When Drafts Are Free, Deletion is the Product

“Less is more” has never been truer.  AI has made scale and volume easier than ever. Unfortunately, both scale and volume are bad news when your strategy is off-kilter. I wrote a piece a couple weeks ago in favor of not publishing faster, and I stand by that.  In this piece, I want to talk about…

Ashley Poynter
Ashley Poynter
— Founder, Content Rewired
5–7 minutes
human in the loop

“Less is more” has never been truer. 

AI has made scale and volume easier than ever. Unfortunately, both scale and volume are bad news when your strategy is off-kilter. I wrote a piece a couple weeks ago in favor of not publishing faster, and I stand by that. 

In this piece, I want to talk about adding value by doing less and even deleting. 

It’s fun to point to all the “markers” of AI writing: the em dashes, the lists of three, the “it’s not x, it’s y” framework. I’ll defend the em dash to my last breath—it’s my favorite piece of punctuation, and I’m keeping it—so let’s set the surface tells aside. They were never the real one. Do you know what the biggest tell actually is? 

It says too much.

AI is great at producing. With the right prompt, a machine could produce a great 1,000-word piece on your topic of choice. But it can’t tell you whether or not you actually needed all 1,000 of those words. It will write it without “wondering” whether or not each sentence is the right, true thing to say to that reader in that moment. And there’s really no checklist or prompt to streamline that reality.

Drafting has become commoditized. Now, deletion is the value add.  

In nearly every instance I read about it, human-in-the-loop is described as a step in a process: a checkpoint where a person reviews what the model made and signs off. That description has the value upside down. The model’s output is the raw input, and the human’s judgment is the thing that matters. The model is doing a fast first pass on the human’s behalf, and the loop closes on the important part.

I should say plainly that I use these models all day. My practice runs on them. I am not writing this from a cabin in the woods with a fountain pen and a grudge. I reach for Claude the way I reach for a calculator, and I would no more give it up than I’d go back to writing in cursive, so this is not a purity argument. I advocate for using AI in ways it makes sense, but I also hold this truth close to heart: drafts are the new outline, and they require more finesse.

Discernment is Expensive

The thing the models genuinely solved is fluency. A confident, immaculately punctuated paragraph used to cost real time and a real person. Now? It costs a few seconds and a fraction of a cent. That is a true and large shift, and pretending otherwise is its own kind of dishonesty.

However, fluency was never the scarce resource. It only appeared that way because the people who had it were rare and slow. When you strip away the slowness, you find out what was actually valuable underneath: knowing when fluency is a cover for lack of substance.

A model will hand you ten plausible, very clean sentences. Picking the one that’s correct for this client, this audience, this paragraph (and seeing that the other nine are competent and dead) is not a downstream review task. Rather, it’s the entire skill. The model raised the floor on fluency to somewhere near the ceiling. Judgment, though—that’s the heart of everything. 

Taste Wades Past Shallow Fluency and Into Deeper Meaning

If you’ve logged onto LinkedIn, you’ve probably heard this bit a handful of times already: A model is built to give you the average of everything it has read. Et cetera. Et cetera. Everyone is familiar with the argument that we’re all converging toward mediocrity. And still, businesses are leaning into this production source to crank out content that is, at best, forgettable, and—at worst—non-compliant. 

A simple editorial pass can save you from the regulators, but taste is what elevates you in the eyes of your audience. Taste is what cuts through the crap and hoists the crux of your content up high where your audience can see it and latch on. Taste says “no.” Less is more. Taste has a point of view and the experience to back it. Taste is what draws a reader or viewer or listener into a story that is unmistakably human. 

When a client hires me, what they are buying (whether the contract says so or not) is the deletions. They’re paying for the eagle eye that slashes unnecessary paragraphs. They’re hiring my ability to recognize and synthesize nuance. That work is invisible by design, which is why it’s so easy to mistake for overhead and so dangerous to automate away.

The Model Can’t Feel the Room

Discernment and nuance are the same muscle aimed at different problems, and the problems share a property: the model can describe them and cannot feel them.

A confident claim reads as authority to a startup founder and as arrogance to a procurement lead at a state agency. A joke that lands in one deck reads as a flippant comment about risk in another. A statistic can be accurate but rhetorically dead on arrival because the reader has seen that same number in nine other posts this quarter. The model has read about all of this, and, in some statistical sense, “learned” it. But knowing that a difference exists and sensing which side of it you’re on are different acts, and only one of them delivers good, trustworthy work.

The model can confirm a sentence is grammatical and factually supported. It cannot tell you the sentence is going to make a careful finance buyer trust you less. It tries, to be sure. But the person who has been in the room with that finance buyer and shaken her hand doesn’t have to try. She knows.

The Cheaper the Draft, the More the Judgment Is Worth

The workflow framing assumes that as the machine gets better, the human’s slice gets smaller. Stop and look more closely. The economics actually run the other way.

When the draft was expensive, judgment was bundled invisibly inside the cost of writing, and clients paid for the bundle without ever pricing the parts. But now that the draft is nearly free, the bundle has come apart in front of everyone. Empty fluency has made discernment more expensive. 

AI writes drafts for me all day. I spend the rest of the week deciding what goes, what stays, and what’s even worth saying.

My clients don’t need a “pile” of homogenized content that their competitors also cranked out that morning. They need the strategy that dictates what buyers need and want to know—and substantive, worthy content that fits the bill. A human touch.

Want to Talk to Someone?

Contact us

When Drafts Are Free, Deletion is the Product – FAQ’s

Ashley Poynter
— About the author
Ashley Poynter

Founder of Content Rewired, a fintech content practice built on twenty years inside payments, treasury, and fintech SaaS. Previously head of content at PaymentWorks. Writes about editorial leadership, AI-enabled content production, and the discipline that separates B2B fintech marketing that compounds from the work that just publishes.

— Keep reading

More on this thread.

  • fintech content revenue engine

    FINTECH CONTENT MARKETING AGENCY

    The Fintech Content Revenue Engine: How to Attribute Content to Real Deals
  • micro-content for fintech

    FINTECH CONTENT MARKETING AGENCY

    Micro-Content in Fintech: How to Repurpose Long-Form into Revenue
  • fintech content operations

    FINTECH CONTENT MARKETING AGENCY

    From Chaos to Clarity: Building a Scalable Fintech Content Operations System
— Studio Notes

If this was useful, the newsletter is too.

One essay every other week. No roundups, no engagement bait. Just one substantive argument about fintech content marketing.

Book a intro call →