Even with artificial intelligence (AI) coming in, and maybe especially because AI is coming in, companies agree that human communication skills remain at the top of what employers want today.

In fact, GMAC’s 2025 Corporate Recruiter Survey ranked communication skills as the third-most important skill, only behind problem-solving and strategic thinking. And it’s in the top six skills employers expect to matter even five years from now.

But here’s the harder question: How do we measure whether someone can handle difficult conversations and whether communication training improves performance?

The Problem With Measuring Communication Skills

I saw this challenge firsthand with one of my company’s pilot customers. They had already invested in workshops to help employees handle difficult workplace conversations. The post-training surveys all said that the workshops were well received. Employees learned frameworks, talked through examples and were even encouraged to practice through peer role-play afterward.

The company was doing many of the right things. The problem was that they had very little visibility into what happened next. They couldn’t easily tell who had completed the role-plays. They couldn’t tell how difficult the conversations were, how well people handled them or whether someone who struggled the first time actually improved with practice.

In other words, they could measure the training experience. They couldn’t really measure the skill.

That gap is surprisingly common. According to the 2025 ATD research report, “The Future of Evaluating Learning and Measuring Impact,” 93% of organizations collect participant satisfaction data after training. Yet only 30% say they are good at using learning data to make business decisions. Even more telling, 90% say isolating the impact of training on results is a challenge.

The Current Limitations of L&D Metrics

Learning and development (L&D) has historically been forced to measure what is easy to capture rather than what it wants to know: Did people attend? Did they like the program? Did they pass a knowledge check? Do they feel more confident?

All of those things can be useful signals. But none of them necessarily answer the question we really care about: Can this person handle the conversation well when the pressure is on?

That distinction is especially important with communication skills because communication is not primarily a knowledge problem.

For example, a manager may know the framework for giving difficult feedback and still become vague when an employee gets defensive. Or a customer success manager might know they should set boundaries and still give in when a valuable client becomes frustrated.

The challenge is not simply knowing what to do. It is being able to do it in real time, while the pressure is high. That is what makes communication skills so difficult to measure — it’s contextual, behavioral, pressure-dependent and multidimensional.

Traditional Role-Play and Its Challenges

So, what solution has L&D traditionally relied on to improve communication skills? Human-to-human role-play. It forces people to think on their feet instead of selecting the “right” answer from a multiple-choice list.

And research supports the value of deliberate practice. In one randomized study of communication training, participants who received deliberate communication practice significantly improved their demonstrated communication skills, while those in the traditional training control group did not. Interestingly, both groups reported increased self-efficacy.

The problem with traditional role-play is not that it is ineffective. It’s that it’s difficult to scale and even harder to measure consistently.

Peer role-plays can vary dramatically. One participant may get a cooperative partner while another gets someone much more challenging. Feedback quality depends on who is observing. Employees may feel awkward practicing difficult conversations in front of colleagues. And unless someone is carefully tracking every interaction, there is rarely a useful dataset showing whether people are improving over time.

Shift to Demonstrated Performance

This is where I think we need to rethink what communication skill measurement should look like. Instead of relying primarily on satisfaction scores or self-reported confidence, organizations should start measuring demonstrated communication performance.

That means putting employees into realistic situations and looking for observable behaviors.

For example, in a difficult performance conversation, did the manager clearly explain the issue? Did they use specific evidence? Could they acknowledge the employee’s perspective without backing away from accountability? Did they manage defensiveness? Did they leave the conversation with clear next steps?

Those behaviors can be measured using AI. And, importantly, they can be measured more than once so that the question becomes less about whether someone performed well on a single assessment and more about whether they are getting better: Baseline. Practice. Feedback. Retry. Improvement.

The Role and Benefits of AI Simulations

Through my experience, we’ve seen that AI-powered simulations make this kind of performance measurement far more scalable than it’s been in the past.

Learners enter private, simulated workplace conversations in which an AI persona responds dynamically to what they say. The interaction can become skeptical, defensive or resistant, requiring the learner to adapt in real time. Their performance is then evaluated against observable communication behaviors, giving the learner immediate feedback and the opportunity to repeat the experience to see whether their performance improves.

That creates something L&D teams have historically struggled to capture: a skills dataset rather than an attendance dataset.

Instead of simply knowing that 200 managers completed a training — because each individual is practicing the same consistent role-play scenario — an organization can begin to see trends such as most managers are strong at explaining expectations but consistently struggle to manage defensiveness. Or that a cohort improved meaningfully after several rounds of practice.

That kind of information is much more useful for both the learner and the business.

Elevating Rigor in Communication Training Measurement

None of this means we should stop measuring satisfaction, confidence or knowledge. Those measures still have value. But if communication is truly one of the skills organizations care about most, we should hold ourselves to a higher standard.

We would never evaluate someone’s ability to code based only on whether they enjoyed a coding workshop. We would not certify a pilot solely on their confidence about flying. Communication skills deserve the same rigor.

The question after training should not be, “Did people learn something?” It should be: “When the conversation gets difficult, can they actually do it?”

And just as importantly: “Are they getting better?”