Perspectives · AI marketing
The 80/20 rule of AI marketing: the last 20% is the whole game.
AI gets every company to competent, and competent is camouflage. What separates you from slop and sameness is taste, and there is no API for it.
AI gets every company to competent, and competent is camouflage. What separates you from slop and sameness is taste, and there is no API for it.
The 80/20 rule of AI marketing is this: AI can now produce 80 percent of the work, fast and cheap, but that 80 percent only gets you to competent, and competent is what everyone else has too. The last 20 percent, the taste, judgment, and editorial kill that make the work unmistakably yours, is now the entire competitive difference.
Anyone can prompt their way to a first draft now. A landing page, a campaign, a positioning statement, a month of social posts. Twenty minutes, no expertise required. That is the new floor. And here is what nobody selling AI tools says out loud: the floor all sounds the same.
I run my practice on AI agents. They produce market intelligence briefs, outbound campaigns, content at scale, competitive analysis, working websites. I am not an AI skeptic. I am the opposite. One of those agents is torn down in detail in the Monday morning brief. The 80% is real, and it has permanently changed what marketing costs.
I have now reviewed thousands of agent outputs, and the pattern is absolute. The 80% gets you to competent. Competent is everywhere. When every company in your category has access to the same models, competent is camouflage. The model was trained on the same category language your competitors are prompting it with, so the drafts converge on the same voice, the same claims, the same shape.
It is a different kind of work. It is knowing that the technically accurate headline is also the boring one. It is catching the claim that reads fine but will not survive a skeptical CFO. It is recognizing that the draft sounds like your category instead of your company. It is the editorial kill, the tenth revision of one sentence, the decision that this piece should not publish at all.
That work does not come from a model. It comes from thirty years of watching what great looks like, what converts, and what quietly erodes trust. Taste is compressed experience, and there is no API for it. I learned the cost of skipping it the hard way, and wrote it up in I killed my own AI outreach program.
One of three places:
My whole practice is built on the third position. Agents for the 80%. Three decades of judgment for the 20%. The same divide shows up in how AI engines talk about you, which I cover in how AI engines describe your company.
It can produce most of the volume: drafts, lists, reports, variations, monitoring. What it cannot do is decide what is worth producing, kill what should not ship, or make the work sound like your company instead of your category. Without that layer, more AI output just means more sameness, faster.
Because the models were trained on the same material your competitors are prompting against, and left unedited they regress to the average voice of your category. Generic is the default output. Distinctive requires a human with taste overriding the default, line by line.
Put a senior editorial layer on top of the machine: a defined voice the agents must write against, a human who kills weak work before it ships, and strategy that aims the volume at the right targets. The tools are the floor. Judgment is the differentiation.
Todd Henderson is a fractional CMO for founder-led companies, pairing 30 years of brand and marketing leadership with a deployed team of AI agents for market intelligence, outbound, reporting, content, and GEO. He is a Co-Founder of Defining.com, the naming and branding agency, and runs his fractional practice as a separate embedded-leadership engagement.
If your AI-assisted marketing feels productive but somehow invisible, that is the 20% gap. Book a call and we will find where it is costing you. Or start with the free GEO assessment: how the AI engines themselves describe you, versus the herd.