Nobody is an AI expert.
I’ll say it again for the people in the back: nobody. A handful of researchers at three or four labs who’ve been doing this since before it was cool. Everyone else is a fast learner with a newsletter subscription and varying degrees of honesty about it. That includes me. That includes the person who just sent you a resume with “AI Expert” in the headline and a certification that’s eight months old. And it absolutely includes the consultant who just quoted you $15,000 for an AI readiness assessment.
I’ve been in IT and EdTech for over 25 years. I have interviewed and procured more bullshit artists than I care to admit, and I want to tell you something: AI just graduated its first class of them. Cap, gown, LinkedIn banner update, the whole ceremony.
They’re not bad people. Some of them are genuinely smart. But they built an identity on a field that didn’t exist two years ago, priced that identity like it was scarce, and now they’re walking into institutions that are scared enough of missing the boat that they’ll pay for it. Hire them or rent them, doesn’t matter. The institution feels good either way because credentials are defensible. You brought in the certified person. You did the thing. If it goes sideways, at least you followed the process.
This is how organizations perform AI adoption instead of doing it.
The Oracle arrives, on payroll or on a statement of work, same difference. They say “large language model” seventeen times in the first meeting. They have opinions about every tool and deep experience with none of them at scale. They’ve never had to explain a system change to a skeptical department head who has forty other things on her plate, zero patience for tech theatre. They cannot tell you why the thing didn’t work. But they can tell you about the next thing. There’s always a next thing with these people.
Six months later you’ve got a digital dream board full of AI opportunities and an invoice that made your VP wince and nothing implemented.
Congratulations, you bought a very expensive roadmap from someone who won’t be around to execute it.
Meanwhile there’s a person down the hall who’s spent years making complicated systems talk to each other. They’ve shepherded projects nobody wanted to own. They think in systems. They know where the bodies are buried in your org chart. They’ve spent their career translating between what a business unit actually needs and what IT can actually build, and they’ve gotten good at it because the consequences of getting it wrong were real and landed on their desk.
They’ve been quietly poking at AI for a year on their own time, figuring out where it actually helps versus where it’s a parlour trick dressed up in a demo. They’ve already killed two AI pilots quietly because the juice wasn’t worth the squeeze. They were right both times.
That person will out-perform the Oracle in six months. Employee or consultant. I’d bet rum on it.
Because here’s the thing about rolling out AI or any new technology in an actual organization: the tech part is not the hard part. Learning the tools takes weeks. Genuinely understanding a workflow, earning the trust of the people in it, knowing which friction points are worth solving versus which ones exist for a reason nobody remembers, but you’ll find out the hard way, that takes years. You cannot certify your way into institutional knowledge. You cannot invoice your way out of not having it.
AI fluency is a multiplier. If there’s nothing underneath it you get nothing. If there’s years of domain knowledge and systems thinking underneath it, you get something that actually works and doesn’t require a maintenance contract.
The quality I’d screen for that nobody puts in a job posting or an RFP: comfortable being the least expert person in the room. Not as a concession. As a feature. The people who built their whole identity on knowing things are exactly the wrong fit for a space that changes every three months. You need someone who can be wrong on a Tuesday and pivot on Wednesday without having an existential crisis about their personal brand.
I’ve started calling this credentiflation. The certifications are breeding faster than the knowledge base is settling. They’ll be irrelevant before the ink dries. What doesn’t expire is the ability to learn in public, sit with not knowing, speak across silos, and figure out what’s actually true rather than what looks good in a slide deck or a proposal.
If you’re hiring or signing a contract: look at what’s underneath the AI keywords. What did this person actually do before AI was the thing? How do they talk about failure? Do they have opinions that cost them something, or just opinions that got them applause? Can they explain a complex system to someone who has no patience for complexity?
The Oracle is expensive, smells like a conference, and is already thinking about their next engagement.
The Analyst is three doors down, has already killed the bad ideas you haven’t had yet, and is quietly making things work.
Hire accordingly.
