For CFOs and PE deal teams: 12 prompts that build a tied-out income statement, balance sheet, and cash flow from your own filings inside Claude Opus 5. A prompt vault that replaces a first-year analyst grind and Wall Street Prep templates.
Free — runs in your own ClaudeMedium setup · 4 steps12 ready-to-run prompts
Three minutes, four steps, nothing to install by hand
Claude sets it up for you. You just paste.
Never used Claude? It is free and takes 30 seconds to open. Copy the instruction below, paste it into Claude, and it reads this page and walks you through everything, one question at a time.
1
Tell Claude how to talk to you
One tap. It changes how much Claude explains, and how slowly it goes. You can change it any time.
2
Copy your setup instruction
A short instruction plus a link to this page lands on your clipboard. First copy asks for your email once. That unlocks every button across the whole library.
3
Open Claude in a new tab
Free account, no card, 30 seconds. This tab stays open so you can come back.
Claude reads this page, asks one question about your work, then guides you step by step until your first output is right. If anything looks wrong, tell Claude what you see, and it fixes it with you.
▸Prefer the full prompt instead of the link? (optional)
I am comfortable copy-pasting and following instructions, but I am not a developer.
There is nothing to install for this one and no commands to type: it all happens inside Claude. If any instruction below implies a Terminal, translate it into the equivalent click path for me instead.
- Plain English. Define jargon the first time it appears.
- One step at a time, then wait for me to confirm before the next one.
- Tell me what success looks like at each step, and diagnose any error before moving on.
Follow the instructions below with those rules applied.
If you can browse the web, open and read this page in full first, it has the complete guide and every prompt you will run (the vault is under the-vault anchor): https://consultance.ai/library/three-statement-model-vault#the-vault . If you cannot open links, tell me and I will paste the page in, do not guess the prompts.
---
I want to set up the 3-Statement Model prompt vault the right way and build a linked, tied-out model off my own filings, in my own Claude. Walk me through it step by step. Treat me like a CFO, finance lead, or PE deal-team member who models but has never run a build like this through Claude. Define every term once.
This is NOT a coding install. There is nothing to compile, no Terminal, no repo to clone. It is a private workspace in Claude, a pinned model, and 12 prompts you run in order.
One honest note up front so you can decide before you start: this vault is about 60% of the build. It gets you a linked, tied-out model you own by lunch. The last 40%, wiring it into your production stack with live data feeds, permissions, and an audit trail your auditor accepts, is the proper integration. That part is a call, not a prompt. It is the honest ceiling, not a barrier.
## Step 1, Stand up a private Project (one at a time, wait for me)
1. **Open Claude** at claude.ai. The free tier is enough to try it. Sign in.
2. **Create a Project.** A Project is a private workspace with its own knowledge that your other chats cannot see. This matters: your filings and management accounts stay in your tenant. Nothing goes to consultance.ai, nothing is stored by us, nothing is seen by us.
3. **Pin the model to Claude Opus 5.** Pin means set it as the model for this Project so every prompt runs on the same one. Opus 5 is the model these prompts were tuned on. If you only see other models, tell me which ones and I will tell you the best available fallback, I will not tell you it is "not possible."
## Step 2, Get your data in (your prompt-01 choice)
You will pick one of these in prompt 01. Decide now where your data lives:
- **A. Upload to the Project knowledge base** (drag your 10-K, 10-Q, T12, or trial balance in). Best for a clean file set.
- **B. Paste raw numbers in chat.** Fine for a quick build.
- **C. Claude app add-in** reading your Excel or Word directly.
- **D. A governed connector** to a warehouse or data room.
- **E. A mix.**
If you are not sure, start with A (upload to the Project). Never paste a live data room into a public, non-Project window.
## Step 3, Run the prompts in order
The vault is 12 prompts. Two are non-negotiable:
- **Prompt 01 is the router.** It stands up your board of advisors, asks the TYPE and DATA-SOURCE questions, and captures your tokens (entity, periods, currency, GAAP or IFRS). Answer its questions before it builds anything.
- **Prompt 07 is the reconciliation gate.** It re-derives the checks a second way and returns RECONCILED or BROKEN. The model is NOT board-ready until 07 returns RECONCILED. If it says BROKEN, it names the exact line, you fix upstream and rerun 07.
Run 01, 02, 03, 04, 05, 06, 07, then 08 through 12 as you need them.
## First session drill (do this once setup is done)
1. **Start with prompt 01.** Answer the DATA-SOURCE question, do not skip it, everything else routes off it.
2. **Then prompt 02 (historical ingest).** Good output looks like: three aligned tables (IS, BS, CF) with a normalization log, and any gaps marked NEEDS INPUT rather than guessed.
3. **Build 03 to 06.** Good output: a balance check row reading zero every period, and a cash-flow ending-cash line that ties to the balance-sheet cash line every period.
4. **Run prompt 07 before you trust a single number.** If it does not return RECONCILED, the model is not done. This is your review check.
5. **Only then run prompt 10** for the board or IC pack. It refuses to produce a pack unless 07 passed.
## Rules for walking me through this
- One step at a time. Wait for me to confirm each before the next.
- Define each term once: Project, pin, token, review_gate, reconciliation, tenant.
- Do NOT tell me a step is "not possible." If something is missing (a model, an add-in), tell me what to check or the manual fallback.
- The official Anthropic docs on Projects and on Claude for Financial Services are a bonus path if I want to go deeper, they are never required to run this vault.
First message to me: ask "which build is this, a public company off filings, a private company off management accounts, or a target under diligence? And where does your data live right now?" Then start with what I answer.
Step 2 · run it on your data
Step 1 set it up. These 12 prompts do the work.
the vault
The 12 prompts
Grab the whole pack as one file, or tap any prompt below to copy it on its own. Placeholders that look like {{THIS}} get swapped for your own numbers — and if you ran Step 1, Claude fills them in for you.
One .md file · all 12 prompts, numbered, in order · nothing left out.
<role>
You are the setup concierge for a 3-statement modeling engagement. You do NOT build anything yet. You ask, wait, and route.
</role>
<task>
Stand up my board of advisors and confirm how we work before any modeling starts. Ask the questions below IN ORDER. Ask the DATA-SOURCE question and WAIT for my answer before proposing any analysis.
</task>
<questions>
1. ADVISOR PERSONA, confirm you will run as a three-person board: a Goldman-style first-year in the model mechanics, a controller who owns the tie-out and balance check, and a PE deal-team associate who reads it for an IC.
2. TYPE, which build is this? (A) public company off filings (B) private company off management accounts (C) a target under diligence (D) a quick operating model, no full balance sheet yet.
3. DATA SOURCE (critical, wait for my answer): (A) upload to a Claude Project (B) paste raw in chat (C) Claude Excel/Word add-in (D) governed connector to a warehouse or data room (E) a mix.
4. TOKENS, capture and read back: entity name, historical periods available, forecast years, currency, reporting basis (US GAAP / IFRS).
5. OUTPUT BAR, confirm every figure traces to a source cell or filing line, every assumption is labeled ASSUMPTION, nothing hardcoded hides inside a formula.
</questions>
<constraints>
- Do not build in this step. Setup only.
- If I skip the DATA-SOURCE answer, ask again.
- State plainly: my data stays in my tenant, nothing is uploaded, stored, or seen by consultance.ai.
</constraints>
<review_gate>
Do not advance to prompt 02 until TYPE and DATA SOURCE are answered and the tokens are read back.
</review_gate>
<role>You are the controller. You turn raw filings or management accounts into a clean historical spine.</role>
<task>From the data source selected in prompt 01, extract the historical income statement, balance sheet, and cash flow across all available periods. Normalize line items to a standard chart. Flag one-offs and reclasses. Do not forecast yet.</task>
<output_format>
1. Historical IS, BS, CF as three aligned tables.
2. A reclass and normalization log: every adjustment, the source line, and why.
3. Gaps where a filing line was missing, marked NEEDS INPUT.
</output_format>
<constraints>
- Use the data source from prompt 01. Do not invent a figure to fill a gap, mark it NEEDS INPUT.
- Every number cites its source line. Hold to the stated reporting basis.
</constraints>
<review_gate>Stop and show me the normalization log before building. I confirm the reclasses before you continue.</review_gate>
<role>You are the Goldman-style first-year. You build the income statement the other two statements hang off.</role>
<task>Build the forecast income statement on the historical spine from prompt 02. Drive revenue off a stated assumption, not a flat guess unless I ask. Carry margins, opex, D&A, interest (placeholder until prompt 05), and taxes.</task>
<output_format>
- Forecast IS, historical and forecast periods side by side.
- An assumptions block, each line labeled ASSUMPTION with value and rationale.
- Interest flagged PLACEHOLDER pending prompt 05.
</output_format>
<constraints>
- Use the data source and spine from prompts 01 and 02.
- Every driver is a labeled, visible assumption. No growth rate buried in a formula. No hardcodes inside a calculated cell.
</constraints>
<review_gate>Confirm the revenue driver and margins before linking the balance sheet.</review_gate>
<role>You are the controller. You build the balance sheet so it BALANCES every period or tells you where it broke.</role>
<task>Build the forecast balance sheet linked to prompt 03. Roll working capital off DSO, DIO, DPO, roll PP&E off capex and D&A, link retained earnings to net income and dividends. Leave debt and cash as links to prompts 05 and 06.</task>
<output_format>
- Forecast balance sheet, all periods.
- A balance check row: Assets minus (Liabilities + Equity), must be zero every period.
- A working-capital driver block, each a labeled assumption.
</output_format>
<constraints>
- Use the data source from prompt 01. Link, do not paste, from prompt 03.
- If a period does not balance, STOP and report the exact line. Do not plug it.
</constraints>
<review_gate>The balance check must read zero for every period before you proceed.</review_gate>
<role>You are the first-year building the debt schedule that closes the circularity: interest feeds the income statement, which feeds cash, which pays down debt.</role>
<task>Build the debt schedule: opening balance, mandatory amortization, cash sweep if any, interest on average or opening balance per my choice, closing balance. Feed interest back to the prompt 03 placeholder and the prompt 06 cash flow.</task>
<output_format>
- Debt schedule, all tranches, all periods.
- The interest number that replaces the prompt 03 placeholder, labeled.
- A note on how circularity is handled.
</output_format>
<constraints>
- Use the links from prompts 01, 03, 04. Interest basis is a labeled assumption I choose. No hardcoded interest.
</constraints>
<review_gate>Confirm the interest basis and circularity handling before flowing cash.</review_gate>
<role>You are the controller. You build the cash flow OFF the income statement and balance sheet, never as a standalone guess.</role>
<task>Build the cash flow statement (operating, investing, financing) derived from the linked IS and BS in prompts 03 to 05. Ending cash must tie to the balance-sheet cash line every period.</task>
<output_format>
- Full CF statement, indirect method, all periods.
- A tie row: CF ending cash minus BS cash, must be zero every period.
- Financing section pulls debt movements from prompt 05.
</output_format>
<constraints>
- Use the linked model from prompts 03 to 05. Derive, do not re-enter numbers.
- If ending cash does not tie to the balance sheet, STOP and report the period and line.
</constraints>
<review_gate>The CF-to-BS cash tie must be zero every period before the model is called linked.</review_gate>
<role>You are an independent reviewer who did NOT build the model. You re-derive the checks a second way and hard-stop on any mismatch.</role>
<task>Independently verify the linked model. Re-derive, from raw inputs NOT the model's output: (1) net income to retained-earnings movement, (2) the balance check, (3) CF ending cash to BS cash, (4) debt closing balances to the schedule. Compare each to the model.</task>
<output_format>
A reconciliation table: Check, Re-derived value, Model value, Match?, and the line that breaks if no.
Then one verdict line: RECONCILED or BROKEN.
</output_format>
<constraints>
- Re-derive from raw inputs, do not copy the model's numbers as your check.
- Any mismatch beyond rounding = BROKEN. Name the exact line and period. Do not smooth it over.
- Do not output RECONCILED unless all four checks match.
</constraints>
<review_gate>The model is not board-ready until this returns RECONCILED. If BROKEN, fix upstream and rerun. This is the sign-off gate.</review_gate>
<role>You are the PE deal-team associate stress-testing the case for an IC.</role>
<task>Run base, downside, and upside scenarios by flexing the labeled assumptions from prompts 03 to 05. Show the effect on revenue, EBITDA, free cash flow, and the debt paydown path. Build a one-variable sensitivity on the driver I name.</task>
<output_format>
- Scenario comparison table across key outputs.
- A sensitivity grid for the named driver.
- A read on which assumption the case is most fragile to.
</output_format>
<constraints>
- Flex only labeled assumptions. Re-run prompt 07 for each scenario, a scenario that does not reconcile is not shown.
</constraints>
<review_gate>Show me which single assumption breaks the case first before writing any narrative.</review_gate>
<role>You are the controller preparing the model for a reviewer who will pull on every formula.</role>
<task>Produce an audit checklist: no hardcodes inside calculated cells, every driver a labeled assumption, all three statements linked not pasted, balance check and cash tie hold, debt schedule feeds interest back. List any line that fails.</task>
<output_format>
- A pass/fail checklist by area (IS, BS, CF, debt, links, assumptions).
- A hardcode hunt: any calculated cell holding a typed number.
- Reviewer notes: where human judgment is still required.
</output_format>
<constraints>Use the built model from prompts 03 to 07. Flag, do not fix silently.</constraints>
<review_gate>Every FAIL must be resolved or explicitly accepted by me before the model leaves the room.</review_gate>
<role>You are the PE deal-team associate turning the reconciled model into a board or IC read.</role>
<task>From the RECONCILED model (prompt 07 must have passed), write the pack: a one-page case summary, the three statements in presentation form, key assumptions, the scenario read, and the two or three risks the numbers surface.</task>
<output_format>
1. One-page executive read. 2. Presentation-form IS, BS, CF. 3. Assumptions appendix (labeled). 4. Risks and what would change the answer.
</output_format>
<constraints>
- Do not produce this pack unless prompt 07 returned RECONCILED.
- Every figure traces to the model. No new number appears that the model did not produce.
</constraints>
<review_gate>This pack needs a named human sign-off before it leaves the firm. State that on the cover.</review_gate>
<role>You are the first-year who defends every assumption when a partner or IC member pushes back.</role>
<task>For each labeled assumption, state the value, the source or rationale, and the range a reasonable reviewer would accept. Flag judgment calls versus filing-anchored assumptions.</task>
<output_format>A table: Assumption, Value, Source/rationale, Accepted range, Judgment call? Then a paragraph on the three most likely to be challenged.</output_format>
<constraints>Use the labeled assumptions from the built model. Anchor to filings where possible, mark the rest as judgment.</constraints>
<review_gate>Confirm the accepted ranges before this memo backs a decision.</review_gate>
<role>You are the controller rolling the model to a new period when fresh actuals land.</role>
<task>Drop the newest actuals from the prompt 01 data source into the historical spine, re-anchor the forecast off the new base, and re-run prompt 07. Report what moved and why.</task>
<output_format>
- Updated model with the new actual period folded in.
- A what-changed log: actuals vs prior forecast, variance, and the driver behind each.
- Fresh prompt 07 reconciliation verdict.
</output_format>
<constraints>Use the data source from prompt 01. Preserve labeled assumptions unless I change them. Do not present until prompt 07 returns RECONCILED again.</constraints>
<review_gate>Show me the variance log before re-anchoring the forecast.</review_gate>
Got the prompts. Want them wired into your actual stack? We map that on a free AI audit.
• Run last quarter's numbers first. Live data is not a test bed.
• Nothing here uploads to us. It runs in your own Claude account, on your own machine.
• A named human reviews and signs every output before it reaches a board, lender, or client.
• Mask account numbers and names to the minimum the task needs.
the fine print
Straight answers on ownership
Prompt set authored by consultance.ai. Goldman Sachs is referenced as the analyst standard the vault matches, no affiliation implied. Model figures in the examples are illustrative. Your filings stay in your own Claude tenant and we never see it. A named human verifies every figure against source and signs before the model is used. Not accounting, audit, or investment advice.
Want this running in your business, not just your laptop? We build it and hand you the keys.
Free 3-Statement Model Builder is a finance and data build in the consultance.ai AI Build Library. For CFOs and PE deal teams: 12 prompts that build a tied-out income statement, balance sheet, and cash flow from your own filings inside Claude Opus 5. A prompt vault that replaces a first-year analyst grind and Wall Street Prep templates. It fits CFOs, controllers, finance leaders, and PE deal teams who need a linked, tied-out three statement model built from filings without paying a first-year analyst six figures to keep it alive. Setup difficulty is Medium, with 4 plain-English steps.
What does Free 3-Statement Model Builder do?
For CFOs and PE deal teams: 12 prompts that build a tied-out income statement, balance sheet, and cash flow from your own filings inside Claude Opus 5. A prompt vault that replaces a first-year analyst grind and Wall Street Prep templates.
Who is Free 3-Statement Model Builder for?
It fits CFOs, controllers, finance leaders, and PE deal teams who need a linked, tied-out three statement model built from filings without paying a first-year analyst six figures to keep it alive.
How hard is Free 3-Statement Model Builder to set up?
Medium to set up — one guided setup instruction covering 4 plain-English steps, plus 12 ready-to-run prompts on the resource page.
How would consultance.ai build this out?
The vault is about 60% of the build, a clean linked model you can trust. We do the last 40%: wire it live to your accounting system or filings feed, lock role permissions, build an audit trail your auditor accepts, and handle multi-entity and multi-currency at production scale. Done with you, then handed over so you own it.
What are the licensing terms?
Prompt set authored by consultance.ai. Goldman Sachs is referenced as the analyst standard the vault matches, no affiliation implied. Model figures in the examples are illustrative. Your filings stay in your own Claude tenant and we never see it. A named human verifies every figure against source and signs before the model is used. Not accounting, audit, or investment advice.
Want this built into your workflow?
Free 3-Statement Model Builder is the starting point. On a free AI audit we map where it fits your stack and what consultance.ai would build around it.