Why Are Clients Cutting Agency Retainers in 2026?
An agency owner shared this story in public this week. A client had been paying him $5,500 a month. The client first dropped the retainer to $3,500. Then another offer arrived: $500, paid once, to teach the in-house editor who would replace his work. He refused.
Most people who saw the story treated the offer as an insult. The data points to a much larger shift. Stensul surveyed 321 marketing technology buyers for its 2026 MarTech Outlook, released in February 2026, and 31% plan to reduce spending on external agencies for campaign execution this year. That same survey found 57% plan to invest in AI reskilling for their own teams. The $500 proposal goes beyond one badly behaved client. The market is resetting the value of agency services, one budget line at a time.
Quick answer: Clients are cutting agency retainers in 2026 because AI pushed down the cost of execution, the part of the bill that pays people to make the work. Stensul's 2026 survey of 321 martech buyers found 31% plan to reduce external agency spend for campaign execution while 57% fund AI training for their own staff. The judgment part of the fee, deciding what should be made and why, still matters. Clients continue to want it. But many are trying to buy that knowledge for the price of one training session. That pricing gap will decide which agencies make it through.
Here is what buyers are doing with the money, in one view:
| What marketing teams are doing | Share | Source, with date |
|---|---|---|
| Plan to cut external agency spend for campaign execution | 31% | Stensul 2026 MarTech Outlook, Feb 2026, 321 buyers |
| Plan to invest in AI reskilling for their own team | 57% | Same survey |
| Expect marketing technology budgets to rise in 2026 | 79% | Same survey |
| Run an in-house agency | 82% | ANA, 2023 edition, 162 members |
| Still also use an external agency | 92% | Same ANA study |
| Average share of work already done in-house | 61% | Same ANA study |
Look at the first two rows together. Companies are cutting execution budgets while putting money into training. The marketing budget remains. Its job changes. Money moves from paying someone else to deliver finished work toward building the skill to make that work internally. That is exactly what the $500 client was attempting. The problem was the price. He offered $500 for a method that had been billed at $66,000 a year.
What did the $500 offer actually price?
The $500 offer put a price on transferring a proven method, the system supporting $66,000 a year in billing, and valued it like one training session. Compare that amount with public wage data and the gap becomes obvious. The U.S. Bureau of Labor Statistics puts the median editor's pay at $75,260 a year, or $36.18 an hour, as of May 2024. At that rate, $500 covers about 14 hours of a median editor's time. The proposal was simple: pay the person who built the method for 14 editor-hours to teach years of accumulated knowledge, then keep using that method through a $75,260 employee for as long as needed.
That math makes the anger easy to understand. But it also reveals the demand. The client was not offering $500 for another batch of articles. In his view, execution had already been handled by AI tools and the editor on staff. The $500 was aimed at the part he still lacked: judgment. Which ideas should get published. Which ones should die. What quality means. Where the bar should sit. He valued the method enough to try buying it after ending the actual service.
That is what most of the industry reaction overlooked. A real rejection would have been no offer at all. This client made a bid.
A bid can be wildly wrong while still proving that someone wants the asset. Anyone who has sold a house understands this. A lowball offer says the buyer wants what you have and hopes you will accept less than it is worth. The answer is a counter. Later in this piece, we will cover what that counter should look like. The useful response takes the request seriously and charges for the asset being transferred.
Why are clients cutting agency retainers in 2026?
Clients are reducing retainers because AI lowered the cost of making marketing work, while many retainers were built around the idea that making the work was the expensive part. For most of the past two decades, agency fees mainly paid for time. Hours spent writing, designing, editing and posting. Once software can create a workable draft in seconds, the part of the fee tied to human hours loses much of its old price floor. Buyers know this. Survey data shows they are changing their spending because of it.
Stensul's 2026 MarTech Outlook surveyed 321 buyers of marketing technology in the U.S. and U.K. in February 2026. AI tools are their number one planned investment for the year. 79% expect budgets to grow. At the same time, 31% plan to cut external agency spend for campaign execution. The report describes the shift as a move toward AI-enabled in-house execution, with reskilling taking priority over outsourced campaign support.
AI did not create the move toward in-house marketing. The Association of National Advertisers has tracked the trend for 15 years, and its 2023 edition shows a clear climb:
| Year measured | ANA members with an in-house agency |
|---|---|
| 2008 | 42% |
| 2013 | 58% |
| 2018 | 78% |
| 2023 | 82% |
Source: ANA, "The Continued Rise of the In-House Agency: 2023 Edition," published 2 May 2023, 162 member companies surveyed.
That study found 65% of members had brought work in-house within three years that had previously gone to an external agency. Cost was the number one reason companies built internal teams. AI has now landed on top of a cost shift that had already been running for 15 years, making an existing trend move much faster.
Another Stensul finding matters here. The buyers cutting agency execution spending are also increasing their budgets. So this is mainly money moving to a different place. Companies are spending less on rented execution and more on capability they can own.
What is an agency retainer actually made of?
An agency retainer is a fixed monthly fee that pays for two different products at once: the hours that produce the work, and the judgment that decides what work is worth producing. For twenty years, those products were sold together because the same people delivered both. There was little reason to give each one its own price.
AI split them apart. The hours now have an obvious market value, and that value is cheap. Judgment still lacks a clear market price, which is where the fight begins. When the client moved from $5,500 to $3,500, he was cutting the value assigned to execution hours. When he offered $500 for training, he made a separate offer for the judgment itself. The agency owner had never sold that judgment on its own either. That meant the buyer got to name the first standalone price, and buyers usually start low.
The ANA numbers show a similar split from the buyer's side, even before the AI surge reached its current level. Between 2018 and 2023, cost savings became less important as the main measure of in-house agency success, dropping from 69% to 62%, while business performance rose from 45% to 59%. Companies discovered that cheap internal execution does not automatically create strong results. The same study found that external agencies still received work for two main reasons: bandwidth and capabilities that did not exist internally. As AI takes on more of the bandwidth problem, the remaining value sits in the second reason: skills and judgment the internal team does not yet have.
That is what the $500 was trying to buy.
Does moving marketing in-house actually work?
Often, the first attempt struggles. The readiness numbers show why. Gartner's 2026 CMO Spend Survey polled 401 marketing leaders and came out in May 2026. 70% of CMOs call becoming an AI leader critical to their 2026 goals. Another 70% say their marketing processes are not mature enough to scale AI. Only 30% say they are ready. CMOs are directing 15.3% of their budgets toward AI while also admitting, in the same survey, that their current systems are not ready to support it at scale.
The internal team carries that gap. The editor taking over from the cancelled agency gets the AI tools immediately. The method is harder to inherit because nobody ever documented it. It existed inside the agency owner's decisions. The client understood that problem, which is exactly why he asked for training after ending the retainer.
Agency data gives us another clue. The RSW/US 2026 New Year Outlook Report came out in January 2026, built on a November and December 2025 survey of marketers plus more than 5,000 agency and professional services firms. It lists "capitalizing on the pullback from in-housing" as one of three opportunities for agencies to regain ground in 2026. That wording matters. There cannot be a pullback unless some companies that moved work inside are now reversing course. Businesses brought work in-house to cut costs, ran into the 70% maturity problem, then learned which part of the old agency fee they still needed.
We covered the buy-or-build choice in Can AI Replace Your Marketing Agency?, and the same conclusion applies here: the tools work, but the replacement only covers part of the job. The hardest part to replace is the part a prompt cannot supply on its own. Consistency also carries value. McKinsey's 2026 B2B Pulse found the top reason buyers switch suppliers this year is inconsistent information across teams, and an internal marketing function still figuring out its standards creates exactly the conditions where that inconsistency grows.
What happens when the trained editor leaves?
The method leaves too, and that risk is larger than many founders account for. The Bureau of Labor Statistics counts 115,800 editor jobs in the U.S. as of 2024. It expects about 9,800 editor openings each year through 2034. Very few of those openings come from growth. Across the full ten-year forecast, the market adds only 700 net jobs. Almost every opening comes from replacing an editor who changes occupations or leaves the workforce.
Now apply those numbers to the $500 plan. Under the client's setup, the entire method sits inside one employee earning $75,260. There are no documents. There is no formal review system. Nobody else carries the standard. The training happens in one afternoon, so the knowledge never gets turned into process. When the editor quits, the method goes with them. A new editor starts with none of it. By then, calling the old agency becomes awkward because the agency's last offer was $500.
An agency's method has to survive individual employees. That is why the method gets stored in checklists, review systems, positioning documents and measurement routines. A writer can leave while the standard remains. That is the version worth purchasing. The useful transfer puts the method into documented processes inside your company, where it survives staff changes. Building that system takes weeks of structured work with the team. That is also why proper enablement has a real price. A $500 session buys knowledge that sits inside one person's head. The fully priced version is the method your company still owns three resignations later.
Why did the client's math feel right to the client?
Because from the buyer's seat, the costs he could see had dropped while the hidden costs stayed hidden. He can see the software bill, the editor's salary, the draft showing up in seconds and the agency invoice disappearing. He cannot see the judgment calls waiting next quarter. The campaign that should be stopped. The topic that will sound like more noise. The claim that falls apart under scrutiny. The quality bar that slips a little each month because nobody senior is checking it.
We see the same pattern in search. In our zero-click analysis, the obvious metric, traffic, went down while the thing that mattered, being the name inside the answer, became the main fight. What people could see and what really carried value moved apart. Buyers who optimize only for visible costs make confident decisions that value the hidden half at zero. Or at $500, which is close enough.
The client's mistake is smaller than much of the commentary makes it sound. He correctly saw that execution is cheap now. He also correctly saw that the method is the piece he does not have. He got one thing wrong: he assumed that method could be handed over in one afternoon. You can pass information in an afternoon. A method comes from applying judgment to case after case until patterns become clear. That needs a structured engagement that lasts weeks.
What should agencies sell when execution gets cheap?
Sell the method itself. Price it and package it on purpose before clients decide the price for you. The $500 story shows what happens when buyers want something that does not appear on the menu. They make up a number. The seller hears disrespect. The answer is simple: put the product on the menu.
The demand showing up in the data is buying four things:
- Strategy and standards. What should be made, who it is for and how good it needs to be. Think of the document the in-house editor was supposed to learn from, except properly built, kept current and owned by someone responsible for results.
- Enablement at a real price. Structured training for a client's team is now its own product line, and Stensul's data says 57% of the market plans to buy it this year.
- Judgment on tap. Review and direction for the work the client's own team creates. It needs fewer hours than the old retainer, but those hours do a different job.
- Measurement the client cannot easily build. Working out what succeeded and why, so the method keeps getting better instead of freezing in place.
This is the model we use at Moxie. We provide the strategy, standards, review and measurement. The client's team handles execution. We chose that position before it felt comfortable. Stories like the $500 offer show why. Execution budgets were always going to shrink. The better place to stand is on the part of the fee that had been undervalued for years. An agency that will not sell its method will watch clients hire the cheapest copy they can find. An agency that packages the method well gets paid for judgment and keeps the relationship even when the execution money moves elsewhere.
The RSW/US report gives agencies another warning if they plan to wait for the old model to return. It found that referrals, networking and past relationships, the traditional sources of agency new business, are becoming less dependable. The agencies that benefit from the pullback will be the ones buyers can discover and judge without help. That is the argument our thought leadership guide makes about becoming the name in the room before the room meets you.
The agency owner in the story was right to reject $500. The bigger mistake would be ignoring what the request revealed: the client wanted his method as a product. Refusing that demand as well would waste the signal.
How should a founder buy the method instead?
Buy it as a proper engagement with a start date, a clear standard, a scoreboard and a price that matches the value it replaces. If you are the client in this story, or you are about to make the same move, here is what the sensible version looks like:
- Buy the operating system, not a session. A method transfer that works takes several weeks. You build your positioning, content standards, review process and measurement system with your team using them as they are created. If the whole thing fits into one afternoon, you bought a slideshow.
- Keep senior judgment on a thread. In-house teams that work well rarely operate alone. ANA's data says 92% of companies with in-house agencies still use external partners. The practical model is internal hands with an external head.
- Fund the reskilling like the majority you belong to. 57% of buyers plan to. Give it a proper budget line instead of treating it like leftover money from a cancelled invoice.
- Measure the standard, monthly. The danger of in-housing appears in month six, when nobody in the room clearly remembers what better used to look like. Week one almost always feels fine.
Use the numbers from the story to see what the counter-offer could have been. The client was paying $66,000 a year. Suppose $40,000 covered execution and $26,000 covered judgment. AI plus an in-house editor can absorb much of that first number. The second number was never the biggest cost, but it was the part that kept building value over time. A counter such as "your team runs the hands, I keep the judgment layer at $2,000 a month" cuts the client's bill by more than half while keeping the method alive and updated. Both parties do better than with the $500 option, where the client saves money for one quarter and spends year two trying to remember what good work looked like.
The chart below sums up the 2026 market in three bars. Budgets are growing, training is getting funded and execution retainers are being cut:
Source: Stensul 2026 MarTech Outlook, February 2026.
Frequently asked questions
Are marketing agencies dying in 2026?
No. The execution-hours side of the business is getting smaller: 31% of buyers plan to cut agency spend for campaign execution this year (Stensul, Feb 2026). But 92% of companies with in-house agencies still use external partners (ANA, 2023), and the 2026 RSW/US Outlook points to a pullback from in-housing that agencies can win back. The fee is splitting into separate products, and the judgment side still has demand.
What is a fair agency retainer in 2026?
A fair retainer prices the two products separately. Work billed around execution hours should cost less than it did in 2023 because the cost of producing that work has fallen. Strategy, standards, review and measurement should carry the price of the senior work they are. If a retainer cannot show which product the client is paying for, the client will set a new price themselves, and their first number will be low.
Should I move my marketing in-house?
Move execution in-house if you have enough volume; 82% of large advertisers already run in-house teams (ANA, 2023). Keep outside judgment until your internal judgment is proven. 70% of CMOs admit their processes cannot yet scale AI (Gartner, May 2026), and quality can slide quietly when no senior person is watching it. The model that works is internal hands with external direction.
Can I just pay my agency to train my team?
Yes. Good agencies now sell this as structured enablement with a clear scope. 57% of marketing teams plan to fund AI reskilling in 2026 (Stensul). But $500 will not buy a working method. A proper transfer takes weeks, with your team creating real work inside the system while they learn it. It deserves professional pricing because it turns an ongoing cost into a capability your company keeps.
Was the agency owner right to say no?
Yes, on the price. The request itself showed where the market is going: clients now want the method more than the hours. Agencies that win from here will recognize that demand and attach a real product and a real price to it instead of ending the conversation with a no.
The bill was always split. Now everyone can see it.
The $500 story will keep spreading as an outrage post, and the anger makes sense. A method that drives a client's growth is worth more than 14 hours of a median editor's pay. But anger does not create a business model, and the numbers below the story keep moving regardless. Execution spending is being cut on purpose by roughly a third of the market at the same time. Training budgets are rising inside those same accounts. The judgment half of the old retainer has become the product, and buyers will purchase it from whoever gives it a clear and fair price.
If you are choosing which part of marketing to own and which part to buy, work out that split before your next renewal or your next $500 offer. Take our two-minute content agency readiness quiz and find out where your team sits on the buy-or-build line. It is free, and your budget is already answering the same question.