Stop Paying Your Agency for Work AI Just Made Free
In the last article, I said most agencies still selling SEO services in 2026 aren't building this kind of content. They're not even close. That's the symptom. The pattern underneath it is bigger than SEO, and it's going to reshape how you spend your marketing budget over the next 12 to 18 months.
A lot of what your agency charges you for, blog posts, email nurtures, landing pages, monthly reports, social content, your own team can now do in-house. Not because they got smarter overnight. Because the AI did.
Forrester forecasts 15% of US agency jobs will be eliminated in 2026 alone after an 8% cut in 2025. Their earlier projection was 7.5% by 2030. They pulled it forward by four years. AI collapsed the cost of the work agencies charge for, and clients started bringing it home.
Meanwhile, global ad revenue grew nearly 9% in 2025 to about $1.14 trillion. At the same time, legacy holding companies like WPP and IPG posted revenue declines. The market for marketing is growing while traditional agencies lose share inside it. That only happens when the buyer rethinks what they're actually buying.
This isn't bad news for you. It's a permission slip, and a budget unlock. The real question is what comes home, what's still worth paying for, and the one thing you have to get right before any of it works. That's what this article covers.
A lot of the work your agency charges for is now work a trained junior on your team can do
Most agency retainers are loaded with assembly work: content production, campaign execution, monthly reporting, and platform configuration. AI made every one of those tasks cheap. A marketing coordinator running an AI system trained on your company can replicate the bulk of what most builder agencies deliver, often faster than the agency itself. And you don't need a director to do it.
Marketing consultant Jack Skeels frames the work agencies sell as two things. There's assembly, the drafting, producing, configuring, and reporting that recombines known patterns. And there's the higher-order work, the judgment that reads the actual situation and decides what to do. He calls it assembler versus architect. AI replaces the assembler. It doesn't touch the architect.
Here's what "in-house" actually looks like now, concretely. Picture an AI system trained on your brand voice, your floor plans and communities, your buyer, and how to write content that's triple-optimized for SEO, AEO, and GEO. A marketing coordinator hands it an approved brief or outline and triggers a skill. Within about ten minutes, there's a first draft, ready for a human to review.
From there the coordinator puts eyes on it. Fact-checks. Does light editing, often with an AI editing tool. Drops in the image. Publishes. The drafting, structure, optimization, and formatting, the work that used to be the agency's billable hours, is done by the system in minutes. The human time is the review and the polish. Not the production.
The quality claim, stated plainly: a properly trained junior operator can be every bit as productive as an agency and produce better deliverables, when three things are true.
- The tools are right.
- The tools are connected to the right knowledge base.
- The operator knows what they're doing.
Miss any one of the three, and it falls apart. Get all three, and the agency's assembly work is no longer worth paying for.
The cost curve is what's driving this. AI inference for GPT-4-equivalent performance dropped from roughly $20 per million tokens in late 2022 to roughly $0.40 per million tokens, about a 10x decline per year. The work that was expensive to produce three years ago is now functionally free.
Who runs it matters. A marketing coordinator runs the system, not a marketing director. According to Glassdoor, marketing coordinators average in the mid-$60Ks, typically $53,000 to $83,000. Add AI tools at under $100 a month. You're not hiring senior talent to do assembly. You're giving a coordinator a system that does the heavy lifting.
The builders who want to be ahead of the pack are following the same path. It's worth naming so you can see yourself in it:
- They learn to produce their content in-house.
- They learn to run their own reporting, so leadership can make better decisions faster.
- They get deeper into the AI toolset over time, so they can produce more work without adding more people to the org.
That's the transition. Quietly happening at the leading-edge builders right now.
What's actually worth paying for now (and it isn't assembly)
Stop paying for assembly. Start paying for two things AI can't hand you out of the box: judgment and strategy, and the work of getting your in-house AI capability built and your people trained to run it. The first is what to do. The second is the engine that lets a junior operator execute it at agency quality, without an agency.
Bucket one: judgment and strategy. Strip out everything AI just made cheap and what's left is judgment. Strategic diagnosis. Pattern recognition across builders. Knowing what to do when the playbook stops working. AI doesn't know what your sales team is hearing from leads. It doesn't know why your last marketing hire didn't work out. It can't see across 30 other builders' data.
Three quick examples of what that looks like in practice:
Buyers in mountain markets respond differently to cost-of-build content than buyers in flat markets. The fix isn't more cost content. It's a different format for it. A single builder's coordinator only ever sees their own data, so they can't spot that pattern. Someone who sees across builders can.
Your leads look fine, but they aren't turning into sales. The instinct is to spend more to get more leads. The real problem might be how fast and how well your sales team follows up. More leads won't fix a follow-up problem. Recognizing that the bottleneck is the sales handoff, not lead volume, is judgment. AI can't catch it because it isn't sitting in your sales meetings.
Reading the sales team's complaints in this month's meeting and turning them into a content shift before next month's numbers reflect the gap. That's anticipating. Not reporting.
Bucket two: getting the AI capability built and your team trained. This is the honest catch-22. Everything in the section above, the trained system, the skill that drafts in ten minutes, the triple-optimized output, doesn't happen by accident. Someone has to build the knowledge base. Encode the brand voice. Build the skills. Wire the SEO/AEO/GEO optimization. Train the operator to run it.
That's real work. It's worth paying for. And it's the opposite of an open-ended assembly retainer. You pay to stand it up and learn to run it. Then you own it.
One more thing here. The system is only as good as what you feed it. A coordinator producing eight blog posts a month from bad briefs is just producing bad content faster. The brief is the strategy. The operator is the executor. That's why judgment and infrastructure go together. Get the strategy right and the system turns it into gold. Get it wrong and AI just helps you fail faster.
This isn't "fire everyone outside the building." It's a re-allocation. You move spend off the work AI made cheap and onto the work that compounds.
The cut, and where the freed-up money should go
The 2026 builder marketing setup is a split. Assembly comes home to a trained coordinator and an AI system. Judgment, strategy, the capability build, and a few specialized technical skills stay as paid outside help. Here's the cut, based on what AI can actually do today. And then the part most builders miss: what to do with the money you free up over time.
Bring in-house (a trained marketing coordinator plus your AI system handles all of this):
- Blog post production from approved strategic briefs
- Email nurture writes
- Landing page copy and standard builds
- Basic social content
- Standard monthly performance reports
- HubSpot day-to-day execution and list management
- Image generation for content (Gemini, ChatGPT, Midjourney)
Keep paying for (outside help, but for the right things):
- Strategic intelligence and buyer journey architecture (the brief that drives every piece of content, not the piece itself)
- Cross-builder pattern recognition and competitive monitoring
- Setting up your in-house AI capability (knowledge base, brand voice, skills, SEO/AEO/GEO optimization) and training your operator to run it
- Technical support for your AI tools and marketing automation platforms. These are specialized skills you don't need full-time. Outsource them instead of hiring for them.
- Custom HubSpot architecture and integrations beyond standard config
- Keyword and SERP strategy across multiple markets
Now the upside. This deserves real estate.
Organic search traffic is heading down. Gartner projects search engine volume will drop 25% by 2026 as buyers shift to AI answers. I covered this in the first article in this series and again in Blog 3 on triple visibility. The reach you used to get for free is drying up. So as your assembly costs fall, redeploy that budget where it still buys attention. Paid ads for exposure, traffic, and leads. Other growth marketing that moves the pipeline.
Here's the honest math, and what it's really for.
Bringing this in-house is not a clip-the-coupon cost cut. At least not in year one. You're funding three things: a coordinator to run the system (mid-$60Ks per Glassdoor), a real investment to stand up and train your in-house AI capability, and a smaller set of outside relationships for judgment, strategy, and technical support. In year one, that can land in the same neighbourhood as a heavy agency retainer.
The difference is what you're buying. A retainer rents you output that stops the day you stop paying. This builds a capability you own. The cost curve bends down in year two and beyond, your output climbs without adding headcount, and the budget you free over time is what you push into paid reach.
One more time, because it's the whole game. All of this is gated by one thing. You have to have the proper AI platform set up in-house first, and your people trained to run it. Without that, "bring it home" is just a slower, worse version of what the agency was doing. The platform is the unlock. It's also the hardest part. And it's not something most builders can stand up alone.
Frequently Asked Questions
What can a marketing coordinator actually produce in-house with a trained AI system?
A trained coordinator running a properly built AI system can produce SEO/AEO/GEO-optimized blog posts, email nurture sequences, landing page copy, basic social content, and standard monthly performance reports. They can also handle day-to-day HubSpot execution and image generation. The system does the assembly work in minutes. The coordinator handles review, fact-checking, light editing, and publishing.
How fast can AI really turn a brief into a usable draft?
About ten minutes to a review-ready first draft, then a human does light editing before it publishes. Ten minutes is for the AI portion. A coordinator then spends time on review, fact-checking, and polish, often using an AI editing tool. Total human time per article drops sharply versus old workflows, and the draft is good enough to start from instead of building from scratch.
What should I still pay for outside in 2026, if not content production?
Three things. First, judgment and strategy: the brief, the pattern recognition across builders, the competitive read your coordinator can't generate alone. Second, the build itself: standing up your in-house AI capability and training your operator to run it. Third, specialized technical support for your AI tools and marketing automation platforms. Skills you don't need full-time on staff.
Do I need a director of marketing to run this, or can a coordinator do it?
A trained coordinator can run the day-to-day. You're not paying senior talent to do assembly anymore. According to Glassdoor data, marketing coordinator salaries average in the mid-$60Ks. That said, you still need senior judgment somewhere, whether that's your VP of Marketing, your CMO, or a strategic outside partner. Strategy doesn't go away. It just stops being bundled with assembly.
What's the realistic cost of running marketing this way, and what do I do with the savings over time?
Year one often lands in the same neighbourhood as a heavy agency retainer. You're funding a coordinator, the platform build, training, and outside help for strategy and technical support. The difference is what you own at the end. Year two and beyond, the cost curve bends down while output climbs. The budget you free up should go into paid reach, since Gartner expects search volume to drop 25% by 2026, and the reach you used to get for free is shrinking.
Why can't I just give my coordinator a ChatGPT login and skip the setup?
Because a raw ChatGPT login doesn't know your brand voice, your communities, your floor plans, or your buyer. It doesn't have your strategic briefs. It doesn't know how to triple-optimize for SEO, AEO, and GEO at the level Google and the AI engines reward in 2026. The output looks like AI wrote it, because it did, without any of the context that makes content actually work for your business.
Won't in-house content quality be worse than what my agency produces?
Only if you skip the setup. A coordinator running a properly trained AI system, with the brand voice encoded, the knowledge base built, and the SEO/AEO/GEO optimization wired in, produces work that's as good as or better than what most builder agencies deliver. The catch is the word "properly." Without the build and the training, in-house quality drops. With it, the quality goes up and the speed goes up at the same time.
Where this leaves you
The named urgency is on your wallet, on a 12-to-18-month clock. The liability isn't your agency dying. It's you continuing to pay $7,000 to $12,000 a month for work that's now a trained coordinator and an AI system away from being free, while the builder down the street brings it home and redeploys that budget into reach and strategy that compound. Forrester says 15% of agency jobs go this year because builders and brands are already doing exactly that. Over the next 12 to 18 months it stops being early and starts being obvious.
The question isn't "should I have an agency?" It's "what comes home, what's worth paying for, and do I have the AI platform set up to pull it off?"
That last question is the whole game. Bringing assembly home only works if you've got the right AI capability in-house and your people trained on it. That's exactly the part you can't buy off a shelf or fake with a ChatGPT login. We've spent the last year building exactly that for builders. The next article is where I show you what it is, and it drops in a couple of days.
Jeff, Founder, Velocity23
