Almost every content team I talk to has adopted AI. And almost every content team I talk to has adopted it backwards.
The inversion is consistent enough that I've started treating it as the default assumption: when an agency tells me they're "using AI in their content workflow," I know the human-in-the-loop sequence is almost certainly wrong before they've described it. Not because AI is wrong to use — it's mandatory now, full stop — but because the steps they've chosen to automate aren't the ones that should be automated.
Here's the giveaway: the content they produce is structurally correct, factually plausible, and completely indistinguishable from a dozen other sites in their category. It's not bad. It's just not useful. And "not useful" is the new bad in search.
"I think you have to be building workflows around content production with AI. I don't think there's really an option anymore. But a lot of agencies are designing their processes the wrong way — the steps they're choosing to use AI for aren't the best ones." — Nick Eubanks, Chief Marketing Officer, Digistore24.
The Wrong Sequence (And Exactly Why It Fails)
Here's what most teams do: they give an AI a keyword and a rough topic, ask it to generate a brief and a draft together, then a human does light proofreading before publishing.
The problem isn't that AI is in the workflow. The problem is where AI is in the workflow.
The brief is the highest-value step in content production. It's where genuine market knowledge, competitive context, strategic judgment, and actual expertise enter the process. When a human writes a brief from a real understanding of the space — from a podcast conversation, from client workshops, from 10 years of watching what moves rankings and what doesn't — the resulting content has an angle no one else is writing. It has perspective. It has something to say.
When AI writes the brief from a keyword and a topic, the brief describes the average of what everyone else has already written. And a brief about the average produces content that says nothing new.
So teams are automating the step that requires expertise and keeping humans on the step that requires grammar. That is almost exactly backwards.
This is not a technology failure. It's a workflow design failure. And it's producing a generation of content that is indexed and forgotten because there's nothing in it worth remembering.
The Right Sequence: Three Stages, Non-Negotiable Order
"Where I've seen the best results from a content perspective is where it's humans actually doing the brief, then it's AI writing the draft, and then it's humans doing the editing. The reverse of how most teams do it," says Eubanks.
Stage 1: A human writes the brief.
This is the stage that determines whether the finished piece has a defensible angle, genuine insight, and something worth a reader's time. AI cannot write this for you — not because the technology isn't capable of generating text that looks like a brief, but because a brief generated from keywords alone produces a brief about what's average.
A strong brief includes: the target keyword and search intent; the specific angle that differentiates this piece from what already ranks; the audience and what they genuinely need to know; key points to cover and — equally important — key arguments to avoid because they're generic or overdone; any client-specific data, first-person experience, or proprietary insight to include; and competitor URLs to analyze and differentiate from.
The brief built from an expert conversation — a recorded podcast interview, a client debrief, a subject matter expert walkthrough — is always stronger than one built from keyword research alone. Keywords tell you what people are searching for. An expert tells you what they should actually know.
Stage 2: AI writes the draft.
With a strong brief in hand, this is where AI earns its keep. Feed the complete brief — not a vague summary, the full brief with all context and constraints — and let the model produce a structurally sound, topically coherent working draft.
What you will not get from AI in Stage 2: distinctive voice, original perspective, proprietary data points, first-person credibility signals, or genuinely novel insight. Those aren't failures of the technology. They're inputs it doesn't have access to. That's Stage 3's job.
What you will get: structure, coverage, fluency, and a usable draft in a fraction of the time a human first pass would take. For a standard 1,500-word article, that's roughly 40–60% of the total human time requirement eliminated. That's a real efficiency gain — and it's what funds the expertise investment in Stages 1 and 3.
Stage 3: A human edits.
This is not proofreading. Stage 3 is substantive revision, and it's where quality is actually produced.
The editor's job is to add everything the AI couldn't: expert quotes, first-person experience, proprietary data, the author's specific voice and perspective, and corrections for any hallucinated claims (which are common and must be caught before publication).
More importantly, the Stage 3 editor needs to audit every paragraph for one specific test: could this paragraph have been written about any client, in any industry, at any time? If yes, it gets rewritten with something specific. Generic paragraphs are the fingerprint of AI content that skipped expert editing. Readers feel it. Search engines are increasingly able to detect it through engagement and return-visit patterns.
The Real Cost-Benefit Picture
I've heard the objection: if we're paying humans to brief and paying humans to edit, what are we actually saving?
The answer is Stage 2 — which, for a standard 1,500-word piece, represents roughly 40–60% of the total human time previously required. You're not eliminating human involvement; you're concentrating it where it matters (strategy and quality control) and eliminating it where it adds the least value (raw drafting of a first pass).
Fully automated content that skips expert input — looks cheaper in the short term. But it accumulates risk fast: reputational risk as readers recognize undifferentiated content; SEO risk as search engines weight expertise and entity signals more heavily in rankings; and strategic risk as your content becomes indistinguishable from every competitor using the same tools.
The agencies I've seen extract real, durable value from AI content services are the ones who use the efficiency gains from Stage 2 to invest more in Stage 1 briefs — not the ones who eliminate Stage 3 and pocket the margin. The latter is a short-term arbitrage. The former is a compounding advantage.
How to Operationalize This System at Scale
If you want to build this workflow across a team — whether you're a solo creator systematizing your process or a content operation producing dozens of pieces per month — here's the structure that actually works.
- Create a brief template that forces Stage 1 to include every required input: keyword, search intent, angle, differentiator, expert inputs, URLs to beat, and specific data or claims to include. A template ensures the brief is complete even when the person writing it is under time pressure.
- Track the ratio of human time at each stage. If you're spending more human hours in Stage 2 (drafting) than in Stages 1 and 3 combined, your workflow is inverted. Rebalance before scaling.
- Establish a Stage 3 checklist for editors. Voice check. Specificity check (flag every generic paragraph). Factual accuracy check. First-person contribution check. Make Stage 3 a defined, scoped job — not an open-ended "just clean it up."
- Measure content performance at the piece level and connect it back to the quality of the Stage 1 brief. Over time, you'll develop a clear correlation between brief quality and content performance. That correlation becomes one of your most valuable operational datasets.
The teams that get this right will produce content that genuinely cannot be replicated by a competitor who just runs keywords through an AI and hits publish. That differentiation is increasingly rare — and in search, increasingly valuable.
