The consensus has arrived, and it's suffocating. Walk into any conference on educational technology, scroll through any L&D department's strategic memo, read any think tank report on the future of certification: artificial intelligence is coming to learning, and that's a settled fact. We've moved past the debate about whether. Now we discuss only how.
This unanimous agreement troubles me. Not because AI won't transform education and training. It probably will. But because consensus this smooth usually means we're all looking at the same problem through the same window, missing what's actually breaking on the other side of the wall.
Consider what we're collectively NOT arguing about. We're not questioning whether the rush to implement AI tools might systematically advantage learners who already have advantages. We're not wrestling seriously with how rapidly deploying AI in certification systems could calcify inequities under the appearance of objectivity. We're not even having honest conversations about whether "overcoming resistance" to AI tools might sometimes mean overcoming legitimate caution from people closer to the actual stakes.
The comfortable narrative goes like this: AI personalizes learning, accelerates skill development, reduces costs, and democratizes access. These things might all be true. But they're not neutral truths that emerge from the technology itself. They're outcomes that depend entirely on choices we haven't yet made.
A manufacturing company using AI to accelerate reskilling for workers displaced by automation is solving one problem. A certification body using the same technology to reduce oversight while maintaining the same pass rates is solving a completely different one, with very different consequences for the person holding that credential.
The consensus hasn't asked: What gets faster at the expense of something else slowing down? When we automate parts of the learning process, what human judgment disappears with it? Who benefits from a certification that's quicker to obtain but harder to interpret? These aren't anti-AI questions. They're the questions that should precede any serious implementation.
Here's what concerns me more: the consensus is already baked into investment decisions, into vendor relationships, into job descriptions for learning professionals. By the time serious doubts emerge, the infrastructure is set. Changing course becomes expensive and embarrassing.
We've seen this before. Online learning was going to democratize education. In many ways it did. But it also created new moats for wealthy institutions while underfunding public alternatives. Nobody intended that outcome. The consensus just didn't account for what it would actually cost institutions to do online learning well, or how that cost would be distributed across different types of schools.
The better question isn't whether AI will transform learning. The better question is: what assumptions baked into our current rush will we regret? What tradeoffs are we not naming? Whose voice is absent from this conversation about "overcoming resistance"?
I'm not arguing for paralysis. But I am arguing that an opinion column written by someone in my position should make you slightly uncomfortable about how comfortable everyone else seems to be.
The actual hard work ahead isn't implementing AI tools. It's holding onto skepticism long enough to ask whether the consensus benefits the people we're actually trying to reach, or mainly the people selling the solutions.
That conversation hasn't really started yet. And that's what should worry us.