New manager training: Practicing the conversations nobody prepares you for




Many new managers are promoted for doing their own job well, then handed a job made of conversations they have never had. New manager training prepares them for that shift. It covers feedback, one-on-ones, conflict, delegation, accountability, emotional intelligence and time management, and it sticks when the hardest conversations get rehearsed before they happen for real.
The first of those conversations usually arrives without a script, whether it is telling a former peer that their work is slipping or raising a conflict between two reports. Without rehearsal, the words get tried out loud for the first time in front of the person who needs to hear them.
New manager training covers the behaviors a person needs once their results depend on other people’s work. Gallup estimates that managers account for at least 70% of the variance in team engagement, so each of these topics affects a whole team:
Formal preparation reaches only about half of new managers. The Cegos 2025 International Barometer surveyed 4,271 first-time managers across 10 countries and found 56% had been trained before stepping into the role. Gallup’s 2025 Global Workplace report found a similar pattern across all managers, with 44% worldwide ever receiving management training.
The weakness shows up in the conversations themselves. Training Industry reported results from the Oji Leadership Poll, run with The Harris Poll. Four in ten workers rated first-time managers weak at reducing conflict, handling difficult situations and providing quality feedback.
Workshops teach what a good conversation contains, and rehearsal builds the habit of having one. A meta-analysis of leadership training points the same way, favoring programs that combine practice, feedback and spaced sessions. Practice-only programs have not outperformed information-only ones on transfer, so the strongest designs combine the two.
The 70-20-10 rule, from the Center for Creative Leadership (CCL), says leaders develop roughly 70% through challenging assignments, 20% through developmental relationships and 10% through coursework. The figures are a rough guide, and practice sits between the 10 and the 70.
Rehearsal turns coursework into something a manager can try on the job. A mentor or coach then supplies the 20 by reviewing what happened.
Face-to-face conversation carries more than the words. Media richness theory ranks it as the richest medium because it adds immediate feedback, body language and tone, though later tests only partly supported that ranking.
Practice benefits from those same cues, since a camera lets a coach see eye contact, posture and nervousness that a voice-only tool never receives. Cues alone are not enough, because Training Industry notes that one-off workshop role-play fades without regular practice. A new manager therefore needs a partner on call before each hard conversation, and Tavus builds the Personified Application Layer (PAL) to be that partner.
The conversations worth rehearsing first are feedback, one-on-ones and difficult conversations. Feedback is closely tied to engagement, since Gallup finds 80% of employees who say they received meaningful feedback in the past week are fully engaged.
Before giving corrective feedback, a new manager should decide whether the problem is ability or motivation. The skill/will matrix from The Tao of Coaching sorts a person’s readiness for a specific task into four quadrants:
The matrix guards against a habit Harvard Business Review documented in “The Set-Up-To-Fail Syndrome,” where managers typically assume poor performance is the employee’s fault. An AI practice scenario can be built twice, once with a report who lacks the skill and once with one who lacks the will.
Hold a weekly one-on-one of about 30 minutes with each report, which Harvard Business Review research found employees rated most desirable and which correlated with the highest engagement. Let the report own most of the agenda, since the same research found one-on-ones work best when topics important to the direct report dominate. Good coaching questions cover what Gallup finds meaningful, such as recognition and strengths.
Each question invites the report to set the topic, which keeps the meeting theirs. A manager can rehearse the opening of a one-on-one with an AI practice partner, including the pause after the first question, which the partner should leave open instead of filling.
Difficult conversations for new managers tend to fall into a few types. Examples include telling a former peer that work is slipping, raising a missed deadline, settling a conflict between two reports and giving a first performance review. Whatever the topic, the same five steps apply.
Each step is a separate skill, and an AI practice partner gives a manager somewhere to rehearse them before the real conversation. The other side can push back, deflect or disagree, and the manager can rerun the same conversation until the facts and the next step come out clearly.
The biggest obstacle to rehearsal is usually finding someone to rehearse with. An AI role-play partner plays the other side of the conversation on demand, so a new manager can try a feedback or conflict opening several times before the real one. A practice partner supplies the repetitions, and a human coach supplies the debrief.
Realistic pushback depends most on how well the partner reads the person, and that is where Tavus’s approach starts.
A PAL is a real-time application you talk to face to face that sees, hears, remembers and responds. In new manager training, a PAL plays the other side of the practice conversation, and each session runs live on the Conversational Video Interface (CVI) with nothing pre-recorded. Four components work together as one loop, and each one covers a part of a real hard conversation.
What the manager practices is set by the scenario, and the learning team that owns the leadership training program builds that part.
Practice only builds skill when it is aimed at something specific and checked afterward.
In Tavus, a learning and development (L&D) team builds each scenario in PAL Maker, the no-code builder, and sets Objectives such as naming an observable behavior. Memories carry each participant’s progress into the next session, and transcripts can route into a learning management system through Tavus’s AI L&D Coach, which also exports competency data.
A first corrective conversation with a former peer is a typical starting scenario. In this hypothetical session, a newly promoted team lead must tell a former peer that his handoff notes keep arriving late and incomplete, with two Objectives to meet. She opens with “I know you are busy, and this is probably nothing, but the handoffs have been rough lately,” and Raven-2 catches the mismatch with her avoidant gaze.
When the former peer pushes back with “Rough how? Everyone has been slammed,” she retreats, leaving no behavior named, no next step and two Objectives open. After a transcript debrief, the retry opens with the same pushback because Memories carry the first attempt forward. This time she answers, “On Tuesday and Thursday, the handoff notes came after the shift change and had no ticket numbers,” and both Objectives close with a next step.
Each format covers a different part of the 70-20-10 model, and practice repetitions are where they differ most. AI can support each format at a different stage, such as rehearsing a workshop’s frameworks or preparing for a mentor review.
No single format covers the full model, so the practical choice is a combination.
Sequence the plan by when a new manager first needs each conversation. Each phase pairs a real conversation with a PAL rehearsal before it happens.
Throughout, the mentor or coach reviews how each real conversation went, and the session transcripts show where to look.
Measure behavior on the job and use completion rates only as context. Six measures give a full picture.
Isolate the training’s effect where you can, because ROI Institute calls results without isolation “only evidence, not proof.” A PAL program also leaves a record of Objectives scores and transcripts for each session, which can be compared against managers who have not trained.
Start small and build around the three conversations new managers find hardest.
AI for managers belongs alongside human coaches, with each doing different work. A sensible split gives a PAL the repetitions and a human coach the debrief and the relationship.
Tavus builds PALs to supply those repetitions. A good first step is to pick one scenario, run it with your next cohort and compare how their first real conversations go.
See it for yourself. Book a demo.
No study gives a target count for feedback conversations. A workable rule is to repeat the scenario until every Objective is met in a single run, then do one final run a few days before the real meeting. Research on practice tests with students found that a practice run a few days before the real thing helped more than one on the same day.
Treat them as employee personal data. In the EU, Article 5(1)(f) of the AI Act has barred systems that infer emotions at work since 2 February 2025, except for medical or safety reasons. Get legal review before enabling perception features, and consider scoring from transcripts only.
A working list for a new manager’s first 90 days is feedback and coaching, one-on-ones, conflict, delegation and emotional intelligence. Sources define the five basic managerial skills differently, so treat this as a practical set for new managers. Each one is a conversation or habit a manager can rehearse before using it on a team.
Extend support past day 90 with a coach, a mentor and fresh scenarios built from the manager’s real stalls. A Gartner survey of nearly 3,200 managers found 40% of those with two years or less of experience were struggling to support their teams.