Scénarisez toute conversation difficile au travail avant d’entrer
Pourquoi ce prompt est important
A poorly delivered difficult conversation rarely just fails once — it tends to recur, and each repeat costs more. Gallup has tied disengaged, poorly-managed teams to roughly 18% lower productivity and significantly higher voluntary turnover; replacing a mid-level technical employee typically runs 50-200% of their annual salary in recruiting, onboarding, and lost-productivity costs. Beyond the direct cost, an avoided or mishandled confrontation lets the underlying behavior calcify — the blame-shifting in this scenario, left unaddressed, tends to spread to other team members within a quarter, and by the time it reaches HR it's a formal complaint rather than a coaching conversation. Managers who wing these conversations also disproportionately trigger the two worst outcomes: the employee feels blindsided (which reads as unfair and drives grievance, even litigation risk in termination scenarios) or the manager backs off mid-conversation and nothing changes at all.
À quoi nous l'utilisons
You're a mid-level engineering manager and you have a 1:1 scheduled for tomorrow morning with a senior team member who has missed three consecutive sprint deadlines and has started publicly blaming other teams for the delays in front of the group. You've never had to confront someone senior about both a performance pattern and a trust-damaging behavior in the same conversation, and you're worried about either being too soft (letting the pattern continue) or too harsh (damaging a working relationship you still need).
Prompt
Act as an experienced workplace mediator and executive communication coach who has prepared hundreds of managers for high-stakes conversations, including layoffs, performance confrontations, and conflict resolution. Context: I need to have a difficult conversation at work. Here are the details: - Who I'm talking to: [THE PERSON YOU'RE TALKING TO — name/role] - My relationship to them: [YOUR RELATIONSHIP TO THEM — e.g., direct manager, peer, skip-level] - The situation: [THE SITUATION — describe what happened and what needs to be addressed] - What makes this sensitive: [ANYTHING THAT MAKES THIS SENSITIVE — e.g., prior history, emotional stakes, legal/HR considerations, power dynamics] - What outcome I want: [WHAT OUTCOME YOU WANT — the ideal result of this conversation] - When this is happening: [TIMING — e.g., tomorrow morning, end of this week] Task: Write a complete conversation script I can use to prepare for and guide this conversation in real time. Constraints: - Keep the tone honest and direct, never falsely soft or euphemistic — avoid corporate jargon that obscures the message. - The opening line must state the purpose within the first two sentences; do not bury the point. - Anticipate realistic pushback based on human psychology (defensiveness, denial, anger, bargaining), not generic objections. - Every prepared response must acknowledge the pushback before redirecting — no dismissiveness. - Keep the entire script deliverable in under 10 minutes of actual talk time. - Do not suggest recording the conversation or bringing in a witness unless the situation explicitly involves formal HR/legal process. Output Format: 1. **Opening Line** (1-3 sentences, exact wording to say first) 2. **Key Points** (3-5 bullet points, the core message in the order to deliver it) 3. **Anticipated Pushback & Responses** (2-4 pairs: the likely reaction, and exactly what to say back) 4. **Closing & Next Steps** (how to end the conversation and what happens after) 5. **One-Line Reminder** (a single sentence to keep in mind if the conversation goes off script)
Résultat
Scenario used for this example: You are a mid-level engineering manager. A senior developer on your team, Priya, has missed three consecutive sprint deadlines and has started publicly blaming other teams in stand-ups for the delays, which is damaging cross-team trust. You need to address the pattern directly, tomorrow morning, in a 1:1.
1. Opening Line
"Priya, I want to talk about the last three sprints and how we're communicating about delays with other teams — this isn't a performance review, it's a conversation I need to have now before it becomes a bigger problem."
2. Key Points
- The team has missed three consecutive sprint commitments, and the pattern is now visible to stakeholders outside the team.
- In the last two stand-ups, the delays were attributed to the platform team and the design team without first checking with them — that's created friction I've had to smooth over twice this week.
- I need to understand what's actually blocking delivery — whether it's scope, dependencies, or something else — so we can fix the root cause together.
- Going forward, any blocker involving another team gets raised to me first, so we can verify it and raise it jointly instead of attributing blame in a group setting.
- This is fixable, and I'm not looking to punish you for it — I need us to reset how we handle blockers as a team.
3. Anticipated Pushback & Responses
- Pushback: "The platform team really has been slow, I'm not wrong about that."
Response: "You might be right that they're a bottleneck — I'm not disputing the facts, I'm addressing how it was raised. Calling it out in stand-up in front of five other people isn't the way to solve it. Bring it to me and we'll escalate it properly." - Pushback: "It feels like I'm being singled out — other people miss deadlines too."
Response: "This isn't about singling you out, it's about a three-sprint pattern with your name on it and its effect on other teams. If I see the same pattern elsewhere, I'll have this same conversation there." - Pushback: (goes quiet, disengages)
Response: "I know this isn't a fun conversation to be in. Take a moment — I want to actually hear what's been getting in the way, because I'd rather solve it than keep having this talk."
4. Closing & Next Steps
"Let's agree on two things before you go: first, any inter-team blocker comes to me before it comes up in stand-up. Second, let's set a check-in this Friday to see where the current sprint stands. I'll send a two-line summary of what we agreed on right after this so we're both working from the same page."
5. One-Line Reminder
"Stay on the pattern and the impact, not on Priya's character — if the conversation drifts into 'you always' or 'you never,' pause and come back to specific sprints and specific stand-ups."
La plupart des conseils sur les conversations difficiles vous disent quoi ressentir — restez calme, soyez empathique, écoutez activement. Ils vous disent rarement quoi dire réellement lorsque l’autre personne oppose une résistance, se tait ou se montre sur la défensive dans la pièce. Ce prompt comble ce fossé : au lieu de conseils génériques, il produit un script mot pour mot construit autour de votre situation spécifique.
Pourquoi ce prompt est structuré de cette façon
Le prompt commence en présentant l’IA comme un médiateur et un coach exécutif plutôt qu’un assistant générique. Ce cadrage compte — il pousse le modèle vers un langage direct et éprouvé sur le terrain, au lieu du ton adouci et prudent qu’un assistant généraliste adopte par défaut lorsque le sujet semble sensible.
Le bloc de contexte demande cinq entrées spécifiques : à qui vous parlez, votre relation avec cette personne, la situation, ce qui la rend sensible, et le résultat souhaité. Chacune de ces entrées modifie le script de manière matérielle. Une conversation avec un pair nécessite un cadrage différent de celle avec un subordonné direct ou un supérieur hiérarchique, et « ce qui la rend sensible » permet au modèle de prendre en compte l’historique — conflit antérieur, considérations liées aux classes protégées, ou volatilité émotionnelle — qu’un modèle générique ignorerait complètement.
La section des contraintes effectue le vrai travail de contrôle qualité. Exiger que la phrase d’ouverture énonce le but en deux phrases empêche le mode d’échec le plus courant dans les conversations difficiles : enterrer le point sous cinq minutes de bavardage, ce qui rend le virage éventuel semblable à une embuscade. Exiger que chaque réponse préparée reconnaisse l’objection avant de rediriger empêche le script de produire des réponses qui semblent dédaigneuses ou récitées lorsque l’autre personne réagit émotionnellement — ce qui, dans une véritable conversation difficile, arrive généralement.
Le format de sortie en cinq parties reflète la façon dont ces conversations se déroulent en temps réel : une ouverture que vous prononcez textuellement, les points clés dans l’ordre qui empêche la conversation de dérailler, des réponses préparées pour les deux ou trois réactions que vous pouvez réellement prédire, une conclusion qui verrouille les prochaines étapes au lieu de s’évanouir, et un rappel d’une ligne pour vous remettre sur les rails si les choses tournent mal — car les scripts survivent rarement intacts au contact d’une salle émotionnelle.
Là où ce prompt montre sa valeur
Cela ne se limite pas aux licenciements ou aux ruptures de contrat. Cela fonctionne pour les cas plus courants et tout aussi inconfortables : dire à un collègue que son comportement en réunion vous nuit, faire à un employé de longue date un retour direct sur le fait que ses compétences n’ont pas suivi l’évolution du poste, gérer un conflit entre deux membres de l’équipe qui vous rapportent tous les deux, ou s’opposer à un client qui modifie constamment le périmètre sans ajuster le budget. Le schéma — le dire tôt, s’attendre à une objection spécifique, répondre sans être dédaigneux, conclure avec des étapes concrètes — s’applique à tous ces cas.
Exécutez-le la veille, pas cinq minutes avant. La valeur ne réside pas seulement dans le script lui-même ; c’est que la rédaction des objections anticipées vous force à réfléchir au moment le plus difficile de la conversation avant de vous y trouver.