Field notes on brand identity in the age of AI.
Four skincare brands launched the same eye patch jar in one week, and one meme decided who looked smart. The speed wasn't the skill. It was what the groundwork looked like from the outside.
Where would an hour of independent work do the most good in AI safety? A gap map, a test for my main hypothesis, and how I would choose a first project.
A person in the loop is not a safeguard by definition. The arithmetic of how much review it takes to catch a rare failure, and what a checkable oversight claim looks like.
I meet a version of this founder more often than you'd think: smart, capable, sitting on a genuinely good idea, still "just polishing a few things" before launch — for the second year running. Nothing's technically wrong. The logo isn't quite right yet. The website copy needs one more pass. The positioning will be finalized "once we've done a bit more research." Every individual delay sounds reasonable. Stacked together, they add up to a business that's been circling the runway for so long it's starting to forget why it wanted to fly in the first place.
Most of what I write about Brand Drift focuses on AI — a chatbot, an automated email, a tool nobody's supervising closely enough. But drift doesn't need artificial intelligence to happen. It happens just as easily, and sometimes far more damagingly, when the humans inside a company simply don't agree on who the company is supposed to be. That version of drift has a name too, in my research — I think of it as internal-external misalignment — and it's exactly what took down one of the most recognizable brands of the last decade.
Here's a test I run on my own clients constantly, and now I'm handing it to you: take your last ten social captions, customer emails, or website blurbs, strip off your logo and your name, and hand them to a stranger. Ask them to guess who wrote it. If they can't get anywhere close, you don't have a brand voice. You have words that happen to be near your logo.
One of the most useful questions I ask in a first strategy call has nothing to do with branding on the surface. I ask: "Tell me about the client who's actually easiest to work with, refers you the most, and pays without a fight." Not your ideal client on paper. The real one, currently, in your inbox. The answers surprise founders more often than you'd think — and every time it happens, I get a little thrill, because it usually means we just found the actual business hiding underneath the one they thought they were running.
I've turned away work. Not often, and never lightly, but enough times that I've stopped feeling guilty about it. Every time, it's for the same reason: the client wanted me to be a slightly different business than the one I actually am, and saying yes would have meant quietly becoming that different business, one compromise at a time, until Brand Bloom Lab stopped being Brand Bloom Lab.
Let's clear something up first, because the term gets misused constantly: a founder-led brand does not mean you have to post on LinkedIn every day, film yourself talking to a camera, or build a following. I say this as someone who is, by nature, more introvert than not — if "founder-led brand" required becoming a content creator, I'd have closed shop years ago and gone back to designing logos in the dark at 1am.
Safety tests built in English may not travel to Darija, Arabic or French. The evidence, what is missing, and a pre-registered pilot I would run.
Niche down to one specific person. No — broaden your appeal, or you'll cap your own growth. Be everywhere your customers are. No — pick one channel and go deep. Build a personal brand around yourself. No — keep the business separate from you entirely, for scalability.
Every case in this research shares one structural gap: no individual whose job clearly included catching the drift before it went public. Ending this series on the question underneath all the others — whose job was it, specifically?
Here's something that stings a little, so I'll just say it plainly, the way I said it to a founder recently who did not love hearing it: the business that wins the deal isn't always the better one. Often it's just the clearer one.
Okay, real talk. Almost every founder who lands in my inbox opens with some version of "I think we need a rebrand." I read it with my coffee in one hand and, let's be honest, probably a piece of chocolate in the other, and I already know what I'm going to find before I even click the link.
I want to tell you about a phrase I coined, because I'm a little proud of it and also because you need to know it exists before it happens to you: Brand Drift.
I designed my first logo in 2013. I was 18, I'd built a student community on Facebook, and I remember sitting there for hours moving a circle two pixels to the left, convinced that circle was the thing standing between me and success. I did not know what positioning was. I did not know what an audience was. I knew Photoshop, mild stubbornness, and how to make a circle look "professional."
A 2024 Nature paper showed AI models trained on their own outputs degrade in a specific way — rare, distinctive patterns vanish first. That's not just a technical curiosity. It's the same mechanism behind Brand Drift, running automatically.
If I had a dirham for every time a founder told me their business coach said "niche down until it hurts," I could probably retire and open a very small, very specific chocolate shop. Which, come to think of it, might be the only appropriate use of that advice.
I'm going to tell you about the project where the founder herself said, more or less, "I still can't believe the numbers" — and then I'm going to tell you why I actually believe them, because I watched it happen slowly, in a very unglamorous way, before it happened fast.
Aviation learns from near-misses. AI mostly learns from screenshots. What a near-miss report for deployers would need, sketched against three public chatbot cases.
Let's talk about the worst possible place to test whether your brand strategy actually works: a live room, in real time, in front of people who did not agree in advance to be nice to you.
I didn't start Brand Bloom Lab because I woke up one day wanting to run another marketing agency. I started it because I kept watching the same thing happen to business after business: more posts, more campaigns, more channels — and somehow, less clarity every single time. I wasn't trying to add to the noise. I was trying to solve the actual disconnect I kept seeing underneath it.
Most of the brands I work with are trying to get louder. More reach, more followers, more impressions, more everything. So it was genuinely refreshing — and honestly a little clarifying for how I think about branding in general — when I worked with JetPlus, a private aviation brand, and realized: the entire strategy had to be about being quieter, on purpose, and doing it so well that quiet itself became the pitch.
Nobody at Sports Illustrated sat in a meeting and decided to destroy 70 years of editorial trust. That's the part I keep coming back to when I tell this story.
Somewhere around brand number six, I remember thinking: this is either going to be the most interesting project of my year, or the moment I finally lose my mind. It turned out to be the first one, mostly.
A compute threshold is a tripwire for paperwork, not a verdict on risk. The arithmetic of why fixed lines age, and a layered design that expects to be revised.
A SOC 2 report and an ISO 42001 certificate can both be real, hard-won, and still never once ask whether your AI-generated content actually sounds like your brand. Compliance and governance are not the same claim.
A model passed its dangerous-capability test. What can you conclude? Four ways a pass misleads, and a seven-question rubric for reading one.
The Collingridge dilemma, two California bills, and what a decision record for rules made under uncertainty could look like.
I could have picked three comfortable countries that would confirm what I already believed about Brand Drift. Here's why I went looking for the one most likely to prove my own framework wrong instead.
Brand Drift doesn't stop at the customer. Amazon's hiring tool quietly taught itself to prefer men — and internal-facing drift like that runs longer undetected, because the people it fails never get far enough to complain.
Open weights, open source and the EU's exemptions are three different things. What the difference means for safety, and how I would structure a release decision.
Won't AI regulation eventually solve most of what I write about? I tested that hope honestly against the EU, the Gulf, and Estonia's actual regulatory philosophies. None of the three answers the question a founder is actually asking.
If a corporate brand can drift from its stated values through an unwatched automated system, can a country do the same to its own national brand? Australia's Robodebt scheme says yes, at a scale no company case comes close to.
Five of Kapferer's six identity facets get discussed constantly. The sixth, reflection, gets skimmed past — and AI image generators defaulting to a narrow stereotype of "customer" are making that an expensive gap to ignore.
Some brands catch drift early and recover cheaply. Some let it run for eighteen months and pay for it publicly. The difference isn't luck — it's a specific capability called sensing, and almost nobody builds it on purpose.
Sports Illustrated didn't write the fake bylines itself — a vendor called AdVon did, undisclosed, across dozens of publishers at once. Your brand is only as governed as your least-watched contractor.
I used to explain away working from Morocco before anyone took my research seriously. I don't anymore — building a framework outside the usual Western default turned out to be the point, not the caveat.
Across the drift cases I've studied, roughly eighteen months tends to pass between a brand's identity quietly slipping and someone outside the building noticing loudly enough to matter. Here's what shows up before that clock runs out.
AI ethics asks whether a system is fair to society at large — a question mostly answered by regulators, on a timeline measured in years. My research asks a narrower one a founder can actually check this Tuesday.
Is Brand Drift a real, structural thing, or just a story I've decided to tell about a pattern I noticed? Answering that honestly meant borrowing a philosophy of science built for something much bigger than branding.
Ask most businesses what AI governance means and you'll get "are we legally exposed." That's the wrong starting question — it protects the company and says nothing about whether the AI is actually good for the people on the other end of it.
Why does my research cover the UAE, Saudi Arabia, Morocco, and Estonia instead of picking one? Because I don't trust a definition of 'authentic brand' that only had to survive being tested in one culture.
Dubai and Riyadh are adopting AI faster than almost anywhere on earth. National strategy was never built to govern the twelve-person team automating its customer chatbot — which is exactly where Brand Drift compounds fastest.
Most AI governance advice lives at the enterprise-compliance altitude. The BIG Framework is three watch-points a ten-person team can actually run, on a schedule, with a name attached to each one.
A car dealership sold a Tahoe for a dollar. A delivery firm's chatbot wrote a poem wishing itself shut down. Neither was an AI problem — both were the same governance gap, just with the volume turned down for now.
CNET quietly published AI-written finance explainers under a vague byline, and more than half needed correction. That's the moment Brand Drift got its name — the silent, compounding gap between what a brand says it is and what it's actually producing.
Never miss a post