Problems with AI literacy efforts
Organizations are rushing to launch AI literacy programs. Employees are participating in webinars. Compliance team publishes policy. The learning team is building a course that explains what generative AI is, how prompts work, and what risks to avoid. But something important is missing. Most AI literacy efforts are about increasing awareness rather than performance.
Even though employees have completed training and know more about AI, their behavior at work remains largely unchanged. They are still hesitant to use AI, even when it could be helpful. They place too much trust in the output, even when it requires scrutiny. They still misuse the tools in high-risk situations. They still struggle to determine when human judgment matters most.
Why most AI literacy efforts fail and what learning and development should do instead.
The problem is not knowledge. The problem is judgment. L&D teams are asking the wrong questions. Instead of asking, “Have employees learned about AI?” we should be asking, “Can employees leverage AI to make better decisions under real-world working conditions?” That change changes everything.
The hidden problem of AI literacy
Most AI literacy efforts follow a familiar pattern.
What is AI? Types of AI Benefits and Risks Ethics and Compliance Prompt Basics Check Your Knowledge
This approach makes sense in theory. Organizations want their employees to understand technology before using it. But there are flaws. Work is not a test. Real work is tedious, time-sensitive, emotional, and full of uncertainty. Employees rarely face situations like multiple-choice quizzes. Instead, you are faced with decisions such as:
Can AI be used to safely summarize this sensitive document? Should this recommendation be trusted or verified? Is this customer communication too sensitive for AI support? Are we saving time or introducing risk?
These are judgment calls. And judgment develops differently than knowledge.
Difference between knowledge and performance
Traditional study programs are optimized for recall. Performance is different. Achieving performance requires people to diagnose situations, adapt to changing conditions, consider tradeoffs, and act despite uncertainty. High performers often succeed not because they know more, but because they think differently. They instinctively adjust how they approach problems. Sometimes you need to be creative. Sometimes I’m skeptical. Sometimes even executions. Sometimes I have to be restrained.
The challenge is not just intelligence. It’s about knowing what kind of thoughts are needed at that moment. This is where many AI literacy efforts fail. They teach employees about tools, but not how to think with them.
A better model: performance intelligence
Organizations need to treat AI literacy not as awareness training, but as a judgment ability. One way to help think about this is through performance intelligence systems.
This is not a scientific theory or a new form of intelligence. It is an applied framework that combines established ideas from adaptive expertise, metacognition, deliberate practice, and performance feedback. The goal is simple. It’s about helping people make better decisions under pressure.
In practice, this means helping employees go through five stages:
Diagnose work status. Trigger the right thinking mode. Practice in uncertainty. Receive feedback. Adjust your movements and repeat.
Here’s what it actually looks like:
Step 1: Teach employees to diagnose context
Most training assumes that the same answer applies everywhere. That’s not the case in real work. Employees must first be aware of their situation. Consider these three common tasks:
Scenario A
Summarize the 90-page policy document. Scenario B
Draft a legal compliance statement. Scenario C
Dealing with frustrated customers.
AI may be suitable for all three situations. But it’s not the same. Your risk profile changes. The need for human oversight changes. The cost of mistakes varies. Rather than teaching blanket rules such as “use AI” or “avoid AI,” organizations need to teach context-specific decisions such as “What kind of problem is this?” What level of risk exists? How much human review is required? That’s a more useful skill than memorizing terminology.
Step 2: Teach your employees to switch thinking modes
Not all problems require the same cognitive approach. One of the biggest risks with AI is that employees use the wrong mode of thinking. for example:
creative mode
Generate ideas, brainstorm, and consider alternatives. analysis mode
Examine discrepancies, compare evidence, and identify patterns. Verification mode
Challenge output, test assumptions, and validate claims. decision mode
Choose your path despite incomplete information. escalation mode
Recognize when human expertise is required.
A major cause of workplace failure occurs when employees remain in creation mode when validation mode is needed. In other words, they generate confidence and trust too easily. The most powerful AI users are not necessarily the most technically proficient. They are often the ones who know when to shift mental gears.
Step 3: Practice under uncertainty
Traditional training often resolves ambiguity. In practice, ambiguity increases. This mismatch weakens the transfer. Imagine the following scenario. A senior leader asks an HR professional, “Can you use AI to quickly summarize concerns about employee performance before tomorrow’s leadership meeting?” Soon, competing pressures emerge.
Limited time Privacy concerns Incomplete information Unclear policy boundaries Pressure from leadership
There is no perfect answer. That’s why scenarios are important. Employees must learn how to navigate trade-offs. Should we use AI? If so, what information is it safe to include? What level of validation is required? What risks outweigh the benefits of speed? This is what workplace capability really looks like.
Step 4: Give feedback on decisions, not just accuracy.
Most training feedback focuses on accuracy. But workplace decisions are rarely binary. A more powerful approach is outcome-based feedback. for example:
Choice 1
Employee uploads sensitive data to unauthorized tools. result
Increased privacy and legal risks. Choice 2
Employees will avoid AI completely. result
Missed productivity opportunities. Choice 3
Employees use approved workflows and verify output. result
Faster execution with risk management.
The lesson is not just whether the answer is right or wrong. The lesson is to understand the trade-offs. Employees improve faster when they understand why their decisions were successful or unsuccessful.
Step 5: Incorporate reflection into your work
Training rarely fails because people forget the content. You will fail because old habits will return. When people reflect on their actual work, their behavior changes. After implementation, organizations should ask employees:
What assumptions have changed? This week, when was AI most helpful? When did you decide not to use it? And why? What went wrong?
Small moments of reflection create stronger judgment over time. Eventually, employees will stop relying on strict rules and start developing better intuition.
Big opportunity for L&D
L&D has long focused on knowledge transfer. But in an environment shaped by AI, rapid change, and uncertainty, knowledge alone is becoming less valuable. The new competitive advantage is judgment. Organizations don’t just need employees who are knowledgeable about AI. We need employees who can:
Diagnose the situation. Recognize the risks. Switch your thinking mode. Make decisions under uncertainty. Learn from the results.
In other words, organizations need adaptable performers. The future of L&D may depend less on telling people what to think and more on helping them learn how to think when strategy breaks down. It’s not just an issue of AI literacy. It’s a performance issue.
Share with
