Knowing What to Say Isn't the Same as Saying It Under Pressure

By Mark Garner · Published October 7, 2026

The moment the script stops working

A department head walks into a quarterly review with a plan to cut her team's budget by twelve percent. She knows the numbers, she has rehearsed the opening, and the first two minutes go exactly as planned.

Then the board chair asks: "If this doesn't hold, what's the next thing you cut?"

She knows the answer. But what comes out is a long, hedged response about "evaluating options across the portfolio" that never names anything. On the drive home, she says the answer she meant to give, clearly, in one sentence.

Most people who have had a high-stakes conversation know that drive home. The problem wasn't knowledge. It was saying it at the moment the conversation stopped following the plan.

The hard part of most important conversations starts after the part you prepared for: when someone pushes back, asks something unexpected, adds new information, disagrees, gets emotional or raises the stakes. So don't just rehearse the answer. Practice what happens when the answer stops working.

The knowing-doing gap in conversations

Jeffrey Pfeffer and Robert Sutton described a problem they saw across organizations: companies rarely lack knowledge about what to do. Their real difficulty is turning what they know into action. They called it the knowing-doing gap (Pfeffer & Sutton, The Knowing-Doing Gap, 2000).

They were writing about organizations, but a similar pattern shows up in individual conversations. Plenty of managers have been through feedback training. They know to describe the behavior, not the person. Then the employee says, "That's not fair, nobody ever told me that," and the model disappears. What's left is an apology or an argument.

Afterward, that manager could probably tell you what they should have said. The gap isn't knowledge; it's delivery under changed conditions, which is a different skill. Courses, frameworks and workshops are much better at building the first than the second.

Training creates knowledge. Practice creates performance. This article is about the second half of that sentence.

Why pressure changes communication

A few things reliably change the conditions of a conversation:

• Pushback: the other person rejects your premise, not just your conclusion.
• The unexpected question: fair, relevant, and not the one you rehearsed.
• New information: something you didn't know changes what the right answer is.
• Disagreement: someone with status or influence says no.
• Emotion: the other person gets angry, upset or defensive, or you do.
• Consequences: the stakes are said out loud. "So what happens to my team?"

When one of these lands, people tend to fall into one of three patterns. Some freeze or ramble, filling space while they look for footing. Some over-soften, backing away from their point so the discomfort stops. Some retreat into generalities, answering a specific question with an abstract one, like the department head above.

These reactions don't necessarily mean the person doesn't know the material. More often, they have never heard that question out loud, under that kind of pressure, and had to answer it in real time.

There's a second reason knowledge alone doesn't hold: unused skills fade. A review of communication-skills teaching in medicine notes that "communication skills tend to decline with time unless they are regularly recalled and practiced" (Junod Perron et al., Swiss Medical Weekly, 2015). A workshop in March doesn't guarantee much about a conversation in October.

Where high-stakes conversations actually break

When conversations go wrong, the failure usually isn't in the opening, the part people prepare. It tends to come two or three exchanges in.

The difficult employee conversation. A manager opens well: specific, calm, focused on behavior. The employee says, "You never told me this was a problem." Now the manager has to decide, live, whether to defend the past silence, concede it, or acknowledge it and still hold the line on what needs to change. The third option is right, and it's the hardest to find in the moment. (How managers practice difficult conversations goes deeper on this.)

The board conversation. The presentation lands. Then: "What would make you change your recommendation?" These questions test whether you've thought one step past your own plan, and whether you can say so in a sentence. (You can try a version of this in the free board challenge scenario.)

The interview follow-up. The candidate tells a well-rehearsed project story. The interviewer asks, "What would you do differently?" The polished answer has run out. (Interview preparation covers what these follow-ups are listening for.)

The structure is the same in all three. The prepared part goes well, the conversation changes, and the real test is what you do next.

What effective conversation practice looks like

If the hard part comes after the opening, practice that only covers the opening is practicing the easy part. Yet many people "practice" by writing down what they want to say, maybe saying it once to a colleague, and calling it prepared.

There's a real difference between having role-played it once and having actually practiced the skill. A one-off role-play is usually a single attempt, against a colleague who is being polite, ending once the opening has been delivered, with no second try. It's useful, but it's a rehearsal of the script, not practice of the conversation.

Practicing the skill is a loop. Each step is there for a reason.

1. Rehearse out loud

Say it, don't just think it. A sentence that sounds clear in your head often comes out longer, softer or more tangled when spoken. Saying it out loud tells you whether your opening works as speech and gives you a baseline.

2. Resistance

This is the step most practice skips, and it matters most. The other side needs to push back, ask the question you didn't want, add information you didn't have, or react emotionally. A partner who accepts your opening and agrees is training you for a conversation that won't happen.

Good resistance is realistic, not theatrical. An employee who says "nobody told me" is more useful than one who storms out. A board member who calmly asks what you'd cut next is harder to answer than one who is simply rude.

3. Observe

After the run, look at what you did, not what you meant to do. Did you answer the question that was asked? Did you take a position or hedge? How long did it take to get to the point? Did you hold the line you came in with, or give it away once the pressure arrived? A coach, a colleague or a recording all help, because people are poor judges of their own delivery in the moment.

4. Debrief one concrete moment

Not "that went fine" or "I need to be more confident." Pick a moment: when she said "nobody told me," I apologized three times and dropped the point I came to make. Specific moments can be changed. General impressions can't.

5. Change one thing

Choose one adjustment. Acknowledge the past once, then return to the behavior. Answer "what would you cut" with a name before you explain it. Stop after your recommendation instead of adding qualifications. Trying to fix everything at once usually means nothing changes.

6. Practice again

Run the same conversation again, against the same resistance, and see whether the change holds. This is the step that separates practice from rehearsal. The first run shows you where the conversation breaks. The second run, with one intentional change, shows you whether you can do something different when it breaks. The goal over several rounds is for the response you had to think about to become one you can reach for.

Here's the loop in a single example. A manager practices a feedback conversation. The debrief shows that when the employee pushed back, the manager shifted from describing behavior ("the last three reports were late") to describing attitude ("you don't seem committed"). The change for round two: when pushed, return to the specific behavior. The second run, against the same pushback, sounds different. Not perfect, but different in the way that mattered. (Employee feedback conversation practice and how to practice difficult conversations cover the preparation side.)

The loop is simple. What makes it work is repetition: repeated, feedback-driven practice against a conversation that pushes back.

What practice can actually reveal

Practice makes visible what otherwise stays hidden until the real conversation: whether your point is clear, whether your answer has structure, whether you take a position or hedge, whether you're concise and specific, and whether you adjust when new information arrives. Some can be scored consistently; others are better judged by a person in a debrief. VoicePrep scores some of them, as described below.

What practice cannot prove

Here's the part that's easy to leave out: doing well in practice is not proof you'll do well in the real conversation.

The other person may do something no scenario anticipated. A strong practice session is evidence of what happened in that practice session—not a guarantee of performance.

A paper by Amazon researchers on Conversation Coach, a voice-enabled system for practicing difficult workplace conversations, reports that more than 40,000 managers used it over six months, across more than 54,000 sessions. That shows real demand for this kind of practice. The authors also list among their limitations that they have no causal proof of downstream effectiveness, and note that "adoption alone does not establish coaching effectiveness" (Wu et al., "Conversation Coach," 2026). That standard applies to every product in this space, including ours.

The communication-teaching review cited earlier notes that whether communication skills transfer into practice is difficult to assess, and recommends a working environment in which their use is valued, supported and rewarded (Junod Perron et al., 2015).

The practical implications:

• Follow up after the real conversation. Where did it break? Practice that next.
• Keep practicing over time. One good session before a big meeting is preparation, not a skill.
• Build an environment that supports it. Managers and coaches who ask about practice and expect it make it more likely to stick.

Practice narrows the gap between knowing and doing. It doesn't close it on its own.

The practice layer between learning and performance

Learning is where you gain the concepts, through courses, books, workshops and coaching.

Performance is the real conversation, with real stakes.

Between them is the practice layer: where people rehearse those concepts under changing conditions, get feedback, adjust and try again before the conversation counts. When that layer is missing, the result is the practice-to-performance gap: people leave training knowing what good looks like, then meet the real conversation without ever having done it.

That layer is what we built VoicePrep for. It's AI-powered practice for high-stakes conversations: interviews, leadership, difficult conversations, negotiation and executive pressure. Practice is live and scenario-based — spoken in the live interview with Amina, typed in the role-play rooms — and the other side of the conversation adapts to what you say. It pushes back when you're vague, can introduce new information or consequences, and reacts when you hold or give up your position. Afterward you get AI feedback, and you can run the scenario again. It's designed around the loop above, especially the resistance and the second attempt. (For a broader view, see leadership development.)

It isn't a replacement for learning or for a good coach.

How VoicePrep observes communication

VoicePrep rooms give different kinds of feedback: leadership rooms score scenario-specific dimensions, and executive rooms give a written, evidence-based debrief. In live interview sessions with Amina, VoicePrep's AI interviewer, VoicePrep also produces the Communication Index, which tracks five dimensions:

• Clarity: whether your point comes through.
• Structure: whether your answer is organized.
• Confidence: whether you take a position and speak to it.
• Conciseness: whether you get to the point without filler.
• Impact: whether you name concrete outcomes rather than describing things vaguely.

Each dimension is scored from 0 to 100. Four (clarity, structure, confidence and impact) are AI judgments of what you said against a fixed rubric, not a ranking against other users. If your camera is on, confidence also includes an on-camera presence signal; otherwise it's based on language alone. Conciseness is calculated rather than judged, from filler words and speaking pace.

Every dimension needs enough evidence: if your answers are too short to judge a dimension fairly, it's left blank rather than guessed, and no Index is produced from fewer than two answers. Each score also keeps the evidence behind it, so feedback points back to what you actually said.

For signed-in users, the Index is kept across sessions, so you can see how your scores move over time. Coaches can see it for linked clients, for the sessions completed through that coaching relationship.

The Index is a consistent, structured way to observe communication in practice. It isn't a clinical or psychometric instrument, and it doesn't predict how a real conversation will go. It's one input to the debrief, not the verdict. For more on how evidence, change and the Index's limitations are handled, see How VoicePrep's Communication Index Works: What It Measures in Practice, and What It Doesn't.

For L&D teams and coaches

For L&D teams, the gap shows up after training. A workshop can teach a feedback model in an afternoon; whether people can use it when an employee pushes back is a separate question. And a completion rate shows that practice happened, not that someone is ready.

The challenge for L&D isn't only delivering useful training, but creating opportunities for people to carry those skills into the conversations where they matter. For a practical L&D approach, see After the Training: How L&D Can Help Communication Skills Transfer to Real Conversations.

For coaches, the same applies between sessions, which is often where the most valuable work happens. A client leaves with a plan, and the next time you hear about it, the conversation has already happened. Assigned practice fills that gap: the client practices the conversation, including the pushback, and you have something concrete to discuss.

The challenge for coaches isn't only giving useful feedback, but helping clients carry a communication skill into the conversation where it matters. For a coach-focused approach, see Why Coaching Doesn't Always Transfer to Real Conversations, and How Practice Helps.

The same review recommends a working environment that values and supports the use of these skills (Junod Perron et al., 2015), and coaches and L&D teams are often the people who build that environment. More on how VoicePrep supports this work is on the coaches and L&D page.

Knowing it and saying it are different skills

Knowing what to say is valuable; frameworks, training and coaching give people the right ideas. But in most of the conversations that shape careers and teams, the hard part comes after the prepared part: the follow-up question, the pushback, the new fact, the emotional reaction. Saying the right thing at that moment is a separate skill, built through repeated, feedback-driven practice.

Training creates knowledge. Practice creates performance. Don't just rehearse the answer. Practice what happens when the answer stops working.