After the Training: How L&D Can Help Communication Skills Transfer to Real Conversations

By Mark Garner · Published October 8, 2026

The workshop went well

Picture a leadership program for first-time managers. One module teaches a model for difficult feedback: describe the specific behavior, explain its impact, agree on what changes next. The session goes well. The managers understand the model and can explain why it works. In pairs, they practice it on a polite colleague and produce clean, well-structured feedback. The evaluations are positive.

A month later, one of those managers sits down with an employee whose reports have been late three times. The opening follows the model. Then the employee says, "That's not fair. Nobody told me the deadline had changed." The manager apologizes, softens, and the conversation ends without any agreement on what happens next.

The training didn't fail. The manager learned the model, understood it and could demonstrate it in a controlled setting. What hadn't happened yet was using it when the conversation changed. That gap, between what training builds and what shows up at work, is what researchers call transfer, and it's where L&D teams can have a large effect, mostly through what they design around the training itself.

Knowing, understanding, rehearsing, demonstrating, transferring

A lot of different things get recorded as "learning." For communication skills, it helps to separate five of them. These aren't formal stages from a single model. They're a practical way to be precise about what a program has and hasn't shown.

• Knowing: the person can recall the model. They could pass a quiz on it.
• Understanding: they can explain why it works and when to use it.
• Rehearsing: they have said it out loud, in their own words.
• Demonstrating: they can perform the behavior when someone is watching, in a practice setting.
• Transferring: they use it at work, in the real conversation, and keep using it over time.

Training research makes a similar distinction. A meta-analysis of training criteria used an expanded version of Kirkpatrick's evaluation model that separates learning into immediate post-training learning, retention, and behavior or skill demonstration, treating each as a different measure (Alliger et al., 1997). Transfer to the job is a further step again. A program can succeed on the first four and still leave the fifth to chance.

What the research says

Start with the most important point: training works. A meta-analysis of organizational training found medium to large effects across reaction, learning, behavioral and results criteria (Arthur et al., 2003). A later meta-analysis focused on leadership training, drawing on 335 independent samples, found improvements in reactions, learning, transfer and results, and concluded that leadership training is substantially more effective than previously thought (Lacerenza et al., 2017).

The same research also shows that how training is designed matters. In the leadership training meta-analysis, the moderator analyses supported needs analysis, feedback, multiple delivery methods (especially practice), spaced training sessions, an on-site location, and face-to-face delivery that is not self-administered (Lacerenza et al., 2017).

The transfer question is different from the learning question, and what happens outside the classroom carries a lot of weight:

• Before and after matter. In a survey of training professionals at 150 organizations, respondents estimated that 62% of employees apply training on the job immediately afterward, 44% after six months and 34% after a year. These are professionals' estimates, not direct measurements of employee behavior, so they're best read as a sense of the problem rather than a rate. The more useful finding is that activities in the work environment before and after training were more strongly related to transfer than activities during training (Saks & Belcourt, 2006).
• Some factors show up consistently. A review of the transfer literature identified the factors with the strongest, most consistent relationships to transfer. In training design, these include behavioral modeling, error management and realistic training environments. In the work environment, they include transfer climate, support, the opportunity to perform, and follow-up (Grossman & Salas, 2011).
• Unused skills fade. A meta-analysis of skill retention research found substantial skill loss with non-practice or non-use, growing the longer skills went unused, and cognitive tasks were more susceptible than physical ones (Arthur et al., 1998). That research covers trained skills in general, not communication specifically, but the practical implication is hard to ignore: a feedback model taught in March and not used until October is at risk.

Taken together, the practical message for L&D isn't "training doesn't work." It's that training is one part of a system, and transfer depends on what surrounds it: what's defined beforehand, how realistic the practice is, whether people get the opportunity to use the skill, and what support and follow-up come afterward.

Why communication skills depend on practice

Conversation skills are especially exposed to this gap because the other person keeps changing the conditions. A model works cleanly when the other side cooperates. Real conversations bring resistance, competing priorities, emotional reactions, unexpected information, pressure from stakeholders, and consequences said out loud. Any one of these can pull someone away from what they intended to do.

That's why a realistic practice environment matters for these skills in particular. Practicing on a polite partner trains people for a conversation that won't happen. The longer explanation of how pressure changes communication, and what to do about it, is in Knowing What to Say Isn't the Same as Saying It Under Pressure.

A practical framework for L&D

Here's a sequence L&D teams can use to design for transfer. It's a practical framework, not a validated model, and it doesn't depend on any particular tool:

Train → Define the behavior → Rehearse → Add resistance → Observe → Debrief → Change one thing → Practice again → Apply → Reflect

The middle of the sequence, from rehearsing through practicing again, is the practice loop described in detail in the practice under pressure article. The steps that are specific to L&D are the ones on either side of it.

Train

Teach the model, the vocabulary and the reasons. This is what training does well, and the rest of the sequence depends on it.

Define the behavior

Before anyone practices, decide what you need to see, in observable terms. Not "gives effective feedback," but "names the specific behavior, and when challenged, acknowledges the point and returns to what needs to change." This is the step that most often goes missing, and it shapes everything after it: what the practice scenario includes, what an observer watches for, and what a manager asks about later. It's also a small-scale version of needs analysis, one of the design features supported in the leadership training meta-analysis.

Practice, with resistance

Rehearse the conversation, introduce realistic pushback, observe what happens, debrief one specific moment, change one thing, and run it again. This is where the defined behavior gets tested under conditions that resemble the real conversation. The cornerstone article covers each of these steps.

Apply

Make sure people actually get the opportunity to use the skill, and soon. Opportunity to perform is one of the work-environment factors associated with transfer (Grossman & Salas, 2011). If the next real feedback conversation is six months away, the program needs to account for that.

Reflect

After the real conversation, look back at the defined behavior. Did it show up? Where did the conversation go differently from practice? Then follow up: a manager's question, a coaching conversation, or another round of practice on the moment that broke. Follow-up is the step that turns a one-time event into something that can continue.

Where practice fits in a program

Practice can sit in several places. Not every program needs every stage; the right choice depends on the skill, the audience and how soon people will need it.

• Before training: a short practice run as a baseline, so participants and facilitators can see where the conversation breaks before the model is taught.
• Immediately after training: while the model is fresh, practice it against realistic resistance rather than only a cooperative partner.
• Between sessions: in multi-session programs, assigned practice between sessions spreads practice over time. Spaced training sessions were among the design features supported in the leadership training meta-analysis (Lacerenza et al., 2017).
• Before a high-stakes conversation: when someone knows a difficult conversation is coming, a focused practice run on that specific situation.
• After a real conversation: reflect on what happened, then practice the moment that didn't go as intended.

An illustrative program example

Illustrative example — not a VoicePrep customer case.

Return to the first-time manager program. This time, the L&D team designs the feedback module for transfer.

Training. The session teaches the feedback model as before. (For how the feedback itself can be structured, see how to give difficult feedback to an employee.)

Defined behavior. The team agrees on one observable behavior for the module: when the employee pushes back, the manager acknowledges the point once and returns to the specific behavior and what needs to change, rather than apologizing repeatedly or dropping the request.

Practice. Within a week of the session, each manager runs a practice conversation with an employee who becomes defensive and says some version of "nobody told me."

The defensive moment. One manager's first attempt follows the familiar pattern: the opening is clear, then the pushback lands and the manager starts explaining the history of the deadline change instead of returning to the behavior.

Evidence and debrief. The manager and a facilitator look at what was actually said in that moment. The debrief focuses on one thing, not the whole conversation.

One change. The manager decides to acknowledge the communication gap in one sentence, then return to the late reports and agree on the next deadline.

Second practice. Same scenario, same pushback. The response is shorter and returns to the behavior. It's not polished, but the defined behavior is visible.

The real conversation. Some weeks later, the manager has a real feedback conversation.

Reflection. In a follow-up, the manager's own leader or a coach asks about the moment of pushback specifically. Did the acknowledgment-and-return show up? What did the employee say that practice hadn't anticipated? That becomes the next practice scenario.

The example shows a design, not a result. It doesn't tell you the real conversation went well, and in practice some won't. (For more of the manager's side, see how managers practice difficult conversations before they happen.)

Measuring honestly

Most L&D reporting stops at the first two levels below. Each level answers a different question, and none of them stands in for the one above it.

1. Participation. Who attended or started. It tells you about reach. It doesn't tell you anything about learning.
2. Completion. Who finished. It tells you the program was delivered. It doesn't tell you whether anyone can do anything differently.
3. Practice evidence. Whether people practiced, how often, and on which scenarios. It tells you practice happened. It doesn't tell you how well.
4. Observed behavior in practice. Whether the defined behavior appeared during practice, as seen by an observer, a recording or structured feedback. This is the closest a program usually gets to "behavior or skill demonstration." It tells you what someone did in practice. It isn't proof of what they will do at work.
5. Workplace transfer. Whether the behavior shows up in real conversations, which usually needs input from managers, the people on the other side of the conversation, or the person's own structured reflection. It's harder to measure, and it's the question that matters most for the skill.
6. Business outcomes. Retention, engagement, performance or other results. Transfer can contribute to these, but many other factors affect them too, so a change in outcomes isn't automatically the result of transfer, and transfer doesn't automatically produce one.

Two research findings are useful here. First, the training criteria meta-analysis cited earlier found that reactions about how useful training was were more strongly related to on-the-job transfer than reactions about how much people liked it, and were stronger correlates of transfer than measures of immediate or retained learning (Alliger et al., 1997). If you collect post-session reactions, it's worth asking about usefulness, not only satisfaction. Second, the organizational training meta-analysis found that the choice of evaluation criteria was related to measured training effectiveness (Arthur et al., 2003). What you measure shapes what you can claim.

The practical rule: report each level as what it is. Practice evidence is evidence about practice. It's valuable, because it's much closer to behavior than a completion rate, but it isn't workplace transfer, and workplace transfer isn't automatically a business outcome.

What AI practice can and cannot do

AI practice tools, including ours, can make the practice steps easier to provide at scale. VoicePrep is a practice layer between training and performance. It gives people adaptive scenarios where the other side of the conversation responds to what they say: it can push back, introduce new information, and react when someone holds or gives up their position. People can rehearse a specific conversation, get AI feedback on what they said, and run it again. That makes repeat practice, including practice between sessions, easier to provide, and it gives a facilitator, manager or coach something concrete to debrief.

What it can't do matters just as much:

• It can't show that someone will use the skill at work. A strong practice session is evidence about that session.
• It isn't a validated assessment, and its feedback doesn't predict workplace performance or certify readiness.
• It doesn't replace facilitators, managers, coaches, organizational support, or the real-world opportunity to use the skill.

That last point is supported by the research above. The leadership training meta-analysis found stronger effects for face-to-face delivery that was not self-administered (Lacerenza et al., 2017). That study didn't evaluate AI practice, and it doesn't mean self-directed practice has no value. But AI practice done alone is a form of self-administered practice, so our reading is that it works best as one component of a program: assigned with a clear purpose, connected to a defined behavior, debriefed with a person where possible, and followed up after the real conversation.

A practical takeaway for L&D

Training matters. Practice matters. Transfer depends on what happens around both.

So instead of asking only "Did they complete the training?", it's worth asking:

• What behavior do we need to see?
• Where can people rehearse it?
• What happens when the conversation gets harder?
• What evidence do we have from practice?
• What happens when they use it at work?
• What support follows?

Programs that can answer those questions don't guarantee transfer, but they give it a much better chance. (For the coaching side of this, see why coaching doesn't always transfer to real conversations.)