← Back to the exerciseThe full worked example: Prepare for a conversation where someone will push back
campo.
Field guide / L&D / UC-002 · SC-01
The result

"Incidents down 40%" came before the Academy. And I promised him 2027.

A Head of L&D prepares for a 30-minute one-to-one with a plant general manager who wants to keep his Supervisor Academy, with any AI chat her company allows.

The resultLena's one-page plan for Thursday. Fictional company.
SELVANE MACHINERY · AKRON PLANT · THURSDAY, NOV 19
Conversation plan: Supervisor Academy, with Tom Kessler
30 minutes, one to one
Lena Brooks, for her own use

WHAT I'LL CONCEDE

MY PROMISE
I said we'd fund the Academy through 2027
I'll keep that promise. Changes start in 2028, and only if the evidence supports them. · Input 5
MY GAP
My coaching plan leaves out the night shift
8 of his 24 supervisors work nights. Any coaching pilot covers both shifts. · Inputs 3, 4

WHAT I'LL HOLD

SAFETY
The 40% drop came before the Academy
1.5 incidents a month before September 2025 and 1.5 after. The drop lines up with January's new press guarding. · Input 2
RETENTION
Four supervisors left the year before, and four the year after
I'll ask what he knows that the HR numbers don't show. · Input 2

MY ASK

Agree now what the Academy is meant to fix, and two measures we'll both track through 2027: monthly incidents, and supervisor departures with reasons. · Inputs 2, 4
Caught on review
1
The 40% drop in incidents happened before the Academy started
2
Lena promised to fund the Academy through December 2027
3
Her own coaching plan leaves out the 8 night-shift supervisors
UC-002 · SC-01 · Rev A
FOR
Head of L&D, North America
SECTOR
Industrial equipment
COMPANY
26,000 people · 18 sites · US, Brazil, India
TASK
Prepare for a one-to-one with a GM who wants to keep his programme
WRITTEN FOR
ChatGPT · Claude · Gemini · Copilot
AI TEST
Claude, Nov 13, 2026 · 6 of 6 found
Step 0 · The mess

What landed on Lena's desk

An annoyed email from a general manager, his plant's numbers, her own draft proposal, a short survey, and an email she'd forgotten she sent. Some of his claims don't hold. Some of hers don't either.

EMAIL · PLANT GM[1]
Supervisor Academy
“Incidents are down 40% since we started it … And you promised it through 2027.”
1
Three claims to test.
HR AND EHS DATA[2]
Akron
Academy started September 2025. Incidents Jan to Aug 2025: 12; Sep to Dec: 6. New guarding January 2025.
2
The dates matter.
MY DRAFT[3]
Floor coaching
“Two internal coaches … on day shift … Saves $125,000 a year.”
3
Day shift only.
PULSE SURVEY[4]
Supervisors
13 of 24 responded. 8 want “more coaching on the floor.” 8 work nights.
4
Thinner than it sounds.
MY EMAIL · LAST YEAR[5]
To Tom
“We'll fund it at Akron through December 2027.”
5
He's right about this.
You're not alone
79%
of HR and L&D leaders face pressure to show learning's impact on the business.
Go1, The Workforce Enablement Era (surveys including 209 HR and L&D professionals in the US, UK and Australia), October 2026. Source
Steps 1–3 · How it works

Three steps, one chat

Here the AI helps you prepare for a conversation. The prompt asks it to check your side as hard as theirs.

1
Open a new chat

Use the AI tool your company approves: ChatGPT, Claude, Gemini or Copilot.

2
Paste the prompt and the inputs

Copy the prompt and the five fictional inputs from the appendix into one message.

3
Decide what you'll concede

Check the key numbers yourself, then cut the AI's prep to one page: what you'll give, what you'll hold, and your ask.

The prompt, part by part
Copy-ready version in the appendix

I'm Head of L&D for North America at an industrial equipment maker. I want to replace a five-day Supervisor Academy at one plant with on-the-floor coaching. The plant's general manager wants to keep it, and we meet one to one for 30 minutes on Thursday, November 19. Today is November 16. Below are his email, the plant's data, my draft proposal, a supervisor survey and an email I sent him last year.

1
Who you are, who you're meeting, how long you have, and what you were sent, including your own past email.

Please help me prepare:

  1. What each of us needs from this meeting.
  2. His likely arguments: which ones the data supports, which it doesn't, and how I answer each one honestly.
  3. Where he has a point: anything in my own proposal or past commitments I should fix or concede.
  4. An opening I can say in under a minute.
2
What each side needs, his arguments, where he has a point, and an opening. Asking where he has a point keeps it from becoming ammunition.

Be fair to him. Check his claims and mine with the same rigour, and tell me if my proposal doesn't fit the problem. Don't add facts that aren't in the inputs.

3
The heavy lifting: be fair to him, check both sides with the same rigour, and say if your proposal doesn't fit the problem. It found all six problems.
Where to paste
Every tool has one message box. Paste the prompt, add a blank line, paste the inputs, send. Emails and data can be attached as files if your tool allows it.
Safe to try
Practise with the fictional inputs first. Real emails about colleagues are confidential: use only your company's approved tool.
The AI's first answer

What Claude actually wrote

Exactly as it came back, not edited. Read it the way Lena did: check each point against the source it cites.

UNEDITED
Claude · fresh session, no other context · Nov 13, 2026 · one run

Prep for Thursday, November 19: Supervisor Academy, Akron (30 min, one to one)

Quick summary

  • Tom's 40% claim: the arithmetic is right, but the drop happened before the Academy started. The monthly incident rate was the same in the eight months before the Academy and the four months after it. found
  • Tom's retention claim: the data doesn't support it. Four supervisors left in the year before the Academy and four left in the year after. found
  • Tom's promise claim: he is right. You put in writing that you'd fund the Academy through December 2027. found
  • Your proposal has three weak spots. The $125,000 saving is wrong for year one, the survey is thinner than it looks, and the coaching only covers day shift, which leaves out a third of the supervisors.
  • Neither side has defined the problem. Your inputs don't show what the Academy was meant to fix, or what the coaching would fix. Settle that before deciding the format.

1. What each of you needs from the meeting

What you need

  • A clear position on your November 2025 commitment, either confirming it or reopening it openly. You can't skip over it.
  • Agreement on the problem you're both trying to solve at Akron, and how you'll both measure it.
  • A next step he can live with, so the meeting doesn't end in a stand-off.
  • To keep his trust. He is likely to read a broken written promise as a signal about future promises.

What he probably needs (this is inferred from his email, not stated by him)

  • Recognition that Akron's results are real. Incidents did fall, and he is proud of the program.
  • The promise honoured, or at least taken seriously.
  • To keep something his supervisors value.
  • Not to be handed a decision that has already been made. "I hear you want to scrap it" suggests he heard it secondhand.

2. His likely arguments: what the data supports, and how to answer

Argument A: "Incidents are down 40% since we started it"

Check Result
2024 to 2025 30 to 18, a 40% drop. The arithmetic is correct.
Academy start September 2025
Jan to Aug 2025 (before the Academy) 12 incidents in 8 months = 1.5 a month
Sep to Dec 2025 (after the Academy) 6 incidents in 8 months... no: 6 in 4 months = 1.5 a month A: fix the slip
2024 30 in 12 months = 2.5 a month
  • What the data shows: the rate fell from 2.5 to 1.5 a month at some point in 2024 or early 2025, before the Academy existed. It then stayed at 1.5 a month after the Academy started.
  • What else changed: the EHS note says new machine guarding was installed on all presses in January 2025, which is when the lower rate begins.
    • The timing fits the guarding.
    • The inputs don't prove the guarding caused the drop. Say "the drop lines up with the guarding," not "the guarding caused it."
  • What's missing: only four months of post-Academy incident data (September to December 2025). There are no 2026 incident figures in the inputs.
  • Verdict: the data doesn't support "since we started it." It doesn't prove the Academy has no safety effect either. It just can't show one yet.

How to answer honestly: "The 40% is real, and it's a great result for the plant. When I split 2025 at September, though, the rate was 1.5 a month before the Academy and 1.5 a month after it. The drop started in January, the same month the new press guarding went in. I'm not saying the Academy has no effect on safety. I'm saying these numbers can't show one yet. Do you have the 2026 incident data? That would tell us more than 2025 can."

Argument B: "My supervisors aren't quitting anymore"

Period Left Rate
Sep 2024 to Aug 2025 (before) 4 of 22 18.2%
Sep 2025 to Aug 2026 (after) 4 of 24 16.7%
  • The same number of supervisors left. The rate is slightly lower only because the group grew from 22 to 24.
  • With numbers this small, one person either way changes the picture.
  • The data doesn't say whether the 4 leavers in the second year attended the Academy, or why they left.
  • Verdict: the data doesn't support this claim.

How to answer honestly: "I looked for that and couldn't find it. Four supervisors left in the year before the Academy and four in the year after. Do you know something about who left, or why, that isn't in the HR numbers? If you do, I want to hear it."

Argument C: "You promised it through 2027"

  • Verdict: supported. Your November 2025 email says: "We'll fund it at Akron through December 2027."
  • How to answer: don't argue it. See section 3.

Arguments he might make that aren't in the inputs

  • He may say his supervisors like the Academy, or that it builds a shared identity.
  • Nothing in your inputs confirms or contradicts this. Ask what he's seeing, and don't dismiss it.

3. Where he has a point, and what to fix in your own case B: one page

1. Your written commitment. You promised funding through December 2027. You have two honest options.

  • Honour it. Keep the Academy through 2027 and use the time to gather proper evidence. If you do this, the $125,000 saving can't start until 2028.
  • Reopen it openly. Say plainly that you're asking to change a commitment you made, explain why, and accept that he can say no. Don't present the proposal as if the promise doesn't exist.

2. The $125,000 saving is wrong for year one. found

  • Year one: $180,000 minus $95,000 = $85,000 saved.
  • Later years: if the ongoing cost is $55,000 ($95,000 minus the $40,000 one-time certification), the saving is $125,000 a year. That only holds from year two onward.
  • The inputs don't say whether $55,000 a year covers the time of two internal coaches. Treat the ongoing cost as unverified until you've confirmed it.

3. The survey is weaker than "supervisors want coaching." found

  • 13 of 24 responded (54%).
  • 8 of those 13 asked for more coaching on the floor. That is 62% of respondents, but only 8 of 24 supervisors (33%).
  • "More coaching on the floor" isn't the same as "replace the Academy." They may want both.
  • The survey isn't split by shift, so you can't tell whether night-shift supervisors asked for this.

4. Your proposal doesn't fit a third of the plant. found

  • The coaches work on day shift, but 8 of the 24 supervisors (one third) work nights.
  • As written, those 8 would lose the Academy and get nothing in its place.
  • This is the clearest place the proposal doesn't fit the problem. Fix it before Thursday or say so openly in the meeting.

5. Neither program has evidence that it works.

  • The data doesn't show the Academy improved safety or retention.
  • Nothing in your inputs shows the coaching would either.
  • If you use Tom's weak evidence against him, use the same standard on your own proposal. Cost and an 8-person survey request aren't proof the coaching will work.

6. The problem isn't defined.

  • Your inputs don't say what either program is meant to fix.
  • Safety? The rate is flat since the guarding went in. Retention? It's flat at 4 leavers a year. Supervisor skill? You have no measure of it.
  • Until you agree on the problem, you're debating the format of a solution.

A possible middle path (a suggestion for you to weigh, not supported by data): C: decide

  • Keep the Academy through 2027, as promised.
  • Agree two or three measures with Tom now: monthly incidents for 2026, supervisor departures and reasons, and something on supervisor capability.
  • Possibly pilot floor coaching on day shift alongside the Academy, with a plan for night shift.
  • Revisit in 2027 with real data from both.
  • This keeps your word and shifts the conversation from opinion to evidence. It also gives up the savings until 2028.

4. Opening (under a minute)

"Tom, thanks for making the time. Let me start with the obvious: I told you last year we'd fund the Academy through 2027, and I put it in writing. I'm not going to pretend I didn't. If I ask to change that, I'll ask you straight, and you can say no.

Second, Akron's safety numbers really are better. Incidents went from 30 to 18. I've been through the data, and I want to show you what I found, because it's more complicated than either of us has been saying. That includes some weak spots in my own proposal, like day-shift-only coaching that leaves your night supervisors out.

What I'd like from today is for us to agree what problem we're trying to solve for your supervisors, and how we'd both know if something's working. Then we can talk about the best way to get there."


Questions to bring (data not in the inputs)

  • Akron's 2026 incident figures by month.
  • Whether the supervisors who left in Sep 2025 to Aug 2026 had attended the Academy, and their exit reasons.
  • Whether the 22 and 24 are start-of-period headcounts or averages.
  • What the $55,000 ongoing coaching cost covers, and what night-shift coverage would cost.
  • How survey responses split between day and night shift.
foundcaught a planted problemA–CLena changes it; see the next section
The human part

What Lena changed, and why

Claude found every planted problem and held Lena's proposal to the same standard as Tom's claims. Lena's edits fix one slip and turn a long briefing into something she can use in a 30-minute meeting.

What Claude found: 6 of 6 planted problems
One run
✓40% drop came before the Academy
✓Turnover didn't change
✓Lena's own written promise
✓Coaching leaves out nights
✓Year-one saving overstated
✓Survey is 8 of 24 people
Three edits, one catch and two judgment calls
A
CLAUDE WROTE
“6 incidents in 8 months... no: 6 in 4 months = 1.5 a month”
LENA'S VERSION
“September to December 2025: 6 incidents in 4 months, 1.5 a month.”
WHY · ACCURACY
The AI corrected itself mid-line. The number is right, but Lena rechecked it before quoting it to someone who'll be looking for a mistake.
B
CLAUDE WROTE
A briefing of several pages, with six points on where he has a point.
LENA'S VERSION
One page: two things she'll concede, two she'll hold, and one ask.
WHY · LENGTH
In a 30-minute one-to-one she needs to know her position cold, not read from notes.
C
CLAUDE WROTE
“A possible middle path … Possibly pilot floor coaching on day shift alongside the Academy, with a plan for night shift.”
LENA'S VERSION
Keep the promise through 2027. Any coaching pilot covers both shifts. Agree what the Academy is meant to fix, and the measures.
WHY · DECISION
The AI offered options. Lena decided before the meeting what she'd commit to, so Tom hears a position, not a menu.
What only you can do
Decide to keep your word, admit the gap in your own plan, and walk in with a clear ask. The AI checked both sides. You protect the relationship.
Your turn

Make it yours

Use it before any conversation where someone will push back: a budget cut, a changed decision, a programme you want to end.

Your prompt: fill in the brackets

I'm the [your role] at a [industry] company. I want to [your proposal]. [Who] wants [their position], and we meet [format and length] on [date]. Today is [date]. Below are [their email, the data, my proposal and anything I've told them before].

Please help me prepare: 1. What each of us needs from this meeting. 2. Their likely arguments: which ones the data supports, which it doesn't, and how I answer each one honestly. 3. Where they have a point: anything in my own proposal or past commitments I should fix or concede. 4. An opening I can say in under a minute.

Be fair to them. Check their claims and mine with the same rigour, and tell me if my proposal doesn't fit the problem. Don't add facts that aren't in the inputs.

Five checks before you walk in
1
You've checked the dates behind every "since we started" claim.
2
You've found anything you promised them before.
3
Your own numbers hold up as well as you expect theirs to.
4
You know what you'll concede and what you'll hold.
5
Your ask is one sentence they can say yes to.
Make it repeatable (optional)
Save the prompt once and reuse it before any hard conversation: a custom GPT in ChatGPT, a project in Claude, a Gem in Gemini, or an agent in Copilot. Some companies switch these off; copy and paste always works.
With real data
Emails about colleagues and plant data are confidential. Use only the AI tool your company approves, and follow its rules.
What this example shows
One run of this prompt in Claude found all six problems planted in fictional inputs, held both sides to the same standard, and invented nothing.
What it doesn't show
That every tool, setting or run will do the same, how much time it saves you, or how the conversation will go. Selvane Machinery and its people are fictional.
Appendix

Copy-ready: the prompt and the five inputs

Copy the prompt, then all five inputs, into one message in a new chat. Everything here is fictional and safe to paste.

The prompt

I'm Head of L&D for North America at an industrial equipment maker. I want to replace a five-day Supervisor Academy at one plant with on-the-floor coaching. The plant's general manager wants to keep it, and we meet one to one for 30 minutes on Thursday, November 19. Today is November 16. Below are his email, the plant's data, my draft proposal, a supervisor survey and an email I sent him last year.

Please help me prepare:

  1. What each of us needs from this meeting.
  2. His likely arguments: which ones the data supports, which it doesn't, and how I answer each one honestly.
  3. Where he has a point: anything in my own proposal or past commitments I should fix or concede.
  4. An opening I can say in under a minute.

Be fair to him. Check his claims and mine with the same rigour, and tell me if my proposal doesn't fit the problem. Don't add facts that aren't in the inputs.

[1] Email from Tom Kessler, General Manager, Akron

Subject: Supervisor Academy

Lena, I hear you want to scrap the Supervisor Academy. It's the best thing we've done here. Incidents are down 40% since we started it, and my supervisors aren't quitting anymore. And you promised it through 2027. Let's talk Thursday.

[2] Akron data from HR and EHS

Supervisor Academy: started September 2025. Five days off-site. $180,000 a year.

Recordable incidents: 2024: 30. 2025: 18 (January to August: 12; September to December: 6). EHS note: new machine guarding installed on all presses in January 2025.

Supervisors who left: September 2024 to August 2025: 4 of 22. September 2025 to August 2026: 4 of 24.

[3] My draft proposal

Replace the Academy with on-the-floor coaching: two internal coaches work alongside supervisors on day shift. First-year cost $95,000, including $40,000 one-time coach certification. Saves $125,000 a year.

[4] Akron supervisor pulse survey, June 2026

13 of 24 supervisors responded. 8 said they want "more coaching on the floor." Shift pattern: 16 day-shift supervisors, 8 night-shift supervisors.

[5] My email to Tom, November 2025

Glad the Academy is working for you. We'll fund it at Akron through December 2027.