When something goes properly wrong at work, how often is it because nobody knew the rule?
I ask because I spent years in learning and capability, mostly on the account side, which meant I heard a lot of these stories second hand. And I'm struggling to think of one that fits. What I remember instead are situations the rule didn't quite cover, where somebody had to make a call and made the wrong one.
I think that's the gap. Training teaches people what the rules say. Real incidents tend to be the case the rules didn't anticipate. Sitting between the two is judgement, and judgement is the thing I'm fairly sure we almost never design for.
Every module has a second curriculum
Here's what I mean by teaching two things, and I'll admit I'm still working out how far to push this.
The stated curriculum is the content. The policy, the procedure, the four warning signs, the escalation path. It's in the storyboard, it's signed off by the subject matter expert, and it's what the assessment measures.
The unstated curriculum is the habit of mind the module trains through its shape. Not what people learned. How they learned to approach it. Educationalists have a name for something similar in schools, the hidden curriculum, and I think the corporate version is just as real and rather less examined.
Here's where it gets uncomfortable. A module you can pass by pattern-matching the longest answer option has a very clear unstated curriculum. It teaches that every problem has one right answer, that somebody more senior has already worked it out, and that your job is to spot it.
I've started calling it the compliance reflex, mostly for want of a better name. Reach for the approved answer, and if you can't find one, wait until somebody hands you one.
If I'm right about that, it's a strange thing to be installing, because it's precisely the instinct that fails people in the situation the policy didn't anticipate.
So teach critical thinking? I don't think it's that simple
The obvious response is to add critical thinking to the curriculum. Buy the module, load it into the LMS, tick a new box.
The research is unkind to that idea, and I found this genuinely deflating when I first read it.
Daniel Willingham's work is the clearest statement of the problem. Critical thinking, he argues, "does not have certain characteristics normally associated with skills", in particular being able to use it at any time. Thinking is tangled up with the knowledge being thought about. You can teach people maxims about how to reason, but without domain knowledge and repeated practice they generally can't apply them, and transfer between subjects is far weaker than most of us assume.
A 2015 meta-analysis in the Review of Educational Research pooled 341 effect sizes and found critical thinking instruction does work, though modestly, at a mean effect of 0.30. What lifted the results were three specific conditions. Dialogue. Authentic, situated problems. Mentoring.
Read that list again, then picture the average compliance module. No dialogue. Invented scenarios with tidy answers. Nobody to ask.
Which brings me to a conclusion I keep landing on and don't much enjoy. If these skills only develop inside real subject matter, then the only place we can build them is the domain training itself. The very training the industry has spent twenty years making shorter and easier. I don't think there's a separate course that fixes this. I think there's only the module you already have, made harder.
Habit seems to beat knowledge, and there's a study I love about this
In 2019, Sam Wineburg and Sarah McGrew published an experiment I think should be required reading in our industry. Three groups were given the same task, evaluating live websites on contested topics while thinking aloud. Ten PhD historians, twenty-five Stanford undergraduates, and ten professional fact checkers.
The historians lost, and not narrowly. They read carefully and stayed on the page, weighing up the logos, the domain name and the tone. Several were taken in by sites built to look official.
The fact checkers left almost immediately. They opened new tabs and worked out who was behind the site before reading a word of it. Wineburg calls this lateral reading. They reached better-founded conclusions in a fraction of the time.
The historians weren't short of intelligence or expertise. They were about as credentialled a group of readers as you could assemble. What they lacked was a habit. The fact checkers had practised one specific move often enough that it had become automatic.
I think that's the whole secondary skills argument in a single experiment. Knowing things isn't the same as knowing what to do when you're not sure. The second one looks like a trained reflex, and reflexes come from repetition rather than exposure.
Why I think this matters more this year than last
If the value of your training sat in transferring content, I'd suggest that value is falling quickly. Any policy, procedure or definition an employee needs can now be summarised on demand by a tool costing a few dollars a month. Retrieval has been commoditised.
What hasn't been commoditised is knowing which question to ask, whether the answer you got is right, and what to do when two credible sources disagree. Those are the secondary skills, and I think they've just gone from a pleasant by-product of good training to the main event.
The risk, as I see it, is that we move in the opposite direction. Microsoft Research and Carnegie Mellon surveyed 319 knowledge workers in 2025 and found that the more confidence people placed in an AI tool, the less critical thinking they applied to the task. Confidence in the tool quietly replaced scrutiny of the output.
Put that beside the compliance reflex and you have what I'd call a poor combination. A workforce trained to expect one approved answer, now holding a machine that produces one instantly, in a confident tone, every single time.
What I'd actually do about it
I'm not arguing for difficulty as a virtue in itself, and I don't think every topic needs this. A hand hygiene refresher doesn't need to develop anyone's judgement. It needs to be short and correct.
But where the point is capability rather than notification, I think three design choices change the second curriculum.
Give people problems the content doesn't fully answer. If the module stated the answer verbatim three screens earlier, you're assessing reading rather than thinking.
Make them go and find something. One question that requires opening the actual policy, standard or register teaches a research habit that probably outlasts the module by years.
Allow more than one defensible answer. Ambiguity is where judgement lives. It's also where most real work happens.
None of that costs more to build than what we currently produce. It costs more to think about, which is a different problem, and an honest one.
Where I've landed
Every module you ship teaches a habit, whether anybody designed one or not. I'm fairly sure most of ours currently teach people to look for the answer somebody else already approved.
Which is a strange thing for compliance training to teach, when compliance failures are almost never caused by people not knowing the rule. They're caused by people meeting the situation the rule didn't cover, and not being in the habit of working it out.
So if you only change one thing in the next module you commission, I'd make it this. Put in one question the content doesn't answer, and see what comes back.