Two memes are fighting on your feed about AI and marketing agencies. Both are loud, both are half-right, and both skip the part that actually matters.
The first is triumphant: “I fired my $5,000-a-month agency and replaced it with a $20 AI subscription.” The second is the backlash, where someone runs a free AI SEO tool, gets generic slop, and decides AI marketing agents are a scam. Both come from something real. The internet is genuinely flooded with AI hype and AI spam right now, and scrolling past it is exhausting. But while everyone argues about which meme is right, the actual shift in the market goes unnoticed, and that shift is the useful thing to understand.
You do not pay an agency to do the work
Take the $5,000 meme first. It falls apart the moment you look at it.
Nobody who has actually paid an agency $5,000 a month trades that for a $20 subscription they have to run themselves. That was never the deal. You are not paying to get the work done. You are paying to hand off the responsibility for it, to have someone whose job is to worry about it, catch the problem before you do, and own the result. A subscription does not worry about you.
So the honest version of the meme is smaller and less flattering. If your agency could be swapped for a twenty-dollar tool you run yourself, it was thin to begin with, and AI just made that visible. A real one is held in place by results and a relationship, and a chatbot supplies neither. Mostly, AI exposed the spots where there was never much value in the first place.
The moat was never the tool
If your advantage was access, a tool or a template or a capability the client could not reach on their own, it disappears the month that capability costs twenty dollars. That has little to do with AI specifically. It is what happens to any moat built on access once the thing behind it gets cheap.
Judgment does not get cheap the same way. A good writer with AI is still a good writer, now faster and working at scale. A senior strategist with AI still knows which of the tool’s confident recommendations to throw out, which is most of the job. You spend a few hours editing in conversation with a model instead of a few days grinding through the same four pages, and that speed lets you take on more clients rather than fewer. The tool got commoditized. Knowing what to do with it did not. We have written before about why the tell in cheap content was never who typed it, because it comes down to whether there is a real point underneath, which is a question of judgment.
And the only way to develop judgment is to live through the consequences of your choices - experiential learning. You’ve got to put in the reps to get the pattern recognition.
The market got bigger at both ends
The market did not shrink when AI arrived. It grew at both ends.
At the bottom, a new kind of operator showed up. Someone grabbed the twenty-dollar subscription, hit the same generic wall everyone hits, then actually read how search and AI work and built a workflow of their own, shaped to a specific kind of client and a little sharper after every engagement. They can serve businesses that were never going to hire anyone. The local services company that cannot handle more than five new customers a month was never a $5,000 client, but a $250 package makes sense for them now, run by someone who could not have offered it a year ago. That is the floor dropping, and new customers walking in under it.
At the top, the senior shops that used that same speed to do more, and better, kept their enterprise work and took on harder problems. The ceiling went up.
What got squeezed is the middle: the layer whose whole pitch was “we have the tool and you do not.” That was always going to compress the moment the tool got cheap. Markets do this every time the cost of basic competence drops, and the industries it has happened to are mostly still here.
The only ones who get hurt are the ones who sit still
The apocalyptic version never arrived. AI did not end writing or design or strategy. The design tools everyone said would clear out the designers just made the good ones faster. What AI actually ends is narrow: an advantage that was only ever access to a tool, and the operator who decides to wait and see.
If your only edge was a capability anyone can now rent, this is a hard year. The good ones, handed a tool that strips the grind out of the parts that were never the point, just got the most leverage they have ever had. The difference is whether you moved.
That is what we mean by AI in the right hands. The value was always the person who knows which of the tool’s answers to trust, who has done this long enough to feel when something is off, and who kept learning while everyone else argued about memes. No subscription replaces that, and none was ever going to.
If you are trying to work out which side of that line your marketing sits on, book a 20-minute conversation. We will tell you honestly where the real advantage is, and whether the people you pay have it. No deck, no pitch.
Common questions
Will AI replace my marketing agency?
It will replace an agency whose only advantage was access to tools or templates you could not run yourself, because those now cost very little. An agency whose advantage is judgment, specific experience in your world, and ownership of the result is a different case. A useful test: could a twenty-dollar subscription your own team runs get you the same outcome? If it could, you were overpaying. If not, AI does not change what you are actually buying.
Should I fire my agency and use AI myself?
Most owners will not, for the same reason most companies do not do their own books even though they could. You are paying to offload the responsibility, not just the labor. The better move is to ask your agency how they are using AI, and whether the speed is reaching you as more and better work or just being kept as margin.
Does this mean AI content and AI SEO tools do not work?
No. The generic, off-the-shelf version is too generic to be a strategy, the same way a stock template never was one. Pointed at a specific problem, with a person editing and deciding, the same tools produce real, useful work. The tool is a starting point; the judgment around it is what you were paying for. We made the fuller version of that case in what a content engine actually is.