← Back to the exerciseThe full worked example: Test a proposed fix against where the problem really is
campo.
Field guide / L&D / UC-001 · SC-09
The result

"Cut onboarding to 1 day?" Cut the 5-day wait instead.

A VP of Talent at a 60-warehouse logistics company tests an Operations proposal against its own numbers and brings back a counter-proposal, with any AI chat her company allows.

The resultWhat Renee sends the CHRO and Operations. Fictional company.
BRANNOCK LOGISTICS · PEAK SEASON
Getting new hires on the floor faster
For: Janet Cole, CHRO, and Dan Reyes, SVP Ops
Thursday · From: Renee Alvarez

THE SHORT ANSWER

Onboarding isn't the bottleneck. Of the 9 days from offer to first shift, the background check takes 5 and onboarding 3. And day 2 of onboarding is the safety day, when 14 of October's 22 injuries involved people in their first 30 days. · Inputs 3, 5

WHAT I PROPOSE FOR PEAK

THIS WEEK
Start onboarding every day during peak
Aims to remove the 1-day wait after the check clears. My team runs the extra start days. · Input 3
HR OPS
Find out what drives the 5-day background check
It's the biggest single block of time. HR operations and the screening provider bring options by next week. · Input 3
WITH OPS
Keep the safety day. Trim days 1 and 3 together
Anything a new hire doesn't need before the first shift moves to the first weeks on the job. · Inputs 3, 5

WHAT THE NUMBERS SAYInputs 2, 4

East 105%Not a rate
Rehires counted twice
October 27%Mixed
Includes temp contract ends
Onboarding isn't workingUnsupported
No link in the data
Site 51Missing
No data since June
Caught on review
1
"Onboarding is the bottleneck" is 3 days of 9; the check is 5
2
October's 27% includes temps whose contracts ended
3
Cutting to 1 day risks the safety day
UC-001 · SC-09 · Rev A
FOR
VP, Talent
SECTOR
Logistics & distribution
COMPANY
18,000 people · 60 warehouses · US, Canada
TASK
Respond to a proposal to cut onboarding
WRITTEN FOR
ChatGPT · Claude · Gemini · Copilot
AI TEST
Claude, Nov 13, 2026 · 6 of 6 found
Step 0 · The mess

What landed on Renee's desk

A proposal from Operations, a CHRO who half-believes it, and four reports. The proposal sounds sensible. Operations' own breakdown says the problem is somewhere else.

EMAIL · CHRO[1]
Ops onboarding proposal
“Dan wants to cut onboarding from 3 days to 1 for peak. He says onboarding is the bottleneck.”
1
A fix, before anyone checked the problem.
ONBOARDING REPORT[2]
October by region
East: 1,240 hires, 1,302 completions, 105%. Site 51 hasn't reported since June.
2
More completions than hires.
NOTE · SVP OPERATIONS[3]
Time to first shift
“Background check cleared, 5 days; check cleared to onboarding start, 1 day; onboarding, 3 days.”
3
His own numbers point elsewhere.
HR ANALYTICS[4]
30-day turnover
18%, 21%, 27%. “Onboarding isn't working.” October includes temps on 6-week contracts.
4
A season called a trend.
SAFETY NOTE[5]
October injuries
14 of 22 recordables involved people in their first 30 days. Day 2 is the safety day.
5
What the cut would remove.
You're not alone
35%
of workers' compensation claims come from employees in their first year on the job.
Travelers 2024 Injury Impact Report (more than 1.2 million claims, 2017 to 2021). Source
Steps 1–3 · How it works

Three steps, one chat

Here the AI tests someone else's proposal. The prompt asks where the time actually goes.

1
Open a new chat

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

2
Paste the prompt and the reports

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

3
Bring a better option

Add up the days yourself. Then turn the AI's ideas into a short proposal your colleague can say yes to.

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

I'm VP of Talent at a logistics company with about 18,000 people in 60 warehouses in the US and Canada. Peak season is starting. Our SVP of Operations wants to cut new-hire onboarding from three days to one, to get people on the floor faster. My CHRO wants my view by Thursday. Below are her email, our onboarding report, the SVP's note, our 30-day turnover figures and a safety note.

1
Who you are, what's being proposed, and who needs your view by when.

Please give me:

  1. Does cutting onboarding solve the problem Operations describes? Answer in two sentences.
  2. What the numbers really show, with the input each point comes from.
  3. A counter-proposal I can take to Operations.
2
A two-sentence answer, the numbers, and a counter-proposal. Saying no is easy; a better option is what moves things.

Be skeptical. Check what each number counts, whether a change is a trend or a season, where the time actually goes, and what we would lose by cutting. If something is missing, say so. Don't add facts that aren't in the inputs.

3
The heavy lifting: what each number counts, trend or season, where the time goes, and what we'd lose. It found all six problems.
Where to paste
Every tool has one message box. Paste the prompt, add a blank line, paste the reports, send. Exports can be attached as files if your tool allows it.
Safe to try
Practise with the fictional reports first. Real injury and turnover data is sensitive: 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 Renee did: check each point against the source it cites.

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

1. Does cutting onboarding solve the problem Operations describes? A: plainer

Only partly. By Dan's own breakdown, onboarding is 3 of the 9 days from offer to first shift, and the background check alone takes 5. Cutting onboarding to 1 day would save at most 2 days, and it would put the warehouse safety day (Day 2) at risk at the moment injuries are concentrated among new starters. The turnover figures don't show that onboarding is failing, so "fixing" onboarding by shortening it doesn't address a problem the data has proven.

2. What the numbers really show

Time to first shift: the biggest delay is the background check, not onboarding [3]

  • 5 + 1 + 3 = 9 days, so the breakdown adds up.
  • Background check: 5 of 9 days (about 56%). The wait from check cleared to onboarding start is 1 day (about 11%). Onboarding is 3 days (about 33%). found
  • 6 of the 9 days pass before onboarding even starts. The claim that "onboarding is the bottleneck" [1] doesn't match Dan's own numbers.
  • Missing: whether these are calendar days or working days. "9 days on average" is an average, and we don't know the spread or how it varies by site. We also don't know whether the 1-day wait is driven by when onboarding sessions are scheduled.

The onboarding report's "%" is not a completion rate [2]

  • East shows 1,302 completions for 1,240 hires (105%). Nobody can be more than 100% onboarded. The note explains why: completions are counted each time someone finishes, and people who leave and are rehired in the same month are counted again. found
  • So completions count events, while hires count people. They are not the same group, and dividing one by the other doesn't tell us what share of new hires completed onboarding. The 92% in Central and West is unreliable for the same reason.
  • East's 62 "extra" completions (1,302 − 1,240) suggest same-month leave-and-rehire activity, but the report doesn't say how many rehires there were, so this is an inference.
  • Missing: Site 51 (Moncton) hasn't submitted onboarding data since June. That is four months missing (July to October), and it isn't flagged in the table. We don't know which region Moncton sits in. found
  • Totals for the three regions: 3,080 hires and 2,995 completions, with the same counting caveat.

30-day turnover: the October jump is mixed with a season and a definition change [4]

  • The figures are 18% in August, 21% in September and 27% in October. The footnote says October includes peak-season temporary hires on 6-week contracts, counted as leavers when the contract ends. So October isn't comparable with August and September: part of the rise comes from planned contract ends, not people choosing to quit. found
  • There's also a definitional puzzle. A 6-week contract runs about 42 days, so a contract ending normally shouldn't fall within a 30-day window. Either the metric isn't measuring the same thing each month, or some temps leave before their contract ends. We need to ask the analyst.
  • Three months, running into peak season, is not a trend. There's no prior-year figure for the same months to show whether this happens every peak.
  • Missing: the denominator (30-day leavers as a share of what?), the October figure with temps removed, and any exit reasons.
  • The analyst's line "Onboarding isn't working" is an opinion. Nothing in the inputs links turnover to onboarding content. Janet's email [1] repeats this as fact. found
  • Even if onboarding were contributing to turnover, cutting it from 3 days to 1 is not a fix for that.

Safety: new starters are involved in most injuries, and Day 2 is the safety day [5]

  • 14 of 22 recordable injuries in October (about 64%) involved people in their first 30 days. found
  • Day 2 of onboarding covers powered equipment, lifting and pedestrian routes. A 1-day programme would either drop this or compress it. Dan's note [3] doesn't say which.
  • Caution: we don't know what share of the workforce was in their first 30 days in October, so we can't yet say new starters are injured at a higher rate. With peak hiring, they may simply be a large share of the people on the floor.
  • Missing: whether the 14 had completed onboarding (including Day 2), whether they were temps, and whether Moncton's injuries are included.

What we'd lose by cutting

  • The safety day, at the point when new starters are involved in nearly two thirds of injuries.
  • The gain is at most 2 of 9 days. Cutting both the 5-day check and the 1-day wait would be outside onboarding's control, but they hold 6 days.

3. Counter-proposal for Operations B: three steps

Proposal: keep the safety content intact and attack the 6 days before onboarding starts.

  1. Target the background check first (5 days). Ask HR operations and the screening provider what drives the 5 days and how much can be removed for peak. This is the largest single block, and it's bigger than the entire onboarding programme.
  2. Close the 1-day wait between check cleared and onboarding start. If onboarding only starts on set days, run more frequent start cohorts during peak so cleared hires start the next shift.
  3. Protect Day 2 (safety) in full. Do not cut powered equipment, lifting or pedestrian routes training during peak.
  4. Review Days 1 and 3 with Operations, line by line. The inputs don't say what they contain. Any content that isn't needed before someone's first shift could move to before start, or to short sessions in the first weeks on the job. That could recover up to a day without touching safety.
  5. Pilot before rollout. Try the changes at a few sites for peak and track time from offer to first shift, injuries in the first 30 days, and 30-day turnover (excluding contract ends).
  6. Fix the numbers before anyone uses them again: C: own team
    • Report onboarding as unique people who completed it divided by unique hires.
    • Report 30-day turnover with temporary contract ends separated out, and with a stated denominator.
    • Add a prior-year comparison for the same months.
    • Get Moncton's data since June.
    • Report injuries in the first 30 days as a rate (injuries per new starter, or per hours worked) and note whether each person had completed Day 2.

Short version for Janet: Onboarding is a third of the delay, not the bottleneck. The October turnover jump is partly temporary contracts ending, and nothing in the data links turnover to onboarding. The day we'd most likely lose is the safety day, at a time when new starters account for 14 of 22 injuries. Go after the 6 days before onboarding first, and trim Days 1 and 3 only where content isn't needed before the first shift.

foundcaught a planted problemA–CRenee changes it; see the next section
The human part

What Renee changed, and why

Claude found every planted problem and even spotted that the turnover footnote doesn't add up. Renee's edits are about turning a good analysis into a proposal Operations can accept.

What Claude found: 6 of 6 planted problems
One run
✓Completions above hires
✓The check is the bottleneck
✓October includes temp contracts
✓“Onboarding isn't working” unsupported
✓One site missing since June
✓Cut would hit the safety day
Three edits, all judgment calls
A
CLAUDE WROTE
“Only partly. By Dan's own breakdown, onboarding is 3 of the 9 days…”
RENEE'S VERSION
“Onboarding isn't the bottleneck. Of the 9 days, the background check takes 5 and onboarding 3.”
WHY · AUDIENCE
A plain answer first. "Only partly" invites a negotiation about the wrong fix.
B
CLAUDE WROTE
Six counter-proposal steps, including a pilot and five reporting fixes.
RENEE'S VERSION
Three steps: daily start dates (Renee's team, this week), the background check (HR operations), and trimming days 1 and 3 together.
WHY · LENGTH
Operations needs something to say yes to before peak. Renee leads with the step she can deliver herself.
C
CLAUDE WROTE
Step 6: “Fix the numbers before anyone uses them again” as part of the proposal to Operations.
RENEE'S VERSION
Handled inside Renee's own team, and left out of the proposal.
WHY · PRIORITY
Fixing reports is her job. Putting it in front of Operations would read as a lecture.
What only you can do
Turn a no into a better yes, lead with what your own team can deliver, and keep the safety day off the table. The AI found where the time goes. You make the deal.
Your turn

Make it yours

Use it whenever someone proposes a fix you're not sure about: cutting a programme, adding a course, changing a process.

Your prompt: fill in the brackets

I'm the [your role] at a [industry] company with about [number] people in [number] sites. [Who] wants to [the proposed fix] to [their goal]. [Your leader] wants my view by [date]. Below are [the proposal and the reports that bear on it].

Please give me: 1. Does [the proposed fix] solve the problem [they] describe? Answer in two sentences. 2. What the numbers really show, with the input each point comes from. 3. A counter-proposal I can take to [them].

Be skeptical. Check what each number counts, whether a change is a trend or a season, where the time actually goes, and what we would lose by cutting. If something is missing, say so. Don't add facts that aren't in the inputs.

Five checks before you send
1
You've added up where the time or money actually goes.
2
Rates and counts measure the same people.
3
A jump is compared with the same season, not just last month.
4
You've named what the proposed fix would remove.
5
Your counter-proposal starts with something you can deliver yourself.
Make it repeatable (optional)
Save the prompt once and reuse it for any proposal you need to test: 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
Injury and turnover data is sensitive. Use only the AI tool your company approves, and follow its rules on employee data.
What this example shows
One run of this prompt in Claude found all six problems planted in fictional reports, spotted an inconsistency we hadn't planted, and invented nothing.
What it doesn't show
That every tool, setting or run will do the same, or how much time it saves you. Brannock Logistics 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 VP of Talent at a logistics company with about 18,000 people in 60 warehouses in the US and Canada. Peak season is starting. Our SVP of Operations wants to cut new-hire onboarding from three days to one, to get people on the floor faster. My CHRO wants my view by Thursday. Below are her email, our onboarding report, the SVP's note, our 30-day turnover figures and a safety note.

Please give me:

  1. Does cutting onboarding solve the problem Operations describes? Answer in two sentences.
  2. What the numbers really show, with the input each point comes from.
  3. A counter-proposal I can take to Operations.

Be skeptical. Check what each number counts, whether a change is a trend or a season, where the time actually goes, and what we would lose by cutting. If something is missing, say so. Don't add facts that aren't in the inputs.

[1] Email from Janet Cole, CHRO

Subject: Ops onboarding proposal

Renee, Dan wants to cut onboarding from 3 days to 1 for peak. He says onboarding is the bottleneck and our onboarding clearly isn't working, given turnover. Your view by Thursday please.

[2] Onboarding report, October
Region Hires Onboarding completions %
East 1,240 1,302 105%
Central 980 902 92%
West 860 791 92%

Note: a completion is counted each time someone finishes the programme. People who leave and are rehired in the same month are counted again. Site 51 (Moncton) has not submitted onboarding data since June.

[3] Note from Dan Reyes, SVP Operations

Offer to first shift is 9 days on average. Onboarding is 3 of those. Cut it to 1 and we get people on the floor 2 days faster for peak. Breakdown: offer to background check cleared, 5 days; check cleared to onboarding start, 1 day; onboarding, 3 days.

[4] HR analytics: people who leave within 30 days

August 18% · September 21% · October 27% Analyst note: "30-day turnover rising sharply. Onboarding isn't working." Footnote: October includes peak-season temporary hires on 6-week contracts, counted as leavers when the contract ends.

[5] Safety note, October

22 recordable injuries in the network. 14 involved people in their first 30 days. Day 2 of onboarding is the warehouse safety day (powered equipment, lifting, pedestrian routes).