The Productive Struggle: Why Great Learning Shouldn't Always Be Easy

Part 2 of The AI Skills Paradox Series

Chief Learning Officer "June" — Follow Her Story

June's team had begun asking a different question about every learning request. Instead of immediately determining what content to create, they first asked what people needed to do differently and what experience would help them become more capable. The shift was already changing how the organization approached learning, particularly when the underlying challenge involved behavior rather than information.

Then June encountered a different problem.

A leadership development program included a difficult coaching simulation. Managers were expected to prepare for a conversation with an underperforming employee, determine how to approach the discussion, conduct the conversation, and reflect on what they could improve. With AI readily available, several participants used it to prepare detailed talking points, anticipate the employee's responses, formulate questions, and recommend exactly how they should respond to resistance.

Their answers were excellent, but June wondered whether the managers themselves had actually gotten better at coaching. The technology had helped them produce a better result, yet it may also have performed some of the cognitive work the exercise was designed to develop.

That distinction introduced the next challenge in June's exploration of the AI Skills Paradox. L&D has spent years making learning faster, easier, more accessible, and increasingly frictionless. Much of that progress has been valuable, but if developing capability requires people to wrestle with decisions, make mistakes, receive feedback, and try again, some of the friction we're eliminating may actually be where learning happens.

L&D Has Spent Years Removing Friction

For good reason, the learning industry has spent decades improving the employee experience. Long classroom programs have been supplemented with shorter digital experiences. Search tools make information easier to find. Mobile learning makes development available wherever employees work. Personalized recommendations help people navigate increasingly large content libraries, while performance support delivers information at the moment of need.

AI accelerates nearly every part of that trend. Employees no longer need to search through multiple resources to find an answer because AI can summarize them instantly. They don't necessarily need to wrestle with how to structure a difficult conversation because an assistant can create the talking points. They can generate examples, clarify concepts, summarize research, draft responses, and receive recommendations without leaving the flow of work.

Removing unnecessary friction is usually a good thing. Employees shouldn't have to navigate a confusing LMS, sit through irrelevant content, spend hours locating a policy, or complete repetitive training simply because that's how the program was designed. Technology should make those experiences easier.

The challenge is that not all friction serves the same purpose. Some friction interferes with learning, while other forms of effort are part of how people develop capability. Treating both as problems to eliminate can create efficient learning experiences that inadvertently remove the very activities employees need to improve.

Some Struggle Is Part of How People Learn

In his memo to Unboxed employees, President Dave Romero describes human development as inherently difficult at times. People struggle to find the right words, wrestle with concepts, stumble through presentations, attempt unfamiliar tasks, and discover through experience that they understand less than they initially thought. He argues that the cycle of "trying, missing, adjusting, and trying again" is often where meaningful learning actually occurs. The Work We Are Really In

That observation has significant implications for enterprise learning.

Consider a new manager learning how to deliver difficult feedback. Reading a framework may provide useful knowledge, and watching an expert demonstrate the conversation may establish a model for good performance. Neither experience guarantees that the manager can respond effectively when an employee becomes defensive, emotional, or unexpectedly challenges the feedback.

Capability begins developing when the manager has to make decisions within the conversation. They must choose their words, interpret the other person's response, adjust their approach, recognize when something isn't working, and decide what to do next. Feedback after the experience helps them understand what worked and what didn't, while another attempt allows them to apply that insight.

The difficulty isn't incidental to the learning experience. In many cases, it is part of the learning experience.

AI Can Produce the Outcome Without Developing the Skill

This creates a new challenge for L&D because AI is remarkably good at helping people produce better outputs.

Ask AI to write an executive email and it can create a polished response. Ask it to prepare discovery questions and it can generate an excellent list. Ask it how to handle an employee performance issue and it can recommend an appropriate approach. Ask it to analyze a business problem and it can organize information, identify possibilities, and suggest a decision.

Those capabilities create real productivity gains in the workplace. Employees should absolutely use them when the objective is to produce a high-quality output efficiently.

Learning, however, sometimes has a different objective.

If the purpose of an exercise is to develop executive communication, having AI write the entire message may improve the email without improving the employee's communication capability. If the objective is to develop critical thinking, immediately asking AI for the answer can bypass the reasoning the employee needs to practice. If the objective is to develop coaching capability, having AI prescribe every question and response may help complete the conversation without helping the manager learn to navigate one independently.

This doesn't mean AI should be removed from learning. It means learning designers must become much more intentional about determining which work AI should perform and which work the learner needs to perform in order to develop the intended capability.

The Difference Between Bad Friction and Productive Friction

Dave makes a useful distinction in his memo when he argues that Unboxed's job isn't to remove every sharp edge from learning. The more important responsibility is understanding which edges represent unnecessary friction and which are part of the experience that helps someone grow. The Work We Are Really In

June began using that distinction with her team.

Bad friction makes development unnecessarily difficult without contributing meaningfully to the desired capability. It can include:

• Confusing navigation or technology
• Difficulty locating relevant information
• Excessive administrative steps
• Redundant or irrelevant content
• Poorly designed learning experiences
• Unnecessary scheduling or access barriers
• Repetition that doesn't increase capability
• Retrieving knowledge instead of immediately receiving the answer

Technology should aggressively remove these obstacles because they consume time and attention without making employees better.

Productive friction requires effort because the effort itself contributes to development. It can include:

The distinction is critical. The goal of modern learning design shouldn't be to maximize difficulty, nor should it be to make every experience effortless. The objective is to remove friction that interferes with development while intentionally preserving the effort that builds capability.

• Retrieving knowledge instead of immediately receiving the answer
• Making decisions under realistic constraints
• Practicing unfamiliar behaviors
• Working through ambiguity
• Responding to unexpected situations
• Receiving corrective feedback
• Reflecting on mistakes
• Trying again after an unsuccessful attempt

Practice Gives Failure a Productive Place to Happen

As June considered this distinction, practice became increasingly important to her learning strategy.

Organizations understandably want employees to perform well. A salesperson shouldn't learn how to handle a difficult objection by repeatedly mishandling real opportunities. A manager shouldn't develop coaching skills entirely through trial and error with employees. A leader shouldn't encounter their first high-stakes conflict without having previously practiced how they might respond.

The problem is that capability still requires experience.

Practice creates an environment where employees can encounter difficulty without exposing the organization to the full consequences of failure. A learner can make the wrong decision, experience the response, receive feedback, reconsider the approach, and try again. What would be costly in the workplace becomes developmental in a simulation.

AI significantly expands what is possible in this area. Instead of limiting realistic practice to occasional classroom role plays, organizations can create dynamic scenarios where employees repeatedly practice customer conversations, leadership situations, coaching discussions, negotiations, or other complex interactions. The scenario can respond differently depending on the learner's choices, allowing the experience to retain the uncertainty and decision-making that make real performance difficult.

Used this way, AI doesn't remove productive struggle. It creates more opportunities to experience it safely.

Feedback Shouldn't Eliminate the Need to Think

AI also changes the speed and availability of feedback. Historically, employees might complete an exercise and wait for an instructor, facilitator, peer, or manager to review their performance. AI can increasingly provide feedback immediately, creating much shorter cycles between action, reflection, and another attempt.

That acceleration can be powerful, but June recognized another design decision hiding inside it.

Feedback can tell someone exactly what to do next, or it can help them understand their performance well enough to determine what to do next themselves. Those approaches may produce very different developmental outcomes.

If a learner receives an immediate correction after every imperfect decision, they may become highly effective at following guidance without developing the judgment required to diagnose their own performance. In other situations, direct guidance may be exactly what the learner needs to prevent misconceptions from becoming established.

Effective feedback therefore requires more than speed. It requires intentionality.

Depending on the capability being developed, feedback might ask learners to reflect on their choices before providing recommendations, compare their approach against observable criteria, identify what they would change, or repeat the scenario with new constraints. AI can support each of these experiences, but the design should determine when the learner needs an answer and when the learner needs another opportunity to think.

The Goal Isn't Difficulty. It's Desirable Difficulty.

This distinction helped June avoid an obvious mistake. Recognizing the value of productive struggle doesn't mean making learning harder simply for the sake of making it harder.

Employees already operate in demanding environments. Adding unnecessary complexity, withholding useful resources, or forcing people to memorize information readily available in the flow of work doesn't create meaningful capability. It creates frustration.

The more useful question is whether the effort required by the learning experience corresponds to something the employee needs to be capable of doing.

If a salesperson needs to think quickly when a customer raises an unexpected objection, practice should require them to formulate a response rather than select one from a list. If a manager needs to diagnose why an employee is struggling, the experience should require interpretation rather than simply presenting the diagnosis. If a leader needs to navigate ambiguity, the scenario shouldn't provide every relevant piece of information before requiring a decision.

The difficulty should reflect the capability.

That gives L&D a practical design principle: make access to learning easy, but don't automatically make the learning itself effortless.

June Began Designing for the Moment Before the Answer

The biggest change in June's thinking was surprisingly simple. Her team started paying much more attention to what happened immediately before a learner received help.

Previously, efficiency often meant getting employees to the correct answer as quickly as possible. Now the team asked whether the learner should first retrieve something from memory, make a decision, attempt a response, explain their reasoning, or experience the consequences of a choice.

That moment became particularly important in AI-enabled experiences.

Instead of immediately asking AI to generate the best response, a learner might first formulate their own approach and then compare it against AI-generated alternatives. Instead of receiving coaching before attempting a difficult conversation, the employee might complete a simulation, review feedback, reflect on their choices, and then use AI to explore different strategies before trying again.

The technology remains present throughout the experience, but its role changes. AI becomes a practice partner, feedback mechanism, and developmental accelerator rather than simply an answer engine.

For June, this represented a much more compelling vision of AI-enabled learning. The objective wasn't to protect employees from every difficult moment. It was to create the right difficult moments in environments where people could learn from them.

Productive Struggle Requires Psychological Safety

There was another important implication. Asking employees to practice, fail, receive feedback, and try again only works when the environment allows them to do so without unnecessary fear.

If every imperfect practice attempt becomes part of a formal performance evaluation, employees will naturally optimize for appearing competent rather than experimenting with unfamiliar behaviors. If simulations feel punitive, people will search for the safest answer instead of exploring different approaches. If managers treat developmental feedback as evidence of failure, practice loses much of its value.

Organizations therefore need to distinguish between developmental practice and performance assessment.

Both have a role. At some point, organizations may need to verify that an employee can demonstrate a required capability. Before that moment, however, employees need space to develop it.

AI-enabled practice can make this distinction particularly powerful because it allows people to repeat difficult experiences privately and frequently before demonstrating capability in higher-stakes environments. That creates more opportunities to struggle productively while reducing the business and interpersonal risks associated with learning exclusively through real-world failure.

Key Takeaways

The pursuit of frictionless learning has produced enormous benefits. Employees can access information more easily, learn within the flow of work, receive personalized support, and eliminate administrative barriers that historically made development unnecessarily difficult. AI will accelerate those improvements, and L&D should continue removing friction that consumes time without contributing to performance.

The mistake is assuming that all effort represents poor learning design. Developing complex human capabilities often requires employees to retrieve knowledge, make decisions, navigate ambiguity, practice unfamiliar behaviors, receive feedback, reflect on mistakes, and try again. Removing those experiences can make learning feel easier while simultaneously making capability development weaker.

The opportunity for L&D is to become much more deliberate about where friction belongs. Technology should remove obstacles around learning while preserving the productive effort within learning that helps people grow. AI is particularly powerful when it enables employees to practice difficult situations safely, receive meaningful feedback, and repeat the experience until stronger behaviors begin to emerge.

As AI makes answers increasingly easy to obtain, designing experiences that develop the judgment required to know what to do with those answers may become one of L&D's most important responsibilities.

Coming Next

In Part 3, June encounters another consequence of unlimited AI-generated learning. When nearly anyone can create a polished course, scenario, assessment, or piece of feedback, producing content is no longer the hardest part of L&D's job. We'll explore Content Rich, Context Poor and why diagnosis, business understanding, empathy, and human judgment become more valuable as AI makes content increasingly abundant.