Half Wild Labs Start With an Audit →
Coherence for everything you publish

Anyone can generate more. Someone has to decide what’s right.

Half Wild Labs is a brand engine studio in Washington, DC, founded by Kat Babic, a designer with twenty-five years of brand identity work behind her. The work starts with a simple one-week audit: every place your published work contradicts itself, found and ranked by what it costs you.

Start with a Drift Audit — $2,500 →
What Is Happening

In the age of AI. Your brand is now being made faster than anyone can check it.

Chapters, vendors, volunteers, and the tools they use are all publishing in your name. Each piece gets made from whatever was at hand that day: an old file, a half-remembered rule, a prompt. Nothing in that chain holds your standard unless somebody put it there.

Researchers writing in the journal Business Horizons gave the problem a name this year. They call it agentic brand drift: “the gradual, unintended divergence between a firm’s intended brand identity and the emergent brand character produced by its AI-orchestrated operations” (van Esch & Cui, 2026).

The finding that should bother you most is that there’s no triggering event. The drift happens while your numbers improve, and nothing tells you it started.

Which is why it takes an outside look. Someone still has to decide what right looks like for you: which blue is yours, which mismatches actually matter, and what the fix should be.

Twenty-five years of identity and production work for institutions where being slightly off was expensive.

Also: Council on Competitiveness · Arthritis Foundation · U.S. Department of Education · U.S. Postal Service · DynCorp International · ExxonMobil · Marriott

The Cause

Why this got worse so quickly.

Brand guidelines were written for a world where making things was expensive. Production cost money and time, so the number of decisions stayed small enough that a PDF could hold them all.

Production is now close to free. Every chapter, every volunteer, every vendor, and every assistant is making decisions in your name, and each of them works from whatever they were handed, not from what you decided.

The tools aren’t doing anything wrong. They’re doing exactly what was asked, with exactly what they were given. Point a generator at a body of work that already disagrees with itself, and it will learn the disagreement and reproduce it at volume, for as long as you let it.

Checking every piece on its way out the door stopped being possible a while ago. The standard has to live where the work gets made: in the files people open, and in a written record your tools can be pointed at.

Files that carry your standard, plus a record that states it, built to work as one system: that pair is what I call a brand engine. Building them, and keeping them true, is what this studio does.

The tools got faster, but nobody ever wrote down what right was.

The Symptom

You already know what this looks like.

Your website, your sub-sites, your decks, your forms, your signage, your social accounts, your regional offices. Dozens of people are making things in different rooms, at different times, from whatever file they found, and lately from whatever a tool produced on request.

Sooner or later it stops matching. The first person to notice is usually a customer.

If you run a network you don’t directly control (chapters, regions, affiliates, franchisees, exhibitors, volunteer committees), you already know you can’t staff a designer for every one of them. You also know a volunteer is never going to follow a PDF style guide. The new part is that you can’t make their tools follow it either. A style guide written for a person to read isn’t something a machine can use.

Built for associations with chapters, conference and trade-show programs, franchise and multi-location networks, and organizations six months past a rename.

Detection Is the Easy Half
The tools score compliance. Nobody supplies the taste they score against — and the mid-market has neither.

AI has gotten genuinely good at checking your work. Point it at your files and it will flag the wrong blue faster than any person could.

But AI only knows the blue is wrong against the value somebody gave it. It cannot know the value itself was wrong. It cannot choose between two versions that are both defensible. It cannot tell you which mismatches are costing you something and which cost you nothing, or what the repair should be.

Every one of those is a judgment call, and judgment is what you are hiring.

Big organizations pay a standing brand team to write the standard, then let AI enforce it. Everyone else turns AI loose on an old PDF and a folder of files that already disagree.

The Business Horizons researchers landed in the same place from the other direction. The major AI governance frameworks govern how a system behaves, not whether your meaning holds, so an organization can run them perfectly and still drift.

None of this is an argument against the tools. They’re about to become the most obedient brand managers you’ve ever had, and they’ll obey whatever you give them.

Says Who

I’ve spent twenty-five years on identity and production work, most of it for institutions where being slightly off was expensive: a national rail brand, a UN agency, a federal department, the governing body of a sport.

I’ve been a designer from the start, but along the way I also ran event expos. That meant recruiting exhibitors, fulfilling sponsor contracts, laying out the floor, and solving the power problem at six in the morning.

The expo years are where I learned which kind of wrong is expensive. A scanner can tell you two things are different. It can’t tell you this one costs you a sponsor and that one costs you nothing.

Your conference is not a design project to me. It’s the thing that pays for the year.

Capacity

One designer. One build at a time.

Half Wild Labs is one designer on purpose. I build one engine at a time, and while yours is on the desk it’s the only build in the studio. There’s no junior learning on your account, and no account manager standing between you and the person doing the work. The person you hire is the person who shows up, all three weeks.

The machines are how one person delivers at this scale. Production is machine work now, all the versions and resizes and volume, and I run it that way. The judgment calls stay with me: which version is right, which mismatch matters, what to fix first.

Machines take the volume. A person keeps the judgment.

One thing about the word engine. The engine is for the work, never for the client. It exists so your material comes out consistent across every chapter, every market, every location, not so you get processed.

I’ve watched what big agencies do to accounts this size. You get pitched by the partner and serviced by the intern, and somewhere in there you become a number in somebody’s machine. There’s no machine here to disappear into. There’s a person, and you have my number.

One at a time also means I’ll tell you the truth about timing. If the desk is full, you get an honest start date instead of a squeezed-in project.

The Offers
01

Drift Audit

One week · $2,500

I go through what you’ve already published and show you exactly where it stops matching. That means the site, the sub-sites, decks, forms, signage, and social. Every contradiction comes with a screenshot and a link, ranked by what it’s costing you, and there’s a bounded fix list at the end. No access needed, no meetings, no interviews. It costs you one email. The audit is diagnosis, and it stands alone. Whatever it finds is yours to act on, with me or without me.

02

Engine Build

Three weeks · $12,000 · One build at a time

The Drift Audit tells you what’s wrong. The Engine Build makes it stop. You get an identity designed by someone twenty-five years into this work, plus the files your people actually open: InDesign and Canva templates, PowerPoint and Google Slides masters, social sizes, letterhead, conference artwork, exhibitor and sponsor kits, signage. You also get the piece most brand systems leave out, which is the written record. Your color, type, and spacing get written down as values instead of described in a PDF, with voice, usage, and the do-not list in plain language. I call that file the decision record. The record is one file you can hand to whatever tool your team already runs, so it follows your standard instead of inventing one. The same record printed on one page is the version your people read. The engine is documented, it runs whether I’m in the room or not, and you keep all of it.

03

Standing Watch

Monthly · $1,500 · Capped at six accounts

Standards drift back, and they drift faster now that every department can publish at volume. So each month I check what’s changed across everything you publish and send you a short list of what needs fixing, while each item is still a five-minute job instead of a reprint. I can also be the one who fixes it. The retainer includes a standing block of studio time, and part of that time keeps your decision record current as your organization and your tools change. The cap of six accounts is on purpose. Six is few enough that I know your brand cold and you never have to explain it twice.

Also

Embedded Design

As needed · Hourly or retainer

When your in-house team is at capacity, I step in on your templates and your standards, with fast turnaround and no onboarding drag. I currently support the in‑house design team at a global investment management firm. Need extra hands? Email me →

Start With a Drift Audit

Tell me where to look, and I’ll show you what your published work is currently saying about you.

Request a Drift Audit →

A week later you get a short written report: every place your published material contradicts itself, ranked by what it’s costing you, with screenshots, links, and a bounded fix list. No access, no meetings, no interviews. It costs you one email. $2,500, and the report is yours to act on with me or without me.