Part 2 of 2
Most of what lived in the Cambrian explosion left no descendants. It left a name instead — pinned to a slab of shale, unpronounceable, discovered by someone half a billion years later who had to guess which end was the head.
Take Hallucigenia. For the better part of a century, palaeontologists had it upside down — the spines along its back read as legs, the legs read as tentacles, the whole animal mounted the wrong way up in every reconstruction. It wasn’t a bad guess. There was no living relative to check the drawing against. The creature had simply vanished, taking the instructions for how to read it along with it.
Or Opabinia. Five eyes, and a mouth on the end of a hose, with a claw at the tip. When it was first properly reconstructed and shown to a room of palaeontologists in 1972, the room laughed. Not unkindly — it just looked like something a committee would design, not something evolution would actually build. And yet there it was, fossilised, having once been somebody’s whole functioning body plan.
We are, right now, living through the data and AI profession’s own Cambrian explosion. Forty-odd job titles, by one recent count, describing perhaps three actual jobs. Prompt Engineer. Context Engineer. AI-Native Developer. Vibe Coder. Verification Engineer. Whatever gets minted between me writing this sentence and you reading it. Most of these will leave no descendants. A few will turn out to have hit on a structure the future actually needs — and the rest of this piece is about how you can already tell which.
There’s a name for this pattern, and it isn’t just chaos. Punctuated equilibrium — Niles Eldredge and Stephen Jay Gould’s 1972 revision of how evolution actually moves — says species don’t drift gradually. They sit in long stasis, then burst into rapid diversification for a short, violent window, then settle back into stasis, with a handful of the new forms fixed for the long run and the rest gone without issue. The burst was never the interesting part. What gets fixed when it ends is.
The Burgess Shale took roughly twenty million years to lay down and settle. Ours is compressing into a window you could staff a project around — eighteen months, if the last cycle is any guide. Which means the useful question stops being “what strange new shapes will appear” and becomes “what, of all this, is already being pressed into the rock, while we’re still too close to see it.”
Here’s the part nobody puts in the recruitment deck: you are, for however long your current title holds, a body plan. Database Administrator-shaped, then Business Intelligence Analyst-shaped, then whatever Data-Something-shaped came after, then whatever I’m apparently called this year. You did not choose the shape any more than Opabinia chose the claw on the hose. Somebody upstream decided what the organisation needed identified, and you grew into fitting the description, because that’s what bodies under selection pressure do.
Nobody warns the trilobite.
What fossilises is never the part that mattered while it was alive. Shells, carapaces, the hard load-bearing bits — those press into stone and last. The soft tissue that did the actual sensing and deciding is gone within days. Which is a fairly exact description of what a job title preserves about you. The title is the carapace. It’s what HR can point to, what a slide can cite, what an accountability chain eventually attaches itself to, if it attaches to anything at all. That’s the drift, at cellular scale: not a strategy quietly rotting in a steering committee, but a body building itself around a name that was never really describing it in the first place.
But here’s where it stops being metaphor. Three weeks ago — 2 August 2026 — Article 14 of the EU AI Act became enforceable. It requires anyone deploying a high-risk AI system to name actual, named people accountable for overseeing it: trained, authorised to override it, competent to recognise when they should. It does not tell organisations what to call these people. That’s sediment forming without a settled genus yet attached — real pressure, real weight, no fossil name.
The IAPP, the body that turned “Data Protection Officer” from a GDPR clause into an actual career via the CIPP and CIPM certifications, launched an AI Governance Professional certification in March 2024. Over four thousand people had enrolled in the training before the exam even went live. Certification bodies don’t build exam infrastructure for a shape they expect to go extinct. That’s a palaeontologist deciding a genus is real enough to catalogue.
Then the strangest finding of all: most of what survives won’t be new. Banking has run a formally regulated model risk function since 2011 — the Federal Reserve’s SR 11-7, hardened by 2014 into a proper three-lines-of-defence structure. Rather than a fresh “AI Risk Officer” displacing it, that existing apparatus simply widened its jaw: regulators extended the same guidance to cover machine learning models from 2022 onward. No new species evolved. An old one moved into a niche that had opened up next door.
ISO/IEC 42001, the new international standard for managing AI systems, does the same thing on paper. Its reference architecture — the checklist every auditor and advisory firm will now measure a company’s AI governance against — names four accountable role categories, and the striking thing about the list is how few of them are new. AI risk officer. AI ethics officer. Model validators. And, doing most of the actual work, data steward — a title that’s sat quietly inside data governance teams for the better part of fifteen years, now simply handed a bigger diet.
Which tells you something the Cambrian framing almost hides: the winners of a mass proliferation event aren’t usually the strangest new arrivals. They’re whichever existing organism already had the right jaw when a new food source appeared. Opabinia‘s claw-on-a-hose was a marvel, and it left nothing behind. The trilobite’s dull, over-engineered exoskeleton is the reason we have three hundred million years of trilobites.
So if you’re trying to work out what you’ll be called in five years, that’s the wrong question — Part I told you as much, thirty years ago onward. The better one is whether you’re already doing the dull, load-bearing thing some regulation or standard is about to widen the jaw around. Verification. Oversight. Stewardship. Whatever it gets called afterwards is decoration.
Old jaws, new niche.
Suggested category: Data and Artificial Intelligence (secondary: Entropy Management)
Note on mode: Speculative/Philosophical Essay, extended metaphor (Cambrian explosion → punctuated equilibrium) now carrying real, checkable evidence rather than pure image. Facts and dates to verify before publishing: EU AI Act Article 14 human oversight requirement, enforceable 2 August 2026; IAPP’s AI Governance Professional (AIGP) certification, launched March 2024, 4,000+ enrolled at launch; Federal Reserve SR 11-7 (2011) and its extension via FHFA AI/ML guidance (2022); ISO/IEC 42001’s four AI governance role categories (compliance, strategic, implementation, operational). No Sources: line in the body itself, consistent with the speculative-essay house style — happy to convert these into inline citations if you’d rather it read closer to Argument Essay sourcing before it goes up.
Leave a Reply