Top AI Prompts for Finance Teams to Save Time and Money

The best AI prompts for finance give the model four things: a role to play, the data and assumptions to work from, one specific task, and the format you want back. You then check the result before anyone relies on it. Finance teams use these prompts for forecasting, the monthly close, cash flow planning, budget versus actual and variance analysis, and CFO or board reporting.

Below are 38 copy-ready prompts for SaaS and startup finance work, grouped into nine jobs. Each one says what to paste in, and each group ends with what to check, because AI still gets financial numbers wrong.

Key Takeaways

• 59% of finance leaders said their finance function uses AI in Gartner’s 2025 AI in Finance Survey of 183 CFOs and senior finance leaders, taken May to June 2025, up from 58% in 2024.
• Among finance functions using AI in that survey, the most common uses were knowledge management (49%), accounts payable automation (37%) and error and anomaly detection (34%).
• Presenting the same 2025 survey in June 2026, Gartner said 84% of finance organizations had implemented or planned to implement AI, but only 7% reported a high or very high impact.
• The Association for Financial Professionals’ prompting framework for finance, updated October 2026, has six parts: role; context, data and assumptions; task; requirements; output; and review.
• In the Finance Agent Benchmark published in May 2025, the best model tested, OpenAI’s o3, answered 46.8% of 537 expert-written finance questions correctly, so every AI figure needs checking.

How to Write an AI Prompt for Finance

The Association for Financial Professionals (AFP) publishes a six-part prompting method for finance work, shared by Paul Barnhurst of The FP&A Guy at an AFP FP&A Series event. A short prompt is fine for a simple question, but for analysis and models, where consistency and accuracy matter, give the AI all six:

• Role. Who the AI should be: “Act as an FP&A analyst supporting the CFO.” The role sets the vocabulary and the level of detail.
• Context, data and assumptions. The business situation and the actual numbers, pasted or attached: a ledger export, a billing report, last year’s plan. Give only what the task needs.
• Task. One specific analysis, calculation or deliverable, not “help with our finances”.
• Requirements. The rules: use formulas instead of typed numbers, follow your metric definitions, and flag missing assumptions instead of inventing them.
• Output. The format, audience and level of detail: a table with months across the columns, a one-page summary, three bullet points for the board.
• Review. How the result gets checked. For complex work, ask the AI to describe its approach before it starts, then validate the calculations, assumptions, sources and conclusions yourself.

Worked example: from one line to a usable prompt

Ask “forecast our revenue for next year” and the AI has to guess your growth rate, churn and pricing, so you get a tidy table built on invented numbers. The complete version takes two minutes longer to write:

Act as an FP&A analyst at a B2B SaaS company. Attached is our MRR by month for the last 24 months, split into new, expansion, contraction and churned MRR. Build a 12-month MRR forecast with a base, upside and downside case. Before the forecast, list every assumption (new MRR per month, monthly churn, expansion) in a table and explain where each comes from in the data. Use formulas in a downloadable spreadsheet, not typed numbers. Flag any month in the history that looks like an outlier instead of smoothing it. Output: the assumptions table, then the forecast by month for each case.

The data part matters most, and it works best when the records are connected: the deal, the contract, the invoice and the payment for the same customer. A finance founder who builds AI checks for enterprise finance teams put it this way:

…in finance it’s all about the linking. You want to make sure that the quote is connected to the opportunity in Salesforce, which connects to the right set of invoices or billing, how you bill the customer. And only when you do that, you can actually ask the data questions and check the data.

Ahikam Kaufman, CEO, Safebooks ($1.5M ARR). Watch at 11:34

In practice: export clean tables (CSV or a spreadsheet, not screenshots), keep the column names, and include the IDs that join one table to another. Before you paste anything, check that the AI tool is one your company has approved for financial data, and leave out personal details the task doesn’t need, such as employee names and home addresses or customers’ bank details.

What AI Does Well in Finance, and Where It Needs You

Most finance teams now use some AI. In Gartner’s 2025 AI in Finance Survey of 183 CFOs and senior finance leaders, taken from May to June 2025, 59% reported using AI in their finance function, up from 58% a year earlier. Among those using it, the most common uses were knowledge management (49%), accounts payable process automation (37%) and error and anomaly detection (34%).

Use is not the same as results. Presenting the same survey in June 2026, Gartner said 84% of finance organizations had implemented or planned to implement AI, yet only 7% reported a high or very high impact.

The prompts below lean on what language models do reliably: structuring a model or a report, drafting commentary and emails, explaining a calculation, and flagging rows that look unusual in data you supply. What stays with you is the final number, any accounting judgment, and anything that goes to a board, an investor, a lender or a tax authority.

38 AI Prompts for Finance, by Job

The nine groups map onto the four steps of a finance month: close the books (monthly close, expense tracking), explain what changed (financial analysis, budget vs actual), update the forecast and the cash plan (forecasting, cash flow, budgeting, modeling), then report (CFO and board reports, investor updates). Each step’s output is the next step’s input: the closed books feed the variance analysis, which feeds the forecast, which feeds the report, and next month starts again with the close. Replace everything in [brackets] with your own details, and paste or attach the data each prompt names. The prompts are written for general assistants such as ChatGPT, Claude, Gemini or Microsoft Copilot.

Job	What the prompts produce	What to paste in
Financial forecasting	Revenue forecasts, driver models, scenarios, reforecast checks	Monthly MRR or revenue history, drivers, the original plan
Cash flow planning	13-week cash forecast, burn and runway, collections	Bank balance, AR and AP aging, payroll and debt schedule
Budgeting	Annual budget, budget vs actual, hiring plan cost	Prior-year actuals, budget, headcount plan
Financial analysis	P&L review, unit economics, anomalies, cohorts	P&L, ledger export, customer and cohort data
Financial modeling	Three-statement model, model review, sensitivities	Assumptions or an existing model file
Monthly close	Close checklist, reconciliations, deferred revenue, accruals, flux	Bank CSV, ledger, invoices, trial balances
Expense tracking	Categorized transactions, policy checks, software spend	Card and bank transactions, chart of accounts, policy
CFO and board reports	CFO one-pager, board deck outline, board questions, KPI dashboard	Month’s KPIs, budget vs actual, your notes
Investor updates and fundraising	Investor update, term sheet review, data room list	Metrics, the term sheet, documents you have

AI Prompts for Financial Forecasting

Forecasting prompts work when the AI gets real history and states its assumptions, so you can argue with them.

Driver-based forecast. For when revenue depends on a few measurable drivers.

Act as a SaaS finance lead. Our revenue depends on these drivers: [website visitors, trial signups, trial-to-paid conversion, average contract value, monthly logo churn]. Here are the last 6 months of each: [paste]. Build a driver-based forecast for the next 4 quarters. Show how a 10% change in each driver moves ending ARR and rank the drivers by impact. Return a table and three sentences on which driver to watch.

Scenario plan. For a board discussion about what could go wrong.

Act as a CFO preparing for a board meeting. Using the attached model, build three scenarios for the next 18 months: [we lose our largest customer], [planned hires start one quarter late], [monthly churn doubles for two quarters]. For each, show monthly ending cash, the first month cash falls below [3] months of net burn, and the decision we would need to make and by when. Keep every other assumption from the base case and list them.

Reforecast against the original plan. For catching a forecast that slips a little every month.

Compare our latest forecast with the original annual plan and with every reforecast since. Data: [original plan by month], [each reforecast with its date], [actuals to date]. For new ARR, churn, operating expenses and ending cash, show the cumulative change from the original plan, not only from the last version. Flag any line lowered in three or more reforecasts in a row and explain what that pattern suggests.

Seasonality check. For businesses whose bookings swing by month.

Using [24 to 36] months of monthly bookings [paste], find any seasonality, estimate its size by month, and show how next year's forecast changes with seasonality versus a straight-line trend. Explain the method in plain English so I can rebuild it in a spreadsheet.

Forecast review. For a second opinion on a forecast someone else built.

Act as a skeptical CFO. Review this revenue forecast and its assumptions [paste or attach]. List the five assumptions most likely to be wrong, what in our history supports or contradicts each, and the question I should ask the person who built it.

The reforecast prompt exists because of a pattern founders describe. In a talk in Founderpath’s resource library, Rami Essaid said his previous company, Distil Networks, went into 2017 planning to add about $15 million of new ARR. Four months later the forecast was down to about $9 million, after a series of small cuts in February, March, April and May, and nobody went back to compare it with what they had planned in December and January. Watch at 18:42.

Check the output: recalculate one month of the forecast by hand from the assumptions the AI listed. If it doesn’t tie, ask for the formula behind that cell before you use anything else in the table.

AI Prompts for Cash Flow Planning

Cash prompts need exact dates and amounts, so they work best on exports from your bank and your AR and AP reports.

13-week cash flow forecast. For a weekly view of cash and the weeks that get tight.

Act as a treasury analyst. Build a 13-week cash flow forecast from: opening bank balance [amount], expected customer receipts by week [paste AR aging and billing schedule], payroll dates and amounts, rent, vendor bills due [paste AP aging] and debt payments. Show receipts, payments and ending cash by week, highlight any week below [minimum cash], and list three payments we could move to fix it.

Burn and runway. For the number every founder gets asked.

Here are our operating cash inflows and outflows for the last 6 months [paste] and our current cash balance [amount]. Calculate gross burn, net burn and runway in months, once at the latest month's rate and once at the 3-month average. Show each formula and its inputs, then name the expense categories that grew fastest.

Collections plan. For overdue invoices.

Here is our accounts receivable aging [paste]. Group overdue invoices by customer and days past due, estimate the cash we can realistically collect in the next 30 days, and draft a polite reminder for invoices 30 days late and a firmer one for 60 days late. Use only the fields in the table; do not invent customer details.

Cash versus revenue. For a quarter where revenue grew but cash fell.

Explain why our cash balance fell this quarter while recognized revenue grew. Data: [income statement], [balance sheet at the start and end of the quarter], [cash flow statement]. Walk through working capital line by line (receivables, deferred revenue, payables, prepaid expenses) and put a dollar amount on each.

Check the output: tie opening and ending cash to the bank statement. Then run the same inputs through Founderpath’s runway calculator or burn rate calculator: they use fixed formulas, so a different answer means the AI used different inputs, a different definition or wrong arithmetic.

AI Prompts for Budgeting

Budget prompts are mostly arithmetic and structure, which is where a clear format request saves the most time.

Annual operating budget. For a first draft by department.

Act as a SaaS finance lead. Draft next year's operating budget by department (R&D, sales, marketing, customer success, G&A) from this year's actual spend [paste], our hiring plan [paste] and a revenue target of [amount]. Keep total operating expenses at or below [x]% of revenue. Show each department by quarter and list the assumption behind every line over [amount].

Budget vs actual. For the monthly or quarterly variance review.

Compare budget and actual for [period] using this export [paste]. For every line off by more than [5]% or [$ amount], show the variance in dollars and percent, say whether it looks like timing (it will reverse) or a lasting change, and draft a one-line explanation for the board pack. Sort by absolute dollar variance.

Hiring plan cost. For what new hires do to burn.

We plan to hire [roles, start months, salaries]. Calculate the fully loaded monthly cost for the next 12 months, adding [payroll taxes and benefits]% to salary, and show the effect on monthly burn and on runway from [current cash].

Check the output: add up the departments and the quarters yourself, or ask the AI to put the budget in a spreadsheet with formulas. The totals should match the revenue target and the expense cap you gave it.

AI Prompts for Financial Analysis

Analysis prompts should ask the AI to explain its working, because the explanation is where you catch its mistakes.

P&L review. For the monthly look at what changed.

Act as an FP&A analyst supporting the CFO. Here is our P&L for the last 12 months [paste]. Summarize the three biggest changes in gross margin and operating expenses, explain the likely drivers using only this data, and list the questions I should ask each department head. Flag anything that looks miscategorized.

Unit economics. For CAC, payback and LTV on your own definitions.

Using this data [new customers by month, sales and marketing spend, average revenue per account, gross margin, monthly churn], calculate CAC, CAC payback in months and LTV to CAC by quarter. Show each formula and its inputs in a table, and note where a common alternative definition would change the result.

Anomaly scan. For a ledger export before the close or an audit.

Review this general ledger export [attach CSV] for anomalies: duplicate payments, round-number entries, entries posted on weekends, first-time vendors paid more than [amount], and expenses coded to an unusual account for that vendor. Return a table of flagged rows with the reason for each. Do not change or summarize the data; only flag.

Benchmark comparison. For seeing where you stand against peers.

Act as a SaaS finance analyst. Compare our [growth rate, gross margin, net revenue retention, burn multiple] with published benchmarks for private B2B SaaS companies at [ARR range]. Cite each benchmark with its source, link and date, and note where its definition differs from ours. If you can't find a source for a benchmark, say so instead of estimating.

Cohort retention. For how retention changes across customer cohorts.

Here is our revenue by customer cohort (signup month) for the last 24 months [paste]. Build a net revenue retention table by cohort, describe how retention differs for newer cohorts, and point out any cohort that behaves differently from the rest. Explain the calculation in two sentences.

Check the output: ask the AI to list every number it calculated and the rows it came from, then spot-check two. Open every benchmark source it cites; drop any you can’t find.

AI Prompts for Financial Modeling

AI can draft a model or explain one, but you own what it produces. Our guide to the SaaS financial model covers what a good one contains.

Three-statement model. For a first build from your assumptions.

Act as a financial modeling expert. Build a three-statement model (income statement, balance sheet, cash flow) for a SaaS company with [ARR], [growth rate], [gross margin], [hiring plan] and [billing terms: monthly or annual upfront]. Put every assumption on one assumptions tab, use formulas rather than typed values, and make the balance sheet balance. Before building, describe the tabs and structure and wait for my approval.

Explain an existing model. For a model you inherited.

Attached is our financial model. Explain how it works tab by tab: which inputs drive revenue, where headcount costs flow, and how cash is calculated. List any typed numbers inside formulas, circular references, and assumptions that are not on the inputs tab.

Sensitivity table. For the two assumptions that matter most.

Using our model, build a sensitivity table of ending cash in 12 months, with monthly churn of [1%, 2%, 3%] down the rows and new MRR per month of [amounts] across the columns. Add one paragraph on which combinations take cash below [minimum].

Funding options model. For comparing equity with debt.

Model two ways to fund [amount] of growth spending: an equity round at a [pre-money valuation], and a [term]-month loan at [rate]% with [repayment terms]. Assume the company is worth [value] at the end of year [3]. Over those [3] years, compare founder ownership, the total cost of each option and monthly cash. Show the formulas.

Check the output: open the formulas. A balance sheet that balances because of a typed plug figure doesn’t balance. AFP’s own model-building example warns not to expect a correct model on the first attempt, so start from a structure you trust, such as Founderpath’s SaaS financial model template, and check dilution figures against the dilution calculator.

AI Prompts for the Monthly Close

Close prompts handle the repetitive part: matching, scheduling and drafting explanations. Our SaaS accounting guide explains the revenue rules behind them.

Close checklist. For a close that lives in someone’s head.

Act as a controller at a [size] SaaS company using [QuickBooks, Xero or NetSuite]. Build a month-end close checklist with an owner, a due day (business day 1 to 5) and the evidence each task needs: bank and card reconciliations, revenue and deferred revenue, payroll accruals, prepaid expenses, fixed assets and the review sign-off. Format it as a table I can paste into a spreadsheet.

Bank reconciliation. For matching the bank to the ledger.

Reconcile our bank statement for [month] [paste CSV] with the cash account in our ledger [paste CSV]. Match by date, amount and description, list unmatched items on each side, and suggest the likely match or the journal entry needed for each. Do not net items together.

Deferred revenue roll-forward. For annual and multi-month contracts.

Using these invoices [customer, invoice date, amount, service start and end dates], build a deferred revenue roll-forward for [month]: opening balance, new billings, revenue recognized evenly over each service period, and closing balance, by customer. Check that closing deferred revenue matches our ledger balance of [amount] and list any difference.

Accruals. For expenses that arrive after month end.

Here are vendor bills received after month end [paste] and our recurring monthly expenses [paste]. Identify expenses that belong in [month] but weren't booked, propose the accrual entries with accounts and amounts, and list anything you're unsure about separately.

Flux analysis. For explaining account movements.

Compare this month's trial balance with last month's and the same month last year [paste all three]. For every account that moved more than [10]% and more than [$ amount], show the change and draft a one-line explanation. Mark any explanation you are guessing at as needing confirmation.

One founder building AI for accounting teams describes the requirement: every step has to be visible to the person who signs off.

…it cannot be a black box AI. We call our AI output and methodology a glass box AI, which means everything is not just auditable but is explainable throughout each and every step along the way, which is a necessity for accounting teams specifically.

Tal Kirschenbaum, Co-Founder and CEO, Ledge. Watch at 20:23

Watch the full interview with Tal Kirschenbaum, with transcript.

Check the output: a reconciliation ends at zero difference, with every matched pair traceable to a line on both sides; an AI summary of a reconciliation is not a reconciliation. Check deferred revenue against a schedule such as Founderpath’s deferred revenue schedule.

AI Prompts for Expense Tracking

Expense prompts turn messy transaction lists into categories and exceptions you can review in minutes.

Categorize transactions. For card and bank feeds.

Categorize these transactions [paste date, merchant, amount, memo] into our chart of accounts [paste account list]. Return the original rows with a suggested account and a confidence label (high, medium or low). Put low-confidence rows on a review list instead of guessing.

Expense report review. For policy checks.

Review these employee expense reports [paste] against our expense policy [paste]. Flag out-of-policy items, missing receipts, duplicates, and purchases split to stay under an approval limit. Return a table with the employee, the item and the rule it breaks.

Software spend review. For subscriptions and renewals.

Here is every software subscription we pay for [vendor, plan, seats, monthly cost, renewal date] and our usage export [paste]. Show annual spend by category, renewals due in the next 90 days and seats we pay for but don't use. Draft a renewal negotiation email for the three largest contracts.

Monthly spend summary. For department heads.

Summarize this month's spending [paste ledger export] by department and vendor, compare it with the 3-month average, and write a five-bullet note for department heads on what changed.

Check the output: the categorized list should have the same number of rows and the same total as the export you pasted. If you don’t have a clean account list yet, start from a SaaS chart of accounts.

AI Prompts for CFO and Board Reports

Reporting prompts save the most time on structure and first drafts. Keep the metric definitions fixed; our SaaS metrics pages define the common ones.

CFO one-pager. For the monthly summary to the CEO.

Act as the CFO of a [stage] SaaS company. Using this month's and last month's numbers [MRR, new, expansion and churned MRR, gross margin, net burn, cash, runway, headcount], write a one-page CFO report: five headline metrics with the month-over-month change, three things that went well, three risks and the decisions we need. Plain language, and no adjective without a number behind it.

Board deck outline. For the quarterly board meeting.

Draft the outline of our quarterly board deck [company context]. Sections: summary and asks, KPIs against plan, revenue and retention, spending and cash, team, product, risks. For each slide, name the chart or table it needs and the data I have to pull. Keep it to [12] slides.

Board questions. For preparing before the meeting.

Act as a skeptical board member at a SaaS company. Read our board pack [attach] and list the ten hardest questions you would ask about the numbers, ranked by importance, with the data I should have ready for each.

KPI dashboard design. For a dashboard everyone reads the same way.

Design a monthly KPI dashboard for a SaaS CEO and board. Pick at most 10 metrics from [list], give the formula and data source for each, suggest a chart type, and write a one-line definition so everyone reads each metric the same way.

Check the output: every figure in the report should match the source system for the same date range. AI writing makes weak numbers sound confident, so read the draft against the table, not on its own.

AI Prompts for Investor Updates and Fundraising

Fundraising prompts help with structure and preparation. Anything legal still needs a lawyer.

Monthly investor update. For a consistent update investors can scan.

Write our monthly investor update from these facts: [MRR, growth, cash, runway, wins, misses, hires, asks]. Structure: a three-line summary, a metrics table, highlights, lowlights and asks. Make the lowlights as specific as the highlights. Under 400 words.

Term sheet review. For understanding an equity or debt term sheet.

Act as an experienced startup finance advisor, not a lawyer. Read this term sheet [paste] and explain each economic and control term in plain English: amount, valuation or price, liquidation preference, board seats, protective provisions, warrants, covenants, fees and prepayment. Flag unusual terms or terms that need a lawyer, and list the questions I should ask the investor or lender.

Data room checklist. For due diligence.

Build a due diligence data room checklist for a [Series A or debt facility] for a SaaS company: the folders, the documents in each, and which are usually requested first. Mark the items we already have from this list [paste].

Investor objections. For pitch preparation.

Here is our pitch and our metrics [paste]. List the ten most likely investor objections and draft a factual answer to each using only our data. Where our data can't answer an objection, say so.

Check the output: an AI reading of a term sheet is a starting point for questions, not legal advice. Our guide to debt covenants explains the terms lenders most often attach.

For the marketing side of the business, see our AI prompts for marketing.

How to Check AI Finance Output Before You Use It

Language models still get finance questions wrong often. The Finance Agent Benchmark, published in May 2025, has 537 expert-written finance research questions, built on a task taxonomy developed with experts from banks, hedge funds and private equity firms, and gives the models tools including web search and SEC filings. The best model tested, OpenAI’s o3, answered 46.8% correctly, at an average cost of $3.79 per question. Models have improved since: on a revised version of the benchmark, Vals AI’s leaderboard (updated June 2026) shows the top model at 64.37%, so about one answer in three is still wrong. The lesson for a finance team is the same: the AI drafts, and a person checks.

Four checks catch most problems:

• Trace every number. Each figure should come from a file you gave the AI or a source you can open. If it can’t say where a number came from, re-run the prompt with the data attached, and don’t use the number until it can.
• Tie the totals. Row counts, column totals, cash and revenue should match your ledger, bank or billing system for the same period. If they don’t, fix the inputs before anything else.
• Read the assumptions. Ask for them in a table before the result, and make sure they are yours, not the AI’s defaults; if they aren’t, replace them and re-run.
• Recompute one result. Rebuild one month, one ratio or one reconciliation line yourself, or in a spreadsheet with formulas.

Then a named person signs off, as they would for work from a junior analyst. AFP’s framework puts it plainly: validate the calculations, assumptions, sources and conclusions before the output is used in a financial decision.

Fund What the Numbers Show

Clean forecasts, a tight close and clear board reporting also make funding easier: every source of capital checks the same revenue, retention and cash figures these prompts help you prepare.

Founderpath provides non-dilutive capital for bootstrapped software founders. Connect your billing, banking and accounting data, and Founderpath shows you what you qualify for, with every figure disclosed up front. There’s no pitch deck, no equity and no board seats.

Check what you qualify for. It takes less than 5 minutes.

AI Prompts for Finance FAQ

What are AI prompts for finance?

AI prompts for finance are written instructions that tell an AI assistant such as ChatGPT, Claude, Gemini or Microsoft Copilot what finance work to do, such as a forecast, a reconciliation or a board report. The best ones give the AI a role, the data and assumptions, one task, rules to follow and the output format, and a person reviews the result before anyone uses it.

How do I write a good AI prompt for financial analysis?

Give the AI a role (for example, an FP&A analyst supporting the CFO), attach the actual data, ask for one analysis, set rules such as using formulas and flagging missing assumptions, and say what format you want back. The AFP framework adds a sixth step: review the calculations, assumptions, sources and conclusions yourself before you use them.

Can ChatGPT or Claude do financial forecasting?

They can draft a forecast from historical data you provide, build scenarios and explain the method, which makes a useful first version. They can also make arithmetic and logic errors, so have them list their assumptions, ask for formulas rather than typed numbers, and recompute at least one period yourself before relying on the forecast.

Can AI build a financial model?

AI can draft a three-statement model or explain an existing one, especially inside a spreadsheet with an AI assistant. Ask it to describe the structure first, keep all assumptions on one tab and use formulas. Expect to correct the first version: a model you present is still your responsibility.

Is it safe to put company financial data into an AI tool?

It depends on the tool and the account. Use the AI tool your company has approved for financial data, check whether it uses your inputs for training and how long it keeps them, and leave out personal details the task doesn’t need, such as employee names and home addresses or customers’ bank details.

How accurate is AI for finance work?

Not accurate enough to skip review. In the Finance Agent Benchmark published in May 2025, the best model tested, OpenAI’s o3, answered 46.8% of 537 expert-written finance research questions correctly. On a revised version of the benchmark, the top model scored 64.37% as of June 2026, so checking every number against its source is still the rule.

What should a prompt for a CFO report include?

The month’s key metrics with the prior period (revenue or MRR, growth, gross margin, burn, cash and runway), what went well, the main risks and the decisions needed. Tell the AI to use only the numbers you supplied, to keep it to one page and to back every adjective with a number.

Which AI tool is best for finance prompts?

The prompts in this guide work in general assistants such as ChatGPT, Claude, Gemini and Microsoft Copilot. Assistants built into spreadsheets help most with models and formulas. Choose the one your company has approved for financial data, and judge it on your own tasks rather than on a general ranking.