We Were Good Before AI. We Got Better With It.

Someone recently told me my follow-up emails sounded like AI. It stuck with me, though not for the reason you’d think. It got me thinking about how little most people understand about what an AI-enabled firm actually is, and how easy it is to mistake the tool for the work.

So let me tell you what fifteen years of this actually looks like, and where the machines come in.

Your writing sounding “too polished” might mean the bar just moved — not that you outsourced your brain.

The bitter joke nobody talks about

I was a journalism major. I’ve been writing for a living for thirty years. My prose is polished because I was trained to make it that way, in newsrooms, on deadline, by editors who’d cover a lazy sentence in red ink. The em dash, the well-placed “moreover,” the contrarian turn that makes you stop and reread, the precise word where a vague one would do. I earned alllllllll of that.

Here’s the joke. The models were trained on writing like mine. So the better you write, the more you sound like the machine that learned from good writers. I now actively have to strip the craft out of my own sentences, the dashes, the words I love, the rhythms thirty years gave me, so a busy reader doesn’t mistake practice for a prompt. I will never forgive them for “delve” — I LOVED that word.

We even built it into policy. We have an internal style guide at the agency, and one of its rules is to pull the em dash out of our writing. Not because it’s wrong. Because it now reads as a machine to people who don’t know better, and we’d rather lose the punctuation than lose the reader. There’s a document at my firm that exists specifically to remove some harbingers of good writing so people don’t think my clients are lazy gits.

That is BASSACKWARDS. I now have to put typos INTO stuff so people know I wrote it (that is actually pretty easy since my nails are super long and use TTT alllllll the time). And the people doing the flagging are usually pattern-matching on surface features, a dash here, a tricolon there, because they don’t have the ear to evaluate the thinking underneath. They’re judging the cover because they never really read the book. Now good writing is like having your signature in Papyrus. It just gives the ick.

The part that came first

And while I hate to toot my own horn (JK I love it mostly, YAY menopause freeing me from the shackles of false modesty) we’re good at AI because we spent all the years BEFORE AI, creating templates, studying best practices, adapting to them when they changed, building workflows that looked like a 6th grader’s Rube Goldberg machine.

Two hundred clients’ worth of pattern recognition is a hell of a drug. Watching what works and what crashes and burns is both humbling and edifying and well we’ve done a lot of the former and our fair share of the latter.

I have analysts I’ve spent almost two decades alongside, earning, learning, becoming friends. When I wonder aloud over ribs at 2 am, if a new category is worth launching, I ask because I am curious and they answer because we love this nerdy shit!

We had/HAVE templates and processes and onboarding forms and a competitive intelligence practice, all built by hand, all refined over thousands of hours of doing the work and getting it wrong and doing it better. We know where the bodies are buried in HR tech because well, we’ve got a room of shovels.

None of that comes from a model. All of it predates the tools by years. That foundation is the thing AI sits on top of. Take the AI away tomorrow and we are still one of the longest-running firms in our category. That matters, because it’s the part nobody can replicate by buying a subscription. You can see some of that in our case studies.

Then AI showed up, and we didn’t panic

Idealist that I am, when AI showed up I thought EUREKA! Yay work will be forever transformed. Well, I was half right. Anyway, we were super early adopters, like Pre ChatGPT adopters. We’ve been working in AI for more than five years now. Five years of building it into how we operate, testing it, breaking it, writing policy around it, and throwing out the parts that didn’t hold up.

So why I do admit, I got lazy with the 2-3 line follow up emails, they aren’t automated. They’re me, speaking to text in Gemini so it cleans it up, removes cussing, and clarifies my mad scientist brain. But before that…

When a prospect comes in, I pull everything public on the company and start mapping where I think the real gaps are. Then I poke holes in my own read, against what I know, what my analyst friends think, what I see in the market, and what shows up in our proprietary view of the landscape. The tools research, gather, surface, sort, rank, and MOST importantly for me, ORGANIZE. But the judgment is mine. AI is predictive, not strategic. It’s very good at telling me what usually comes next. It has no idea what should.

I saw this really good cartoon that showed history’s great thinkers running their ideas by ChatGPT (Galileo, Darwin, Einstein…no women because well lezbehonest, they’re not always IN the history books but I digress.) and because AI is filled to the brim with what we ALREADY KNOW, and what the MAJORITY of published works AGREE ON on that subject… you see the problem?

Okay, where was I, oh yeah, THEN I run our calls through a custom transcript processor I built to catch the useful phrases, small and large tasks and todos, insights and between the lines moments, specific ideas I had in the moment, so nothing gets lost while I’m actually listening to the person in front of me. As time goes on, we add all our insights, client and company observations, insights, etc into a robust profile of how to work best with them. Every prospect gets measured against our competitive landscape. Every piece of it gets pulled together from a dozen scattered places into something coherent.

I also have long nails and carpal tunnel, so I quite literally talk to my computer all day (double tap CMD FTW!). The tools turn my scattered, spoken, twenty-years-of-opinions thinking into something organized. They do not have the opinions. I do. OBVS

Being “AI-enabled” doesn’t mean the humans left the building. It means the humans finally stopped drowning in busywork.

The human in the loop is not a slogan

This is the part people who sneer at “AI-enabled” never bother to check. Like, no one gets ticked off when the sniper needs a scope.

When I read something the tools helped me assemble and I notice a point I care about didn’t make it in, I stop. I go disprove it. Not only that, I dig into why something like that would come up. I pressure-test it against real clients, real analysts, and people I’ve known in this industry for twenty years. I have filters built specifically to keep the machine from blowing smoke at me. I push back on it constantly. And you wanna know what? That dogged process has saved a lot of clients a lot of money and bubbled so many opps we never would have thought to look for. Because AI isn’t always wrong or hallucinating, sometimes it’s showing you how it SEES you/the client/company/vendor/etc in the digital world. And like it or not, that is something worth knowing.

And we QA the whole thing on a constant loop, because the ‘trick’ isn’t any single tool. It’s AOS, our agent operating system, and we refresh it constantly. The strange and wonderful result is that we know our clients better than we ever did before. The tools didn’t put distance between us and the work. They cleared away the busywork that used to keep us from it.

So about that email

It probably did read flat. Follow-up emails are the one part of this I treat as logistics, because I am almost impatient to get to the real work. The note that schedules the call is not where I put twenty years of craft. The strategy is.

That’s the thing worth understanding before you write off anyone for “sounding like AI.” A polished sentence is not evidence of anything. A referral IS.

The presence of an em dash isn’t proof you should pounce on like some modern day Hercules Poirot AHA! But a verified case study (or upwards of 50 of them) is proof worthy of further investigation.

This contrarian sentence I am writing right now, despite my long ass nails and carpal tunnel is not evidence of an LLM and a reason to go to the next name on the list. Because I AM THE WINE LIST. IYKYK.

AI can produce a sentence that sounds right, but is it? I mean, my hair LOOKS brown…but it’s not really (it’s gray and no I am not ready to go silver yet, although I cheer on those who are ready, I am just not.)

We’re honest about our tools and rigorous about our thinking.

Judge the thinking. That’s the work. The rest is just the keyboard, and these days, I’m barely even using that.