10 ChatGPT prompts that prep you for hard 1:1s in 2026

The 1:1s that haunt managers — performance concern, scope mismatch, promotion-not-yet, peer-conflict mediation, layoff news, comp disappointment, role redirect to IC, return-from-leave check-in, burnout signal, and "I'm thinking of quitting" — share one feature: the manager dreads them, so they write nothing down, walk in cold, and produce a worse outcome than the report deserved. Ten ChatGPT prompts close that gap. Each takes a small de-identified input, produces an opening line, a branching decision tree for the report's likely responses, a sample script, and the things you must not say. Grounded in Camille Fournier's <em>The Manager's Path</em>, Lara Hogan's resilient-management work, Kim Scott's <em>Radical Candor</em>, and Lattice's manager-effectiveness research.

<a href="https://aipromptshub.co/chatgpt-prompt-generator?utm_source=aipromptshub&utm_medium=blog&utm_campaign=hard-1on1s-2026" style={{display:"inline-block",padding:"10px 18px",background:"#0a66ff",color:"white",borderRadius:"6px",textDecoration:"none",fontWeight:"bold"}}> Generate a custom hard-1:1 prep prompt → </a>

Hard 1:1s fail at a predictable point: the opening 90 seconds. Lattice's 2025 Manager Effectiveness research found that reports' perception of fairness correlated more strongly with the specificity of the opening — naming the behavior, naming the impact, naming what changes — than with anything that happened later. Managers who walked in without a written opening defaulted to generic framing, the report heard "this is about me as a person", and the conversation lost its anchor.

Camille Fournier's The Manager's Path frames this as the "prepare for the hard conversation like you'd prepare for a board meeting" rule. Lara Hogan's resilient-management writing decomposes the opening into observation + impact + question. Kim Scott's Radical Candor framework provides the care-personally / challenge-directly axis the prompts use to keep tone from drifting brutal or mealy.

OpenAI's prompt engineering guide and the GPT-5 model documentation treat conversational pre-mortems as a structured-synthesis task. Each prompt below takes a small de-identified input (role, tenure, the specific behavior, the desired outcome) and emits four artifacts: opening line, decision tree on likely responses, sample script for the 70% case, and the boundary — the thing you will not say no matter how the conversation drifts.

The most-feared 1:1 is the one where performance is sliding and you have to name it before it crosses into PIP territory. The report's first interpretation will be "am I being fired?" — if you do not address that in the opening, the rest of the conversation runs on a fight-or-flight track.

**The prompt:**

``` You are an organizational psychologist prepping a manager for a performance-concern 1:1. INPUT (de-identified): - Report's role and tenure: <role, months in seat> - The specific behavior or output gap (not personality): <2-3 sentences, observed behavior only> - The impact on the team or product: <one sentence> - What has already been tried (if anything): <list> - Is this conversation a PIP precursor? <yes/no/not-yet-decided> - Manager-report relationship strength (1-5): <number> OUTPUT four artifacts: 1. OPENING LINE (1-2 sentences, max 35 words) using Lara Hogan's observation+impact+question structure. The report must hear "this is about specific behavior" not "this is about me as a person" within the first 20 words. 2. DECISION TREE on the 3 most likely report responses (defensive, agreeing-but-vague, asking-am-I-being-fired) with how to navigate each in 2-3 lines. 3. SAMPLE SCRIPT (~250 words) for the 70% case, written for a manager with relationship-strength <input>. End with a concrete next step (next 1:1 date + 1-2 measurable changes). 4. THE BOUNDARY: the 2-3 things the manager will NOT say no matter how the conversation drifts (e.g., do not promise no-PIP if PIP is not-yet-decided; do not give a percent-likelihood of being fired). Rules: - Frame behavior, never personality. "Submitting reviews late 3 of last 4 weeks" not "you seem disengaged". - If is_pip_precursor is "yes", the opening must explicitly say a formal performance plan is being prepared. Do not soften this. - If is_pip_precursor is "not-yet-decided", the opening must say "we are not at a PIP yet AND I want to be honest that we are closer than I want us to be". - Do not produce script lines that begin with "I'm just" or "I feel like" — they erode the signal. - Do not reference Radical Candor, Camille Fournier, or any framework by name in the script itself. The frameworks shape the script; they do not appear in it. ```

**Why it works:** Observation+impact+question in the opening prevents leading with a feeling ("I'm a little concerned") that the report reads as softening bad news. The PIP-status branch removes the manager's incentive to hedge — hedging produces the "you blindsided me" reaction at termination.

**Sample opening:** *"I want to talk about the last four sprints. Three of your reviews landed after the deadline, and the team has been merging without your sign-off, creating risk on next week's release. What's making the timing hard right now?"* — 38 words. Behavior, impact, question.

**When to use:** The day before. Write the opening down. Read it aloud once.

A Staff or Principal IC stops doing Staff-level work. The output is fine — but it's Senior-level output. The conversation has to recalibrate scope without sounding like a demotion.

**The prompt:**

``` You are a manager prepping a scope-mismatch 1:1 with a senior IC. INPUT (de-identified): - Report's level and tenure-at-level: <e.g., Staff, 14 months> - Three behaviors expected at their level that are not happening: <list, behavior only> - Three things they ARE doing well at the next level down: <list> - The strategic project that needs their level-of-impact: <one sentence> - Have they had a scope conversation before? <yes/no> OUTPUT four artifacts: 1. OPENING LINE (1-2 sentences, max 35 words) that anchors on the gap between current output and level expectations, not on the person. 2. DECISION TREE for these 3 likely responses: "the work I'm doing matters" (validate, then re-frame), "I'm not sure what Staff-level looks like here" (lean in — this is the win path), "I don't want the political work" (offer the IC track for that level if you have one; be honest if you don't). 3. SAMPLE SCRIPT (~250 words) that uses Camille Fournier's level-expectations framing without naming the book. End with two concrete behaviors to demonstrate by the next quarterly check-in. 4. THE BOUNDARY: do not promise a promotion or de-promotion in this conversation; do not benchmark against a specific peer by name. Rules: - Use "level expectations" not "performance" — they are different conversations. - Cite the strategic project as the place to demonstrate level, not a vague "more impact". - If the IC has never had this conversation, the opening must explicitly say "this is a recalibration, not a warning". ```

**Why it works:** Separating level from performance is the highest-leverage move in the Fournier playbook. Most scope-mismatch 1:1s fail because the report assumes "my work isn't good enough" when the message is "your work is good but it's not the work your level requires".

**Sample opening:** *"You're shipping high-quality Senior-level work. I want to recalibrate on the Staff-level work I haven't been seeing — specifically the cross-team architecture decisions on the data-platform migration. Can we map what that looks like?"* — 38 words.

**When to use:** A regular 1:1, not a special meeting. A special meeting elevates the stakes and triggers demotion anxiety.

The report has been pushing for a promotion. The promotion isn't happening this cycle. Saying "not yet" without a concrete delta turns the next quarter into resentment fuel.

**The prompt:**

``` You are a manager prepping a promotion-not-yet 1:1. INPUT (de-identified): - Report's current level and tenure-at-level: <text> - Two strengths that DO meet next-level bar: <list> - Two specific gaps that do not: <list, observable behaviors only> - Is the gap closeable in 1 cycle, 2 cycles, or unclear? <text> - Has the report been explicit about wanting promotion? <yes/no> - What signal did the calibration committee give? <one sentence> OUTPUT four artifacts: 1. OPENING LINE (1-2 sentences, max 35 words) that delivers the not-yet decision in the first sentence and the path forward in the second. Do not bury the decision. 2. DECISION TREE for 3 likely responses: "this isn't fair" (do not get defensive; restate the gap), "I'm going to look elsewhere" (acknowledge the validity; do not counter-offer in this meeting), "what do I do specifically" (the win path; lean in). 3. SAMPLE SCRIPT (~250 words) that names the 2 specific gaps with examples, the 2 specific demonstrations needed in the next cycle, and the timeline. End with the manager's commitment ("here's what I will do"). 4. THE BOUNDARY: do not say "definitely next cycle" unless calibration has signaled that explicitly; do not enumerate peers who were promoted; do not promise a stretch project you can't deliver. Rules: - The not-yet decision must arrive in the first 60 seconds. Hiding it inside a strengths-sandwich produces a worse outcome and a more bitter report. - Gaps must be behaviors with examples, not traits. "In the last 2 design reviews you deferred to the architect when you had the strongest opinion in the room" — not "you lack executive presence". - If calibration was a clean no with no path, say so. Do not invent a path. ```

**Why it works:** Lattice's research found that reports rate not-yet conversations as "fair" when the gap is specific and observable, "unfair" when abstract. Naming the manager's own commitment ("here's what I will do") is the Radical Candor care-personally lever applied to a hard decision.

**Sample opening:** *"You're not in this promotion cycle. There are two specific gaps the committee landed on, and I want to spend our time on what closes them in the next two quarters rather than on the decision itself."* — 40 words. Decision in sentence one, work in sentence two.

**When to use:** Within 48 hours of the calibration outcome. The report can sense the cycle ended.

Two of your reports are stuck. They've gone from disagreeing on technical decisions to avoiding each other in standup. You can't fire one. You need to mediate the next 1:1 — first with each separately, then together.

**The prompt:**

``` You are a manager prepping back-to-back conflict-mediation 1:1s with two reports. INPUT (de-identified): - Report A's role and tenure: <text> - Report B's role and tenure: <text> - The trigger event or pattern (de-identified): <2-3 sentences, observed behavior only> - The work that's blocked because of this: <one sentence> - Have you spoken to either individually about this before? <text> - Power asymmetry between A and B (e.g., one is more tenured, one has skip-level support): <text> OUTPUT four artifacts: 1. TWO OPENING LINES — one for the 1:1 with A, one for the 1:1 with B (max 35 words each). Each must name the observed pattern, the work impact, and ask for the report's view first. 2. DECISION TREE for the 3 likely responses you'll hear from each separately: "the other person is the problem" (validate without endorsing), "I don't think this is a big deal" (name the blocked work), "I want them off the team" (do not commit; surface the request neutrally). 3. SAMPLE SCRIPT (~300 words) for the JOINT 1:1 you'll run after the two individual ones. Open with a shared statement of the work that's blocked. Have each share the observed pattern from their side using "I noticed X, the impact on me was Y" structure. Close with one specific changed behavior from each and a 30-day check-in. 4. THE BOUNDARY: do not take sides in the individual 1:1s; do not promise the other will change; do not skip the joint 1:1 because the individuals feel resolved. Rules: - The opening must avoid the word "conflict" — it makes both reports defensive. Use "pattern" or "friction". - The joint 1:1 must have a structured format. Unstructured "let's talk it out" sessions reliably make peer conflict worse. - If there's a real power asymmetry, name it in your own prep, not in the meeting. ```

**Why it works:** Lara Hogan's conflict writing centers on this structure: name behavior, name impact, ask first. Individual 1:1s before the joint one prevent the ambush dynamic — neither report should hear the other's framing for the first time at the joint meeting. The 30-day check-in closes the loop that otherwise reopens within a week.

**Sample joint opening:** *"The data-platform migration is blocked because the two of you are routing around each other. We'll spend 30 minutes on what each of you observed, what the impact was, and one thing each of you will do differently."* Names the work, the cost, the structure.

**When to use:** Individual 1:1s same day, joint 1:1 within 48 hours after. Longer gaps let positions harden.

The hardest 1:1 a manager will ever run. The decision is not yours, the timing is not yours, but the conversation is. This prompt prepares the 15-minute call that respects the human you're talking to.

**The prompt:**

``` You are a manager prepping a layoff-delivery 1:1. The decision is final and not yours to reverse. INPUT (de-identified): - Report's role and tenure (years/months): <text> - The official reason language HR has approved you to use: <verbatim, max 30 words> - The severance and benefits package summary: <text> - Who else is being affected on the team and when they will be told: <text> - Will the report be expected to keep working? <yes — for how long / no — effective immediately> - Day-of logistics (laptop return, accounts, timeline): <bullet list> OUTPUT four artifacts: 1. OPENING LINE (max 30 words) that delivers the news in the first sentence. The report should know within 15 seconds that their role is ending. Do not start with small talk. 2. DECISION TREE for the 3 likely responses: silence (sit with it, do not fill the silence), anger (acknowledge, do not defend the decision), questions about why (use only the HR-approved reason language verbatim — do not extrapolate). 3. SAMPLE SCRIPT (~280 words) covering: news, reason (verbatim HR language), package summary, day-of logistics, and what you as their manager will do to help them next (specific intros, references, recommendation letter). End with the next concrete step in the next 24 hours. 4. THE BOUNDARY: do not apologize for the decision in a way that implies you disagree publicly; do not predict outcomes for other team members; do not extend the conversation past 15 minutes; do not offer your personal opinion on the decision; do not use the phrase "this is hard for me too". Rules: - Use the HR-approved reason language exactly. Improvised reasons create legal exposure for the company and confuse the report. - The script must include the manager's specific post-layoff support commitments. "I will introduce you to three people in my network this week" is the difference between a delivery and a betrayal. - Never schedule this conversation as the last 1:1 of the day. The report needs the rest of their day. So do you. - Never deliver layoff news over Slack, async, or written-only. The 1:1 is the minimum standard. ```

**Why it works:** The 15-second rule comes from resilient-management writing — reports who heard the news within 15 seconds rated the conversation "hard but respectful" at much higher rates than those who heard it after small talk. The HR-verbatim rule protects the report from extrapolated reasons that contradict the official narrative.

**Sample opening:** *"I have hard news. Your role is being eliminated as part of a reduction across the company, effective two weeks from today. I want to walk you through what this means and what I'll do to help."* — 38 words. News and commitment in the same breath.

**When to use:** First thing in the morning, not Friday afternoon, not before a weekend. The report needs a workday to act and absorb with daylight ahead.

Comp letters went out. Your report's number is below their expectation. The cycle is closed. The conversation isn't a negotiation; it's a recalibration of trust.

**The prompt:**

``` You are a manager prepping a comp-disappointment 1:1. INPUT (de-identified): - Report's role and tenure: <text> - Was the comp change a raise, flat, or below-expectation: <text> - The compensation philosophy your company uses (market percentile, equity emphasis, etc.): <one sentence> - The two factors that drove the specific number (rated calibration, market move, budget): <list> - Has the report been explicit about a specific comp expectation? <yes/no — and what was it, rounded> - Is there any flexibility (stock, title, bonus timing)? <yes/no> OUTPUT four artifacts: 1. OPENING LINE (max 35 words) that acknowledges the report likely sees the number as low and signals the conversation will address it directly. Do not pretend you don't know they're disappointed. 2. DECISION TREE for these 3 likely responses: "I'm underpaid vs market" (engage with data if you have it; do not handwave), "my peer makes more" (do not confirm or deny; redirect to their package), "I'm going to look" (acknowledge validity; do not counter unless you have explicit flexibility). 3. SAMPLE SCRIPT (~250 words) that names the two driver factors, walks the report through how comp decisions get made at your company, and ends with one concrete thing: a comp review date, a flexibility offer, or an honest "this number is fixed for the cycle, here's what could change next cycle". 4. THE BOUNDARY: do not promise next-cycle outcomes; do not share peer compensation; do not blame HR or the budget committee in a way that implies you would have given more; do not say "I fought for you" unless you can describe the specific advocacy. Rules: - Use rounded numbers and percent changes, not exact figures, unless the report introduces them first. - The opening must signal that the meeting is for the report to be heard. Comp conversations where the report doesn't speak in the first 5 minutes feel like a verdict reading. - If you advocated and lost, you can say "I advocated for a different number" — do not say what number, do not blame the decision-maker by name. ```

**Why it works:** Comp 1:1s erode trust at the opening, not at the number. Reports who hear "I know this is disappointing" before the manager defends the process rate the conversation fair more often. The peer-comp guardrail is non-negotiable — confirming or denying a peer's number creates a downstream cascade.

**Sample opening:** *"The number is below where I think you wanted to land. I want to talk through how the decision got made and what's possible from here — including what's not changing this cycle."* — 36 words. Acknowledges the gap, signals honesty.

**When to use:** Within one week of the letter, never on the day it lands. Week 2 is the window.

A reluctant manager is struggling. They miss the IC work and their reports are suffering for it. The conversation has to offer the IC track as a real option, not a demotion in disguise.

**The prompt:**

``` You are a manager prepping a role-redirect 1:1 with a struggling manager who would be a stronger IC. INPUT (de-identified): - The struggling manager's tenure-in-role: <text> - Two patterns showing they're struggling with management work (behaviors, not personality): <list> - Two patterns showing they were stronger as an IC: <list> - Does your company have a credible Staff+ IC ladder? <yes/no> - Has the struggling manager hinted at missing IC work? <yes/no> - What signal did calibration or skip-level give? <text> OUTPUT four artifacts: 1. OPENING LINE (max 40 words) that frames this as a fit conversation, not a performance conversation. The redirect should be presented as a strong option among real options, not as the only path. 2. DECISION TREE for these 3 likely responses: "I'm being demoted" (separate fit from performance, name the IC level explicitly), "I want to keep trying" (offer a 90-day plan with measurable behavior changes IF the company can support that — do not offer it if it's not real), "actually I want this" (the win path; move into specifics fast). 3. SAMPLE SCRIPT (~280 words) that names the two struggle-patterns and the two strength-patterns side by side, frames the IC track honestly (level, comp implication, scope), and ends with a decision-by date no longer than 2 weeks out. 4. THE BOUNDARY: do not present the IC track if your company doesn't have a real Staff+ ladder — that's a face-saving demotion and the report will sense it; do not promise the same comp band if it doesn't apply; do not let the conversation end without a decision-by date. Rules: - Use the phrase "role fit" not "role change". Fit reads as situational; change reads as judgment. - Name the comp implication, even if it's neutral. Leaving comp ambiguous in this conversation creates anxiety that crowds out the actual decision. - If your IC ladder caps below their current manager level, this is not a redirect conversation, it's a transition-out conversation. Do not run the redirect script as cover. ```

**Why it works:** The most common failure in IC-redirect conversations is dressing a demotion in development language. Reports detect it within minutes. Naming the comp implication and IC level explicitly is the only path to the report experiencing this as a real choice.

**Sample opening:** *"I want to have a fit conversation. The last six months have shown me where management work is taxing for you and where IC work makes you come alive. I want to talk about both honestly and what options are real."* — 42 words. Fit, both sides, options not verdict.

**When to use:** After two performance-shaped conversations and a stable pattern. Running this 1:1 too early bypasses developmental work the manager deserves.

A report is back from parental, medical, or bereavement leave. The hard 1:1 is the first one — too brisk feels like the leave didn't matter; too soft feels like they're being treated as fragile.

**The prompt:**

``` You are a manager prepping a return-from-leave 1:1. INPUT (de-identified): - Report's role and tenure: <text> - Type of leave: <parental | medical | bereavement | sabbatical | other> - Length of leave: <weeks> - What changed on the team while they were out (org, headcount, roadmap, leadership): <bullet list> - Is there a phased-return plan or accommodation in place? <text> - The first deliverable on their plate when they're back: <one sentence> OUTPUT four artifacts: 1. OPENING LINE (max 30 words) that welcomes them back and signals you've thought about what they need this week, not just what the team needs. 2. DECISION TREE for these 3 likely responses: "I'm ready, give me everything" (slow it down without infantilizing — first week is for context, week 2 is for ownership), "I'm overwhelmed" (name the phased plan or build one in the meeting), "can we talk about what happened" — appropriate for bereavement or medical (listen; do not redirect to work prematurely). 3. SAMPLE SCRIPT (~250 words) covering: what changed on the team (factual, brief), what they own this week (light), what they own week 2-4 (ramping), the accommodation or phased-return reality, and a 30-day check-in date. 4. THE BOUNDARY: do not ask about the personal reason for the leave unless they raise it; do not relay team members' opinions about the leave; do not promise a project on day one that you'll later have to walk back; do not skip a return 1:1 because "they seem fine". Rules: - The first 5 minutes belong to the report, not the org update. Reverse that order and the rest of the conversation lands wrong. - Use "what changed" not "what you missed". The first signals information; the second signals deficit. - For medical and bereavement leaves, the phrase "there's no rush, and there are also no shortcuts" is honest and kind. Use it if the situation fits. ```

**Why it works:** Returns are mis-managed at both extremes — overloaded and tip-toed-around. Building the phased plan into the conversation, with concrete week-1 and week-2-4 ownership, gives the report both agency and protection. Lattice's data shows return-from-leave 1:1s are lowest-rated when the manager defaults to either extreme.

**Sample opening:** *"Welcome back. I want to spend the first part of this hour on you and what kind of week you need, and the second part on what changed and what you'll be picking up — in that order."* — 40 words. Centers the report.

**When to use:** The first 1:1 back. Not day one — day one is for logistics.

You're seeing the pattern: missed deadlines, withdrawn in meetings, late-night Slack messages, weekend commits. You believe they're burning out. The 1:1 has to surface it without putting words in their mouth.

**The prompt:**

``` You are a manager prepping a 1:1 to surface a possible burnout pattern. INPUT (de-identified): - Report's role and tenure: <text> - The 3 specific behavior changes you've observed in the last 4-8 weeks: <list, observed behavior only, with rough dates> - The workload reality (over-allocated, on-call rotation, recent on-call, etc.): <one sentence> - Did the report just ship a major project, lose a teammate, or absorb new scope? <text> - Is the manager-report relationship strong enough to ask direct questions? <1-5 scale> - What can you actually offer (time off, scope reduction, on-call swap)? <list — be honest about what is real> OUTPUT four artifacts: 1. OPENING LINE (max 40 words) that names the 3 observed behavior changes specifically, frames them as a question not an accusation, and signals you're not asking them to perform wellness. 2. DECISION TREE for these 3 likely responses: "I'm fine" (offer one observation back, do not push past a soft no), "yes, I'm exhausted" (move to specifics — what would help in the next 2 weeks), "this is just the job" (validate the constraint, then name what you can actually change). 3. SAMPLE SCRIPT (~250 words) that uses observation+impact+question structure for the burnout signal, offers the 2-3 specific things you can actually do (from input), and ends with one concrete change to take effect within 7 days. 4. THE BOUNDARY: do not diagnose; do not require disclosure; do not promise a workload reduction you can't deliver; do not push a soft no into a yes — the report owns their own framing. Rules: - Open with the observed behavior, not the word burnout. "Burnout" lands as a label; observed behavior lands as care. - Offer only what you can deliver. "Take some time off" with no operational plan is not an offer; it's a wish. - Respect the soft no. If the report says "I'm fine" twice, the meeting becomes a check-in not an excavation. Schedule a follow-up. ```

**Why it works:** Naming observed behavior, not naming burnout, is what distinguishes a caring 1:1 from an awkward one. Observations open the door; labels slam it. Offering only real interventions is the credibility test the report uses to decide whether to engage.

**Sample opening:** *"I've noticed you've been online past 11 most nights for three weeks, you were quieter than usual in the architecture review, and you missed standup yesterday. I'm not assuming I know why — I want to ask how you're doing and what would actually help."* — 47 words. Three specific observations, an explicit refusal to assume.

**When to use:** Inside the regular 1:1 if emerging; outside if acute. A scheduled-around meeting signals stakes the report didn't ask for.

Your report says they're considering leaving. The wrong moves — panic, counter-offer reflex, guilt — will accelerate the departure. The right move is a conversation that earns honesty.

**The prompt:**

``` You are a manager prepping a 1:1 in response to a report signaling they may leave. INPUT (de-identified): - Report's role and tenure: <text> - How did the signal arrive (direct statement, indirect hint, second-hand): <text> - What you believe the top 2 drivers are based on the last 90 days (comp, growth, manager, project, life event, recruitment): <list> - Is there a project, decision, or comp action in flight that bears on this? <text> - What can you ACTUALLY change (not what you wish you could): <list> - Are you ready to accept the resignation if it comes? <yes/no> OUTPUT four artifacts: 1. OPENING LINE (max 35 words) that thanks the report for the honesty, signals you're not going to react with panic or pressure, and asks them to walk you through their thinking — without leading. 2. DECISION TREE for these 3 likely responses: "I've decided to leave" (accept it; ask about transition), "I'm exploring" (the engagement zone; ask what would have to be true to stay), "I want something specific" (engage on the specific; do not promise on the spot). 3. SAMPLE SCRIPT (~280 words) that listens first (~half the script is structured questions), names what you can actually change (no fabrication), is honest about what you can't, and ends with a follow-up in 5-7 days regardless of which way the conversation leans. 4. THE BOUNDARY: do not counter-offer in the meeting; do not guilt the report on team impact; do not promise a promotion; do not threaten that recruiters will be in touch with HR; do not let the conversation end without naming the next conversation. Rules: - The first half of the meeting is questions, not answers. Managers who fill the air with retention pitches accelerate departures. - Counter-offers should be discussed with HR before the meeting, not invented in it. - If the report has already decided, switch to a graceful-transition conversation. Trying to talk them out of a decided exit damages the relationship that survives the departure. - Honesty about what cannot change is more retentive than vague promises about what might. ```

**Why it works:** The "thinking of quitting" 1:1 is won by the manager's first 90 seconds. Panic accelerates the exit; structured curiosity gives the report room to say what's actually wrong. Honesty about what cannot change preserves trust whether they stay or go — the relationship that survives a departure delivers referrals and reputation later.

**Sample opening:** *"Thank you for telling me. I'm not going to try to talk you out of anything today. I want to understand what's driving this, and I'll be honest about what I can and can't change."* — 36 words. Courage acknowledged, pressure removed.

**When to use:** Within 24 hours. Delay reads as disinterest; same-day as overreaction. Next-day is the window.

The comparison table below summarizes each prompt's scenario, the report behavior or signal that triggers it, and the artifact set the prompt produces. The 10 prompts collectively cover the conversations Lattice's manager-effectiveness research identified as the highest-stakes and lowest-prepared.

<a href="https://aipromptshub.co/blog-post-outline?utm_source=aipromptshub&utm_medium=blog&utm_campaign=hard-1on1s-compare" style={{display:"inline-block",padding:"10px 18px",background:"#0a66ff",color:"white",borderRadius:"6px",textDecoration:"none",fontWeight:"bold",marginTop:"12px"}}> Build a custom hard-1:1 prep template → </a>

The chain that turns dread into prep:

1. **T-24h (5 min human).** Pick the prompt matching the scenario. Write the de-identified input as a paragraph — role descriptor, observed behavior, what's at stake, what you can actually offer. No real names, no salary figures tied to a person, no verbatim review snippets. 2. **T-24h (3 min compute).** Run the prompt. Read the four artifacts once. 3. **T-24h (5 min human).** Edit the opening line to sound like you. The script is shape; the voice is yours. 4. **T-12h (5 min human).** Read the opening line aloud. If it doesn't feel honest, rewrite. Do not walk in with words you don't believe. 5. **T-1h (3 min human).** Re-read the decision tree. Rehearse the script for the response you most fear. 6. **T-0.** Open with your line. Listen. Use the tree as a map, not a script. 7. **T+24h (5 min human + 3 min compute).** Re-run the prompt with the actual outcome, asking for a written follow-up message that anchors what was agreed. The written follow-up does as much as the conversation itself.

Total: ~30 minutes of prep across two sittings, replacing the 0 minutes that produces walked-into-cold disasters. Compute cost runs under $0.10 per 1:1 at current OpenAI pricing.

<a href="https://aipromptshub.co/chatgpt-prompt-generator?utm_source=aipromptshub&utm_medium=blog&utm_campaign=hard-1on1s-chain" style={{display:"inline-block",padding:"10px 18px",background:"#0a66ff",color:"white",borderRadius:"6px",textDecoration:"none",fontWeight:"bold",marginTop:"12px"}}> Get the 30-minute hard-1:1 prep template → </a>

### Which ChatGPT model should I use for hard-1:1 prep?

GPT-5 is the right default for prompts 1, 5, 9, and 10 — performance, layoff, burnout, quit-risk — because script tone is the entire artifact. GPT-5-mini handles prompts 2, 3, 4, 6, 7, and 8 because they are structurally tighter. See the OpenAI model documentation for context windows and pricing.

### Is it safe to put my report's information into ChatGPT?

Treat the input as if it were a public document. Use role descriptors, rounded numbers, and observed-behavior snippets — never names, salary figures tied to a person, or verbatim review text. OpenAI's data usage policies cover what's retained; your company's AI policy binds on top. If a private LLM deployment exists, run these there.

### Will the script sound like me?

Not on the first pass. The prompts produce structure; the voice is yours. The 30-minute prep ritual above includes a step to rewrite the opening line in your voice — do not skip it. A script that sounds like a manual lands as a manual.

### What if the conversation goes off-script?

It will. The decision tree handles the three likeliest responses; real reports produce a fourth. Fall back on the structure — observation, impact, question; listen first — rather than searching for the prompt's exact line. Camille Fournier's The Manager's Path treats this same way.

### How is this different from running these conversations with HR?

HR sets policy and compliance language; the prompts prep the human delivery. They work together — the layoff prompt requires HR-approved reason language verbatim. HR is the source of truth on what you can say; the prompt structures how you say it.

### Do these prompts work for skip-level 1:1s?

They work for direct reports as written. For skip-levels, drop the relationship-strength input, expand the boundary list, lean more on observation than interpretation, and end with smaller commitments.

### What if I don't have the relationship strength to run these honestly?

Run relationship-strength as a 1 or 2. The script will be more cautious, lean on observation, and end with smaller commitments. Hard 1:1s with low-trust reports are smaller in scope and longer in horizon, not impossible. Lara Hogan's resilient-management writing covers the rebuild work.

- Camille Fournier, *The Manager's Path* — level-expectations framing and hard-conversation preparation discipline. - Lara Hogan, resilient-management writing — observation + impact + question feedback equation, noticing in 1:1s. - Kim Scott, *Radical Candor* — care-personally / challenge-directly axis used to keep tone calibrated. - Lattice Manager Effectiveness research (2025) — opening-specificity correlation with perceived fairness; return-from-leave ratings. - OpenAI prompt engineering guide — structured-synthesis patterns. - OpenAI model documentation — GPT-5 and GPT-5-mini selection. - OpenAI pricing — per-1:1 compute cost reference. - OpenAI data usage policies — input-handling and retention guidance for sensitive employee data.

---