- Every “AI dos and don’ts” list floating around right now says roughly the same thing: use AI for outlines, don’t use it for final copy, keep a human in the loop somewhere. Almost none of them explain why those rules belong together, or what happens when a team ignores one.
- Here’s the pattern underneath. Trust, memory and a voice a reader recognizes are relationship outputs, built between a brand and a specific reader over time, not something a model can generate on demand. AI can assemble information faster than any team could on its own. It can’t manufacture that relationship, because it has no relationship with your reader, only patterns from what it’s read.
- That’s exactly where AI earns its keep, in the first 80% of the work: starting points, outlines and first-pass drafts. The risk shows up when a team hands the other 20%, the voice and judgment calls, to the same tool.
- More content teams are leaning on AI for first drafts than they were a year ago, and when everyone starts from the same tools trained on the same public web, content converges. The brands that pull ahead over the next few years will be the ones that decide, on purpose, exactly where AI stops and a person picks up.
Where AI earns its keep
AI is genuinely good at starting points: frameworks, outlines and first drafts. Feed it your messaging platform and a specific prompt and it hands you something to work from in seconds, not a blank page.
Ask it for five headline options, a rough outline or a first pass at summarizing a customer call, and you’ll have something usable before you’ve finished your coffee. Same with research synthesis: feed it five competitor posts or a messy customer transcript, and it hands back a workable brief instead of an afternoon of note-taking. That’s real time back, not a shortcut worth feeling guilty about.
It also gets most drafts to about 80% fast. That’s real value.
“You should never let the AI create content and send it directly to the web,” says Tom Neff, Principal Product Manager at Brightspot. “You need human oversight to keep your content trustworthy, consistent and high-quality.”
That’s the 20% where a brand’s actual differentiation lives.
Where AI earns its keep
Starting points. Frameworks. Outlines. Feed it your messaging platform and a specific prompt, and it hands you something to work from in seconds, not a blank page.
Ask for five headline options, a rough outline or a first pass at summarizing a customer call, and you’ll have something usable before you’ve finished your coffee. That’s not nothing.
Same with research synthesis. Feed it five competitor posts or a messy customer transcript, and it hands back a workable brief instead of an afternoon of note-taking. That’s hours back, not a shortcut worth feeling guilty about.
It also gets most drafts to about 80% fast. That’s real value.
“You should never let the AI create content and send it directly to the web,” says Tom Neff, Principal Product Manager at Brightspot. “You need human oversight to keep your content trustworthy, consistent and high-quality.”
That’s the 20% where a brand’s actual differentiation lives.
If you embrace it and look to find out how to make it work to your benefit, I think you’re going to be winning with AI.
The don’ts aren’t a laundry list
Unique perspective. Trust. A memorable experience. AI can’t produce any of these because they aren’t information problems. They’re relationship outputs, and a model trained on public data has no relationship with your reader. It has patterns.
Say a reader has followed your blog for two years. They notice when a piece references something you said six months ago, holds a position you’ve held consistently, or sounds like the same person wrote it. AI has no memory of that reader, and no memory of your last 500 posts unless you feed them in this session. Even then, it’s matching patterns, not remembering a relationship.
The real risk is handing relationship work to a tool built for information work. Do that enough, and your content starts sounding like everyone else’s AI-assisted content: correct, but forgettable.
Jessie Kennedy, Brightspot’s VP of Marketing, is bullish on the use of AI as a tool, not spooked by it. “If you embrace it and look to find out how to make it work to your benefit, I think you’re going to be winning with AI,” she says. “It is time to embrace AI and make it work for you, productivity-wise and in creating better customer experiences.”
She’s right, for the 80%. Which brings us back to Tom Neff’s assertion that a human should always be the last one in the ownership and governance loop. It’s that last 20% that matters, plus the intangibles of what AI can never replace: being in the room or reporting on real-world facts and moments.
The leadership opportunity
The brands that pull ahead in the context of world in which is an embedded reality will be the ones that put a real standard in writing: exactly where AI stops and a person picks up, and why. A standard like that is something you can hand to a content team on day one. “Use good judgment” isn’t a standard. It’s a hope.
A written standard reads something like this: AI drafts every outline and first pass, a senior editor reads every piece for voice before it publishes, and nothing ships with a claim that isn’t sourced. That’s specific enough for a new hire to follow on day one.
For more on where AI fits into daily production, see how AI is transforming content workflows.
And on why protecting that 20% pays off, see rethinking content ROI in the age of AI.
When the 20% needs a specialist, that’s what our creative services team is for. Learn more about our team and services here.
What should content teams never hand to AI? Final voice and trust decisions. The calls that make a piece sound like your brand instead of a template.
Does that mean skip AI for drafting? No, not at all. Use it for outlines, frameworks and getting most drafts to 80% fast, then keep the last 20%, the polish and judgment, human.
Who decides where the line sits? The content lead, not a blanket company policy. It’ll flex by piece and by risk, but it needs to be written down somewhere the whole team can see it.
How is this different from a normal editorial review pass? A normal review checks grammar and facts. This checks something narrower: did the piece keep your actual point of view, or did it flatten into something a competitor could publish just as easily?
What’s the actual risk of getting the split wrong? Assigning voice and trust decisions to a tool that can’t hold either. That’s what turns good content into filler.
Does using AI for the 80% mean less human involvement overall? Less time typing. The judgment calls still take a person, they just happen earlier and more deliberately.
Should a small team skip this distinction to move faster? No. Even a one-person content team decides, formally or not, which parts of a piece get their direct attention. Team size changes who enforces the split, not whether the split exists.
