How to Build a Simple Healthcare Admin Pro Forma From a Spreadsheet of Assumptions With AI

The problem and who this is for

This workflow is for healthcare operations leaders, finance partners, department managers, strategy staff, and analysts who need a first-pass pro forma from non-sensitive operating assumptions.

The job is narrow: take a spreadsheet containing assumptions such as service volume, staffing, labor cost, supplies, fixed costs, and other known drivers, then turn those assumptions into a simple model that leadership can inspect before formal finance review.

The best current ChatGPT path is no longer automatically "upload the workbook into a chat." OpenAI now offers ChatGPT directly inside Excel and Google Sheets, where it can inspect formulas, values, assumptions, and multiple tabs without separating the analysis from the workbook. For organizations that cannot use the spreadsheet add-in, standard ChatGPT data analysis with an uploaded XLSX or CSV remains a strong fallback.

Prerequisites

• A non-sensitive assumptions workbook in Excel or Google Sheets.
• Clear labels for assumptions, formulas, and outputs.
• A duplicate working copy of the workbook so the original stays untouched.
• ChatGPT for Excel or ChatGPT for Google Sheets if your account and organization allow it.
• If the spreadsheet add-in is not available, standard ChatGPT file upload and data analysis access.
• A finance or operational owner who can verify the model before leadership uses it.
• Organization approval for the specific AI surface you plan to use.

Do not upload patient-level data, claims detail, employee-sensitive data, or other restricted material unless your organization has approved that exact workflow.

How to capture or gather the source material

Start with one workbook that makes the assumptions visible.

A simple structure is enough:

1. Inputs: volume, price or reimbursement assumptions, staffing levels, wage assumptions, supply costs, fixed costs, ramp timing, and other drivers.
2. Calculations: formulas that transform the inputs into monthly or annual results.
3. Outputs: revenue, expense, contribution margin, operating result, or whatever summary your finance team actually uses.
4. Notes: source, owner, date, or confidence level for assumptions that are still provisional.

If the current sheet is messy, clean it before asking AI to model from it. Use descriptive headers, avoid unrelated tables on the same tab, and keep one clear unit for each assumption.

Step-by-step workflow

1. Duplicate the workbook

Make a working copy before using an AI tool that can edit the sheet.

Keep the original as the comparison point. This matters because spreadsheet-native ChatGPT can make changes directly in Excel or Google Sheets.

2. Mark what the AI is allowed to change

Tell ChatGPT which tabs and ranges are inputs, which formulas must stay untouched, and where it may add new calculations.

A useful rule is:

• Inputs may be read but not overwritten unless you approve.
• Existing formulas should be preserved.
• New formulas should go in clearly labeled cells or a new scenario tab.
• Formatting changes should be minimal unless you ask for them.

3. Ask for an assumption map before any edits

Open ChatGPT in Excel or Google Sheets and ask it to identify:

• The major assumptions.
• Which cells drive the result.
• Which formulas connect the inputs to the outputs.
• Missing assumptions.
• Ambiguous units.
• Any formulas that appear inconsistent.

Do not let the first action be a rewrite of the workbook. First make sure the model understands the sheet.

4. Ask ChatGPT to propose the pro forma structure

Give it the exact output leadership needs.

For example:

• Monthly volume.
• Monthly revenue.
• Labor expense.
• Non-labor operating expense.
• Fixed expense.
• Contribution margin.
• Net operating impact.
• Base, upside, and downside cases.

Have ChatGPT show the proposed calculation logic before it edits the workbook.

5. Build the model in a new output or scenario tab

Once the logic is approved, ask ChatGPT to create the model in a new tab or clearly separated range.

For each important output, require:

• The formula.
• The source assumptions.
• The unit.
• The period.
• A note when the result depends on an unverified assumption.

If you are using standard ChatGPT instead of the spreadsheet add-in, upload the workbook and ask it to return the calculation table and formulas for you to place back into Excel or Google Sheets.

6. Run simple sensitivity cases

Ask for at least three scenarios when the assumptions are uncertain:

• Base case.
• Upside case.
• Downside case.

Change only the assumptions you intentionally selected. Do not let the model create arbitrary optimistic or pessimistic numbers.

7. Identify the assumptions that matter most

Ask ChatGPT to rank the assumptions by impact on the final result.

The point is not to create a mathematically elaborate model. The point is to show leadership which assumptions deserve the most scrutiny before the organization spends more time refining the business case.

8. Perform an independent finance check

Before circulation, verify:

• Every important formula.
• Every source assumption.
• The time period.
• Units.
• Sign conventions.
• Staffing math.
• Revenue logic.
• Fixed versus variable cost treatment.
• Scenario assumptions.

ChatGPT can accelerate the model. Finance still owns the model.

Tool-specific instructions

Primary path: ChatGPT for Excel or Google Sheets

Use the spreadsheet-native ChatGPT experience when it is available and approved. OpenAI currently documents support for building, updating, explaining, and summarizing spreadsheets directly in Excel and Google Sheets, including multi-tab workbooks with formulas, references, and assumptions.

For large changes, ask for a plan before edits. Specify what must stay unchanged and review formulas and changed cells before saving or sharing.

Fallback: standard ChatGPT data analysis

Upload an XLSX, XLS, or CSV file to ChatGPT. OpenAI currently documents spreadsheet analysis, Python-backed calculations, tables, charts, and review of assumptions.

This path is useful when you want analysis without giving the tool direct control of the workbook.

Copy and paste prompt block

{
  "role": "You are an internal healthcare operations modeling assistant working only with the non-sensitive workbook I provide.",
  "task": "Build a simple first-pass administrative pro forma from the existing assumptions while preserving the source workbook for human finance review.",
  "rules": [
    "Do not overwrite source assumptions unless I explicitly approve.",
    "Do not invent volume, reimbursement, staffing, wage, cost, or timing assumptions.",
    "Before editing, identify the key assumptions, formulas, units, and missing information.",
    "Show the proposed calculation logic before making changes.",
    "Place new calculations in a clearly labeled output or scenario area.",
    "Flag every output that depends on an unverified assumption.",
    "Do not treat the model as final financial advice or an approved budget."
  ],
  "requested_outputs": [
    "assumption map",
    "base-case pro forma",
    "upside case",
    "downside case",
    "top assumptions ranked by impact",
    "open questions for finance review"
  ],
  "review_checklist": [
    "formula accuracy",
    "units and time periods",
    "source assumption traceability",
    "staffing math",
    "revenue logic",
    "fixed versus variable cost treatment",
    "scenario assumptions"
  ]
}

Quality checks

• Every output traces back to an explicit source assumption or formula.
• The original workbook remains available for comparison.
• Existing formulas were not silently overwritten.
• Units and time periods are consistent.
• Base, upside, and downside cases differ only where you intended them to.
• High-impact assumptions are visible to leadership.
• A finance owner reviewed the formulas and assumptions before the model was circulated.

Common failure modes and fixes

ChatGPT changes cells you wanted preserved

Fix: work from a duplicate file, specify protected tabs and ranges, and ask for an edit plan before changes.

The model looks complete but contains invented assumptions

Fix: require a separate assumptions table and mark missing inputs as unresolved rather than filling them in.

The results are numerically correct but operationally unrealistic

Fix: add capacity, staffing, ramp, and timing constraints before rerunning the model.

The scenario analysis becomes arbitrary

Fix: explicitly name which assumptions may change and provide the values for each scenario.

The workbook is too messy for reliable analysis

Fix: clean headers, remove unrelated tables, normalize units, and separate inputs from outputs before using AI.