ways, which is exactly why so many teams are going to be surprised by their end-of-year numbers.
