The Training Lie

Here's a pattern I've watched play out at 8 different health systems:

Phase 1: The Launch. Hospital invests $1.8M in a clinical AI platform. Launch goes fine. Initial adoption hits 34%.

Phase 2: The Plateau. Six months later, adoption drops to 19%. IT department panics. The vendor's response: "The users need more training."

Phase 3: The Training Investment. Hospital hires training consultants. Builds an LMS. Creates video tutorials. Schedules mandatory sessions. Total investment: $400K over 9 months.

Phase 4: The Disappointment. Post-training adoption rate: 22%. That's $133K per percentage point. For a 3-point lift that could be noise.

Phase 5: The Real Diagnosis. Desperate, they finally watch actual physicians use the system. Turns out:

• The AI recommendations appeared in a different tab than the patient chart
• Physicians had to click 7 times to see the reasoning behind each recommendation
• The interface didn't show what data the AI analyzed
• There was no way to document "I saw this but disagree because..."

No amount of training would fix those problems. Physicians weren't confused about HOW to use the system. They were refusing to use a system that didn't fit how they work.

Phase 6: The Interface Fix. Three months of UX redesign. $120K. Moved recommendations to context. Showed reasoning in one click. Made AI data visible. Added defensibility documentation.

Adoption: 61% within 4 months.

The math: $400K in training bought 3 percentage points. $120K in interface design bought 39 percentage points.

Training was the expensive distraction. The interface was the actual problem.

Here's the 5-minute diagnostic:

Question 1: Can users explain what the system does? If yes, it's not a comprehension problem. Training solves comprehension problems.

Question 2: Do power users have significantly better adoption? If power users and regular users have similar (bad) adoption, training won't help.

Question 3: Have users developed workarounds? If users are actively avoiding the system, they understand it well enough to work around it. You don't train away deliberate choices.

Question 4: What do users say when you ask why they don't use it?

• "I don't know how" → Possible training problem
• "It takes too long" → Interface problem
• "I don't trust it" → Trust architecture problem
• "It doesn't fit my workflow" → Workflow problem
• "I forget it exists" → Integration problem

Only one of those is a training problem.

Question 5: Does the interface force extra work without visible benefit? If using the system creates more work than not using it, no training will overcome that.

Here's the uncomfortable truth: training programs often exist because they're easier to propose than admitting the interface is broken.

Saying "we need more training" implies: The software is fine. Users just need to understand it better.

Saying "the interface is broken" implies: We approved a flawed design. Someone didn't do their job. We need to spend more money fixing it.

Which conversation do you think is easier to have in a steering committee?

Training becomes the path of least organizational resistance. It's not malicious. It's human nature. But it's expensive when the real problem goes unsolved.