We use AI in almost every part of how we work. We're coming up on year three of building some of the tech behind it, and it hasn't been plug and play.

This series is us showing our work: where AI sits in how we make things, where it doesn't, and how we keep an eye on it.

"We Used AI" Doesn't Tell a Client Anything

For a while we called ourselves "AI driven." It sounded good. We knew what we meant by it, but nobody reading it did. It's like telling a client you used a human on their project. Well, what does that mean? It doesn't mean anything.

What actually helps is knowing which part of the work AI touched and how it was used there. Is it checking someone's work? Proofing? Running a compliance pass? Or writing the whole thing? Those are very different.

Clients Are Right to Ask

I don't blame anyone for being skeptical. It's a technology a lot of people aren't familiar with yet, and most of what you hear about AI right now is the slop and the poor work. Some clients aren't comfortable with AI in their work at all, and that's fair.

If you hand that person a blanket statement, of course they're going to be uncomfortable. That's why we'd rather show how we're using it, so they can see it, understand it and decide for themselves.

Why We Don't Trust AI Blindly

AI makes mistakes. On our back-end builds, we've had it skip a step it was told to follow, or tell us a process doesn't exist when it's right there in the instructions.

So we don't just take its word for it. We've built checks around it: compliance reviews built into the process, plus human experts reviewing the work. When something gets flagged, we go back and look at how it happened.

It's a learning process. Every miss is a chance to tighten things up.

Where People Lead and Where AI Helps

We have two kinds of copy. Creative copy, like the core messaging, is crafted by a human, then run through AI for review and compliance.

Day to day copy, like social posts, usually starts with AI writing from a library we've built for each client. Then a human reviews and revises it.

It's not a formula. It comes down to what the job requires.

Why Now

Watermarking is coming. What does that mean for us, for our clients, for their content and for our process? We want to spell out how we work, start to finish, so clients know what to expect when it gets here.

We've seen how much work it takes to get quality out of AI and keep the guardrails in place. It's a big lift, and we're glad to show it.

See How We Work

L2 Vantage is how we put this into practice: guardrails on what AI can touch, a record of what it actually did, and a trail back to the source when something needs a second look. See how L2 Vantage works.

Want to walk through it for your own work? Talk to us.

Next in the Series

Part two gets into what an AI disclosure actually needs to say to be useful.