I often say I treat AI like an overzealous intern who thinks he knows everything but is inexhaustibly available to make changes. I have never fired an intern for a bad first draft. I told him what a finished one looks like, and I told him I’d be checking.

So the adaptation senior leaders need to make is the same one. Set the expectations for AI output and then hold people to them. In my own use that means five to ten iterations before anything is dialed in, a source for any claim I did not put in myself, and a rewrite of most of it so it reads like me and not like the machine. That last part is where the tells matter, because they are what a reader now uses to decide whether to trust you.

The tells, from the list I keep for my own writing:

Structure

• Sections that open by announcing themselves (”When it comes to pricing...”) and close with a summary line restating what you just read.

• Lists padded to three or five items. Triads show up everywhere: “clear, concise, and compelling.”

• Headers, bold, and bullets on content that should be two paragraphs.

• Symmetry. Every section gets the same depth whether it matters or not.

• Sentence fragments for drama, used more than once a piece. “Every. Single. Time.”

Sentence patterns

• Contrast framing: “It’s not just X, it’s Y.”

• Payoff lines at the end of a paragraph that restate the point as a slogan.

• Rhetorical-question pivots: “So what does this mean for your team?”

• Throat clearing before the point: “It’s worth noting,” “It’s important to remember,” “Let’s dive in.”

• Parallel openers in a run: “It’s about trust. It’s about speed. It’s about people.”

• Transition stacks: moreover, furthermore, additionally, ultimately.

• Em-dashes all over the place, often two or three per paragraph.

• Hedges stacked on hedges: “can potentially help,” “may often.”

Vocabulary

• The usual word list: delve, robust, seamless, leverage, elevate, unlock, landscape, tapestry, testament, pivotal, crucial, realm, foster, navigate, “in today’s fast-paced world.”

• Intensifiers that add nothing: truly, genuinely, incredibly, very, really.

• Abstract nouns where a specific one belongs: “drive meaningful outcomes” instead of “cut CAC 22%.”

Content

• No specifics. No names, numbers, dates, or firsthand details, and no claim the writer could be wrong about.

• No position. Both sides get presented and nothing is committed to.

• Confident claims with no source, or sources that don’t exist.

• Fabricated precision. A statistic with no source, usually a suspiciously specific one like 73 percent.

• The analogy that arrives from nowhere. “Think of it like a symphony,” on a piece about pipeline.

• Nothing only this writer could know. Swap in another company’s name and the piece still works.

Tone

• Relentless positivity and an eager, helpful register. It opens by praising the question (”Great question!”) and closes with a menu of offers.

• The “whether you’re a founder or a Fortune 500 CMO” opener that addresses everyone and therefore nobody, and the “What’s your take?” close that begs for engagement.

• The emotional temperature never changes, including on topics where a human would get annoyed or funny.

Mechanical giveaways

• Leftover prompt residue, like “Certainly! Here’s a LinkedIn post...” or “[Insert name].”

• Curly quotes and apostrophes mixed with straight ones, a sign the text was pasted from somewhere else.

• Title Case Headers on content that doesn’t need headers.

If your team’s work has more than a few of these, you have an expectations problem. It’s just interns being interns.

I ran this through an AI detector and it did not love the list. Fair enough, it was written with one:)

Subscribe now

So when custom GPTs started I was right on board, and several of those early ones still sit in my library. Claude projects have since become my poison of choice. They are how a one-man practice turns around value as fast as clients expect.

As an avid user, I noticed something happening across my projects. They were getting slower, less accurate, and I was doing way more corrections and deeper rewrites than before. A few times I gave up and did it old school because I could do it better on my own.

How, in a world where new models are supposed to be taking over and ending humanity in a few years, could my AI be getting Dumb and Dumber? So I dug in.

With models changing this often, and some of us jumping between tiers with different access, we are all running a hodgepodge of the latest and greatest. The tools encourage it. Retiring old models, giving away free use of newer ones for short periods. When you build a project or a GPT, you build it around whatever model you chose at the time. If that project lives for a while, you switch, maybe a few times. Some of mine were built to be evergreen.

So the old project is giving the new model instructions in the old way which is total something I would assume a new model would fix, apparently not. The new model follows instructions more literally, so every workaround I wrote to force the old one to do things the way I wanted now gets obeyed to the letter and fights defaults that might be better than my original rules. The instruction files and reference docs grew for months to cover things the old model could not figure out on its own, and all of that now sits in context on every turn, which is where some of the slow comes from. And the examples I pasted in to show it what good looked like are now a ceiling, because the new model might handle everything in a new way.

The engine got “smarter” but it could not adapt to scaffolding I built around the old one aged. If the project is old enough and complex enough, the thing that used to be kick ass at its job is now slow and less accurate than the day you tuned it, and using the latest, greatest and more expensive model is making it worse.

Kind of a bummer this isn’t just handled for you. The fix- I spent a few hours last weekend retuning three or four of my worst offenders, mostly by deleting, a little bit of rewriting and they are humming again. A new step for projects that are more than a generation or two old. A small price for all this AI wonder, but I am not so sure the AI apocalypse is five years out anymore. Maybe we will get ten before Skynet becomes self-aware.

An appropriate place to start:

Career pivots and the value of “yes, and.”

I have enough grey in my beard to have seen my fair share. For myself, for friends, for family. Some are voluntary, like the early days when you’re trying new things, or deciding to get an MBA with a two- and four-year-old at home. Highly recommend you think twice on the latter. MBA great. Two little ones add a 5x difficulty factor.

Some are involuntary, like so many people in my network recently. And some land in the middle. You read the tea leaves and jump before anyone asks you to.

What I’ve found is that no matter how you get there, the golden rule of improv works. Take what comes at you as a yes, then “and” your next move on top. That’s how I’ve found every next thing I’ve done. Sometimes it puts you in places you never thought possible or never even considered.

April was a yes. After three-plus years as CMO at Case Status, a job and a group of people I love, I moved on. It was one of my favorite stops in a wonderful and unconventional career, at least most of the time. We built the marketing team and the GTM process from scratch, scaled from $2M to over $10M in ARR, and made the Inc. 5000 three times. Then the company entered a new chapter, and it was time.

Roman Growth Advisors was the and. Fractional CMO work for B2B SaaS companies where growth is happening but still feels fragile. Six months in, I have run three client engagements, learned that prospecting for yourself is harder than any pipeline you ever built for someone else, and handed more of the work to AI than I would have admitted to or been comfortable with in April. The learnings and the speed have been life-changing, but that’s for future installments.

Sometimes things move quicker than we expect, and on October 1 there is another “and.” I am partnering into something bigger than a one-man practice, and I will tell you what in a couple of weeks when they let me. The fractional work continues, with a bench behind me.

Call it faith in the universe, God, or luck. The yes is usually decided for you, but the “and” is all yours.