Prompts that do the work.
You do the thinking.

Reconcile this bank statement PDF. Follow these steps precisely:

STEP 1 — Extract Statement Metadata
Extract: Bank Name, Account Holder Name, Account Number (last 4), Statement Period, Opening Balance, Closing Balance.

STEP 2 — Extract and Audit All Transactions
Build a structured table: Date | Description | Debit (-) | Credit (+) | Running Balance | Type | Extraction Notes
Assign Debit/Credit strictly from the statement columns. Cross-check every line — Debit always reduces, Credit always increases the running balance.

STEP 3 — Categorize and Flag Transactions
For each transaction: Transaction Type | Likely GL Category | Confidence (High/Med/Low) | Flag/Review Needed

STEP 4 — Statement Reconciliation & Subtotals
Populate: Opening Balance + Total Credits – Total Withdrawals – Total Service Charges = Calculated Closing Balance. Flag any variance vs. stated closing balance.

STEP 5 — Reconciliation & Audit Summary
Summarize deposits, withdrawals, fees, interest, and closing balance. List all flagged or ambiguous transactions.

STEP 6 — Output & Documentation
Deliver a CSV/Excel download with columns: Date, Description, Debit (-), Credit (+), Running Balance, Type, Likely GL Category, Confidence, Flag/Review Needed, Extraction Notes.

Identify unpaid bills over the last 3 months and propose the most likely payment matches based on amount, date proximity, and memo evidence.

For every proposal, show your reasoning (amount / date / memo / vendor) and return a list of ready-to-post cash applications plus any residual balances.

Analyze the Accounts Payable balance for the last month.

Reference the A/P aging report for last month and the balance sheet for last month. Reconcile both sources for each vendor and identify potential discrepancies in amounts, aging buckets, or classifications.

Return:
- Confirmed matching balances or discrepancies requiring investigation (with reasoning)
- Total AP by aging bucket (0–30, 31–60, 61–90, 90+ days)
- High-risk accounts for payment priority

Identify unpaid invoices over the last 3 months and propose the most likely receipt matches based on amount, date proximity, and memo evidence.

For every proposal, show your reasoning (amount / date / memo / customer) and return a list of ready-to-post cash receipts plus any residual balances.

Analyze the Accounts Receivable balance for the last month.

Reference the A/R aging report for last month and the balance sheet for last month. Reconcile both sources for each customer and identify discrepancies in amounts, aging buckets, or classifications.

Return:
- Confirmed matching balances and discrepancies requiring investigation (with reasoning)
- Total Accounts Receivable by aging bucket (0–30, 31–60, 61–90, 90+ days)
- High-risk accounts for collection priority

Using expense patterns from the last 3 full months, identify likely unbilled costs to accrue this month (subscriptions, recurring expenses, expected bills).

Propose amount estimate (average / proration), evidence, and journal entries (debit Expense, credit Accrued Liabilities) with next-period auto-reversal. Ensure the journal entries are broken up by vendor.

You are an internal controls assistant reviewing journal entries posted this period.

Flag any entries that match the following risk patterns:
- Round-dollar amounts over $5,000 with no supporting reference
- Entries posted outside business hours
- Entries posted by users who do not typically post to that account
- Entries that reverse or offset a prior entry posted in the last 3 days
- Entries to suspense or clearing accounts that remain open at period end

For each flagged entry, provide the risk reason, severity (high/medium/low), and recommended follow-up action. Ask if there are other known or suspected risk patterns to look for.

Review the current prepaid and accrual schedules alongside this month's GL activity.

Identify:
- Prepaid entries that should have amortized this period but did not post
- New invoices that appear to be prepaid in nature and need a schedule created
- Accruals from prior months that were not reversed
- Accruals that should be set up this period based on recurring vendor patterns

For each item, draft the correcting or new journal entry with account, amount, description, and effective date.

Identify last 3 months' transactions over $2,000 with descriptions, memos, or vendors suggesting computers, hardware, equipment, etc.

Propose which to capitalize and suggest fixed asset entries based on the transactions found, including: asset type, in-service date (if unknown, default to transaction date), useful life, and depreciation method.

Given an uploaded payroll CSV containing employee-by-pay-period summary data, extract the gross pay, employer payroll taxes, net pay, and total payroll cost for the specified employee and pay period.

Draft a manual journal entry using only valid Puzzle chart of accounts:
- Debit 60100 Salaries for the employee's gross pay
- Debit 60500 Payroll Taxes for the employer payroll taxes (not employee taxes)
- Credit 13511 Salaries & Benefits: Clearing for the total payroll cost

Use the pay date from the CSV. Confirm with user before posting. Provide a clickable in-app link to view the manual journal entry. Use only actual CSV amounts — do not estimate or round.

Review last month's transactions for the following risk patterns:

1. MISCATEGORIZATION RISK — vendor coded to a different account than prior months, or description suggests a different account than used.

2. ANOMALOUS AMOUNTS — recurring vendor whose amount this month is more than 25% higher or lower than their trailing 3-month average; or unusually round amounts (e.g., exactly $10,000) with no memo.

3. FIRST-TIME LARGE TRANSACTIONS — any vendor appearing for the first time with a transaction over $2,500 (SMB) or $10,000 (growth-stage).

4. TIMING FLAGS — transactions dated in the last 2 days of the month that are large or first-time; transactions dated on weekends or holidays unusual for this client.

5. DUPLICATE RISK — same vendor, same amount, same or adjacent date as another transaction in the same period.

For each flagged item: Vendor, Date, Amount, Account coded to, Flag type, Reason (1–2 sentences), Suggested action. End with a brief summary: total reviewed, number flagged, top 1–2 issues by risk level.

Review all transactions in my accounts for the past month. Identify any that look like internal transfers — money moving between my own bank accounts, transfers from checking to a credit card account, payments from one treasury/cash account to another, or any scenario where funds are not leaving or entering the company.

Flag any transactions not already linked as transfers, inter-account payments, or bank sweeps. For each, provide: date, description, accounts involved (from/to), amount, and whether it needs reconciliation or linking.

End with specific recommendations: should I link any transactions, create a transfer, or adjust how these appear on my books?

Perform a data quality scan of my accounting ledger focused on:

1. Duplicate vendors/customers — multiple separately-listed names with high similarity (identical except for capitalization or punctuation). Show exact names found.
2. Uncategorized transactions — transactions coded to "Uncategorized," "No Category," or similar. Give specific transaction details.
3. Inconsistent vendor naming — e.g., both "datadog" and "Datadog" exist. Quote the records.
4. Open invoices/bills older than 90 days with no recent activity.
5. Broken or missing class/location tags where actual missing segment data is observed.

Important: Do not infer or suggest issues unless they are visible in the current data. Group and present only facts found directly in the system. After presenting findings, ask me to confirm the actions I want to take.

For the last full month, flag any transactions that are inconsistent with prior months' patterns for the same vendor, including: category, class/location/department, and typical amount range.

Also flag recurring vendors that are missing or appear more frequently than expected.

Group findings by inconsistency type and show the historical pattern versus the current transaction.

For the last three full months:

1. Identify all transactions posted to "No Category," "Uncategorized Revenue," "Uncategorized Expense," or other default holding accounts.

For each found transaction:
- Flag: List date, vendor, amount, transaction descriptor, and a suggested account category based on recent company activity.
- Explain: Compare to prior months' patterns to suggest the most likely correct category and root cause (e.g., sync issue, missing vendor mapping, lock period conflict, or new vendor).
- Include: A direct in-app link to review or edit each transaction.

Draft a summary review request for the owner, highlighting flagged transactions and probable causes. Propose remediation and prompt for confirmation before making any categorization updates.

Please perform these actions on my uploaded bank statement or CSV file:

1. Extract and transform transactions:
   - Remove any rows labeled "Beginning Balance" or "Ending Balance"
   - Remove the "Balance" column
   - Reorder columns: Date, Debits, Credits, Description
   - Add a new "Amount" column: Amount = (Debits as negative) + (Credits as positive). Format to two decimal places with dollar sign.
   - Output only: Date, Amount, Description

2. For each remaining transaction, create a manual transaction in the specified account. Auto-map the category for each transaction based on best fit from the Description (e.g., AWS → hosting, Zoom/Slack → software, CPA → tax_and_accounting).

3. Present a summary table: Date, Amount, Description, Chosen Category.

Extract two fields from each receipt image provided:
- Vendor Name — the business name (restaurant, retailer, service provider, etc.)
- Total Charge — the final amount charged, inclusive of tax and tip if present

Return each with a confidence level (high, medium, low) and a brief reason. Only extract what is visible and unambiguous — never hallucinate or infer values.

For the Total, report the final charge as labeled (e.g., "Total", "Amount Charged", "Total Paid"). If ambiguity exists, flag for review with "medium" confidence.

If a value cannot be determined with at least "medium" confidence, set it to null and explain why. Never return partial totals (subtotals, tax-only lines). Report only the final cardholder charge.

Give me a summary of the business health and 3–5 talking points for the client. Use plain English for the titles and assume the business owner is not familiar with accounting.

You are an advisory intelligence engine analyzing a client's full GL history. Find the single most valuable insight this client has never been told — something true, specific, and actionable that a great CFO would have flagged months ago.

Scan for: non-obvious margin patterns, hidden concentration risks, cash flow behaviors that are costing money silently, positive trends the owner doesn't know to celebrate, and early warning signals that haven't become problems yet.

Select the single highest-value insight. Then write a short, warm, plain-English client email from the accountant delivering this insight — specific numbers included, no jargon, ending with one question that opens an advisory conversation. The email should feel like it came from a trusted advisor who has been paying close attention. Because now they have been.

You are a financial analyst assistant. Given the trial balance for the current month and the prior month, perform a flux analysis for every account with a variance greater than 5% or $1,000 (whichever is lower).

For each flagged account, provide: the dollar and percentage change, a plain-English explanation of the likely cause based on underlying transaction detail, and a confidence level (high/medium/low) in your explanation.

Format the output as narrative commentary suitable for inclusion in a monthly close report for a client.

Summarize the top 5 variances across Revenue, COGS, and OpEx (materiality > $20k or 10%).

For each, show: top contributing categories and the 3 largest underlying transactions. Tag each variance: Seasonal / One-time / Needs follow-up.

Estimate runway in months based on current cash and last month's burn rate. Produce a summary table showing projected cash over time and include the last month of full cash.

Use the "cash ÷ absolute value of burn" approach: Runway (months) = Starting Cash ÷ |Last month's total burn|

Construct a table for each projection month showing: Month, Starting Cash, Burn Amount, Ending Cash, Runway Remaining (months). Run until cash turns negative.

Include brief metric definitions: Runway (months), Burn Amount, Ending Cash.

Using the client's last 6 months of general ledger activity, build a rolling 13-week cash flow forecast. Base inflow projections on historical accounts receivable collection patterns and recurring revenue. Base outflow projections on recurring accounts payable, payroll cadence, and known upcoming obligations.

Identify the lowest projected cash balance and the week it occurs. Forecast in a table by week for 13 weeks.

Then write a 3-paragraph plain-English cash flow narrative the accountant can send to the client, suitable for a business owner with no accounting background.

Flag any weeks where the balance drops below a 4-week operating expense buffer and suggest one specific action to address each gap.

You are a profitability advisor. Using the client's revenue and cost data from the general ledger, perform a margin analysis at the most granular level available — by product line, service type, customer segment, or location if available, else use the income statement.

For each segment, calculate gross margin percentage and contribution to total profit. Rank segments from most to least profitable. Identify the single largest margin drag and explain the likely drivers in plain English.

Then draft three specific, actionable recommendations the business owner could implement in the next 90 days to improve overall gross margin, with an estimated dollar impact for each.

You are a working capital advisor. Using the client's accounts receivable, accounts payable, and inventory data (if available) from the general ledger, calculate the following metrics for the current period and the prior 3 periods: days sales outstanding (DSO), days payable outstanding (DPO), days inventory outstanding (DIO), and cash conversion cycle (CCC).

Identify the metric that has deteriorated most significantly and quantify the cash impact in dollars. Then generate three specific recommendations to improve the cash conversion cycle, ranked by ease of implementation.

Write the output as a client-facing memo with a clear headline number: "Improving your DSO by 10 days would free up $X in cash."

You are a financial appreciation engine. Before the accountant's client call, analyze the client's recent general ledger activity and identify one specific, genuine, data-backed financial behavior worth celebrating.

It must be:
1. True and specific — cite the actual pattern, amount, or streak
2. Non-obvious — not "your revenue went up" but something the owner might not have noticed about themselves
3. Quantified where possible — translate the behavior into its dollar or time impact
4. Human — frame it as a reflection of a decision or discipline, not just a lucky outcome

Deliver it as a single paragraph the accountant can read verbatim or paraphrase to open the call. End with a suggested follow-up question that turns the compliment into a forward-looking advisory conversation.

Give me an honest, unvarnished evaluation of the financial health of my business based on all available financial statements and recent data.

Use every available financial statement and expense trend — not just top-line numbers — to surface any hidden risks, warning signs, or noteworthy anomalies (even small but material ones).

Your review must be brutally honest — no sugarcoating, hedging, or positive spin. Write as if you're warning a founder facing potential failure, not as a consultant trying to keep a client happy.

Highlight all major risks to cash runway, revenue growth, expense structure, and vendor/expense concentration, using specific data points. Provide a concise "if you do nothing, here is the likely outcome" forecast.

Give your top 3 actionable recommendations, each with a concrete, measurable target (e.g., "reduce payroll by 30%" or "categorize all expenses within 30 days"). Use a direct, CEO-to-CEO tone (think Elon Musk). Avoid jargon.

You are a milestone detection engine and chief celebration officer. Monitor the client's GL data for financial milestones — both absolute (first $1M revenue, $100K profit month) and relative (best margin ever, longest streak of profitability, fastest collections month, lowest expense ratio).

When a milestone is detected, generate a celebration announcement that includes:
- The milestone name and what makes it significant
- The exact number that crossed the threshold
- Historical context (how long it took to get here, what the trend looked like)
- A "only X% of businesses reach this" benchmark stat where possible
- A preview of the next milestone on the horizon with current progress toward it

Format as something the owner would genuinely want to share with their team. Tone: warm, proud, specific.

List the top 5 bright spots or strengths in the company's recent financial performance, based on everything you know from my statements, metrics, and trends.

Highlight revenue wins, improvements, cost-saving successes, or anything trending up — even small victories.

Frame each as a quick 'win' a business owner should feel proud of, preferably with a brief rationale or supporting financial fact (e.g., 'Revenue grew by 10% last month' or 'Your software costs are stable and predictable').

Exclude negatives and avoid caveats — this should be an ego boost, plain and simple. Number each point and keep it punchy and motivating.