AI is being baked into the learning stack at breakneck speed. We’re seeing it everywhere: personalized paths, automated assessments, “smart” recommendations.
But there’s a massive hidden risk here. Because we’re used to traditional software being deterministic (2+2 always equals 4), we start trusting AI the same way. But AI is probabilistic. It’s a “vibe” machine. It’s guessing based on patterns, and it isn’t consistently accurate.
If your team starts trusting the output without understanding the why, you aren’t building a smarter workforce—you’re building a dependent one. To fix this, leaders need a simple, monthly ritual to keep the “human-in-the-loop.”
1. Show the Receipts
Don’t let your AI be a black box. If the system suggests a specific course, a new career path, or a performance assessment, it needs to be able to “show its work.”
- What data is this recommendation based on?
- What is the model actually optimizing for (Efficiency? Engagement? Safety?)?
- Who in this building actually owns the logic behind it?
If you can’t answer these, you don’t have a learning strategy; you have a vendor’s algorithm running your talent development.
2. The “Human Sanity Check”
Because the same AI input won’t always produce the same output, we have to stop treating recommendations as “the truth.”
Make it a standard operating procedure to sanity-check what the AI is spitting out before acting on it. If a “personalized path” looks weird or ignores a learner’s actual experience, a human needs to have the authority—and the habit—to override it. We need to build a culture where “The AI said so” is never considered a valid excuse for a bad decision.
3. Kill the Vanity Metrics
In the old world of L&D, we measured “certificates issued” or “completion rates.” Those are vanity metrics.
AI makes it even easier to generate more certificates, but that doesn’t mean anyone is getting smarter. We need to look downstream. Don’t ask how many people finished the module; ask what changed in their actual work. Are they making fewer errors? Are they resolving tickets faster? Are they staying with the company longer? If the “personalized learning” isn’t moving an operational needle, it’s just noise.
4. Teach People to Interrogate the System
One of the most important skills we can teach right now isn’t “how to use AI”—it’s how to question it. When a learner gets confused or receives a recommendation they don’t understand, the system should provide “cues” for the next question they should ask. We want learners to stay in the driver’s seat. If they just follow the prompts like a GPS, they aren’t learning; they’re just following directions. Real learning happens when a person can see how the tool thinks, spot where it’s shaky, and know exactly when they need to step in and take over.
Why This Matters
This ritual isn’t about slowing things down; it’s about building actual skill rather than blind dependence. People learn faster and work better when they understand the tools they’re using. By exposing the “mechanics” of the AI, you earn the trust of your team and ensure that the technology is actually supporting human judgment, not replacing it.
Author Bio
Vin Mitty, PhD, is a data and AI leader with over 15 years of experience helping organizations move from analytics ambition to real business impact. He advises executives on AI adoption and decision-making, is an AI in Education Advocate, and hosts the Data Democracy podcast. As the Senior Director of Data Science and AI at LegalShield he leads their enterprise-scale AI and machine learning initiatives.
