Choose one of three approaches. The steps for that option appear below.
Choose an option above. On a phone, swipe the tabs left or right to see all options →
ChatGPT runs a code-backed break-even and sensitivity test. You inspect the code and recalculate samples.
Get inputs, rules ready
Draft in Python
Check and save pack
Use: You
Have ready: assumptions table (currency, fixed costs per period, price, variable cost per unit, approved test ranges, expected volume); check a single-product linear model fits.
Stop if upload permission is absent. Verify date, currency, completeness and the supplied rules; do not replace missing rules with guesses. Keep inputs in an access-controlled local draft.
Use: ChatGPT
Only the approved input package from step 1: paste the assumptions table and rules. ChatGPT shows Python code, row counts, assumptions and exception rows.
What to paste into the tool
Task: Produce a draft base-case and sensitivity table with assumptions, break-even units and an explicit limitations note. Inputs: [PASTE: selling price per unit] [PASTE: variable cost per unit] [PASTE: fixed costs per stated period] [PASTE: test ranges] [PASTE: expected volume and its period] [PASTE: currency shared by all monetary inputs] [PASTE: confirmation that a single-product model with constant unit price and variable cost and fixed costs over the stated period is appropriate]. Use only these inputs. Show the Python used and each calculation step per test case. Use contribution per unit = price minus variable cost, and break-even units = fixed costs divided by positive contribution per unit. Round required whole units upward. If fixed costs are zero, report 0 units as the minimum sales needed to cover costs at zero output and flag zero or negative contribution separately. If fixed costs are positive and contribution is zero or negative, show no finite break-even under this model. Keep periods and units aligned. Do not assume demand responds to price or recommend a price. Never silently drop or repair rows. Report missing facts as 'not stated'. Do not access accounts or external sources, send messages, post entries, change the master ledger, sign, publish or pay. Produce a draft for human review only.
Use: You
Check the output, correct or reject it. Test zero fixed costs. Save the checked pack locally as draft.
Recalculate base and edge cases: zero/negative contribution, single and combined changes; whole-unit rounding; expected-volume comparisons use the same period and currency. No recommendation to launch, borrow, invest or set prices is made. Do not send, sign, pay, publish, post journals or update master records.
Copilot in Excel suggests formulas from synthetic sample columns only. You apply checked formulas locally to the real draft.
Get inputs, rules ready
Suggest formulas
Test, build, check
Use: You
Have ready: assumptions table (currency, fixed costs per period, price, variable cost per unit, approved test ranges, expected volume); confirm a single-product linear model fits.
Stop if upload permission is absent.
Use: Copilot in Excel
Only the approved input package from step 1. Ask in chat mode for formulas; it must not modify the workbook or make a finance decision.
What to paste into the tool
Task: Suggest spreadsheet formulas and test cases for a private draft worksheet, not a finance decision. Inputs: [PASTE: 3-5 synthetic example assumption sets, not real figures] [PASTE: column names] [PASTE: approved rules]. Use only these inputs. Use contribution per unit = price minus variable cost, and break-even units = fixed costs divided by positive contribution per unit. Round required whole units upward. If fixed costs are zero, report 0 units as the minimum sales needed to cover costs at zero output and flag zero or negative contribution separately. If fixed costs are positive and contribution is zero or negative, show no finite break-even under this model. Keep periods and units aligned. Do not assume demand responds to price or recommend a price. Explain references, rounding and error conditions (zero or negative contribution). Give formulas as text, with sample expected results; do not edit files. Missing inputs must produce a 'not stated' note or an explicit exception. Do not choose accounting policy or recommend action. Do not send, post, sign, publish or pay.
Use: You
Test formulas on synthetic normal and edge cases, including zero fixed costs; build the draft yourself. Save the checked pack locally as draft.
Do not send, sign, pay, publish, post journals or update master records.
You do all matching and arithmetic first. Claude only summarises a small, checked text note.
Calculate by hand, write note
Summarise the note
Check and save pack
Use: You
Have ready: assumptions table (currency, fixed costs per period, price, variable cost per unit, approved test ranges, expected volume); confirm a single-product linear model fits.
Stop if upload permission is absent. Calculate the base case and approved sensitivity cases by hand; check model limits; record cases with no finite break-even.
Use: Claude
Only the approved input package from step 1: the checked note, not the data file. Claude summarises it and lists open questions, no recalculating.
What to paste into the tool
Task: Write a plain-language explanation of the human-checked sensitivity cases and model limitations, without choosing a selling price or business decision. Input: [PASTE: the human-checked note from step C-1]. Use only this note. Preserve every amount, unit, date, label and source ID exactly. Do not recalculate, infer missing facts or add a judgement. Use contribution per unit = price minus variable cost, and break-even units = fixed costs divided by positive contribution per unit. Round required whole units upward. If fixed costs are zero, report 0 units as the minimum sales needed to cover costs at zero output and flag zero or negative contribution separately. If fixed costs are positive and contribution is zero or negative, show no finite break-even under this model. Keep periods and units aligned. Do not assume demand responds to price or recommend a price. Say 'not stated' for missing information. Separate facts from open questions. Do not draft an external message or instruct anyone to act. Do not access accounts, send, post, sign, publish or pay. Return internal draft text only.
Use: You
Check every sentence and number against your note; remove unsupported claims; keep not-stated facts. Test zero fixed costs. Save the pack locally as draft.
Do not send, sign, pay, publish, post journals or update master records.
Best when: A small model needs repeatable batch calculations you can inspect.
What it does: ChatGPT calculates in code; you recalculate samples.
Strong at: Shows the Python code and row-level audit trail.
Limit: Free plan limits file uploads and data analysis.
Free plan: Yes, limited (file uploads, data analysis)
Best when: You want to build and own the formulas.
What it does: Copilot in Excel suggests formulas; you build and test them.
Strong at: Explains formulas and gives synthetic test cases.
Limit: On iPhone and Android, chat mode only.
Free plan: Paid plan needed
Best when: You want to do all the arithmetic yourself.
What it does: You calculate by hand; Claude summarises the checked note.
Strong at: Preserves every amount, date and source ID exactly.
Limit: Usage is capped in 5-hour and weekly windows.
Free plan: Yes, limited (usage capped in 5-hour and weekly windows)
Before you paste into any AI tool, remove any confidential information. Put it back at the end.
Deciding between ChatGPT and Claude for this? See the comparison.
Never paste credentials or full payment details.
Confirm authorised inputs and rules before use.
Recalculate the base case and every edge case, including zero fixed costs and zero/negative contribution. Confirm whole-unit rounding and that expected-volume comparisons use the same period and currency.
Reconcile completeness and totals, preserve open questions, and save only a draft.
Check the result against your original inputs. Approve it before sharing or acting.
Based on official product documentation checked on 2026-10-29.