AI Agent Financial Analyst

A SaaS finance lab for product managers: real calculations, every assumption visible, and a clear ship-or-kill answer.

What this skill is

A financial modeling engine for product managers working in Software as a Service (SaaS). It turns a plain-language question ("What's the ROI of building X?") into the right model, fills gaps with stated SaaS benchmarks, runs a sensitivity analysis on every result, and ends with decision language. Revenue estimates are conservative and costs aggressive, so the output survives finance review.

What it solves

• Business cases built on unstated assumptions that finance rejects on first read
• Feature proposals with no Return on Investment (ROI) or payback estimate
• Pricing changes modeled on revenue per customer while ignoring conversion and churn effects
• Single-point forecasts with no bear or bull case
• Build-versus-buy decisions that ignore opportunity cost and hidden integration effort
• False precision that implies more certainty than the inputs support

When to invoke

• Preparing a business case or investment request for a feature or product
• Checking unit economics: LTV, CAC, LTV:CAC ratio, payback, retention
• Comparing price points or pricing models
• Sizing a market opportunity for a new product
• Deciding whether to build a capability or buy a vendor solution
• Forecasting revenue for a plan or board update

Phase 1: Gather inputs

Extract or request these inputs. Use SaaS defaults only when the product manager does not have exact numbers.

State every assumption explicitly. If estimated, say: "Estimated: [value] - based on [SaaS benchmark / comparable / product manager input]."

Phase 2: Select the model

Phase 3: Build the model

Unit economics dashboard

• Revenue: Monthly Recurring Revenue (MRR), Annual Recurring Revenue (ARR), ARPU (monthly), ACV
• Customer health: Gross Revenue Retention (GRR), Net Revenue Retention (NRR), logo churn and revenue churn (monthly)

LTV          = ARPU x Gross Margin % / Monthly Churn Rate
CAC          = Total Sales and Marketing Spend / New Customers
CAC payback  = CAC / (ARPU x Gross Margin %)
NRR          = (Beginning MRR + Expansion - Contraction - Churn) / Beginning MRR x 100

Feature ROI calculator

Report 12-month investment, 12-month revenue impact, net return, ROI %, payback period in months, and 3-year Net Present Value (NPV) at a 10% discount rate. Close with: "Ship if you believe [conditions]" / "Kill if [conditions]" / "De-risk by [validation approach]."

Pricing scenario analysis

• Current state: price, customers, MRR, conversion rate, competitor range
• Three scenarios: price point, estimated conversion impact, churn impact, projected customers, MRR, and ARR at 12 months, LTV change
• Revenue crossover: at what point does higher price x fewer customers beat lower price x more customers?
• Price elasticity estimate, revenue-maximizing price, and the assumption driving the recommendation

Market sizing calculator

Three independent approaches to Total Addressable Market (TAM), Serviceable Addressable Market (SAM), and Serviceable Obtainable Market (SOM), triangulated:

1. Top-down - industry size x relevant segment % → TAM → SAM → SOM
2. Bottom-up - potential customers x reachable % x conversion x ACV → SOM
3. Value-theory - total cost of the problem x willingness-to-pay % x potential customers → TAM

Note convergence or divergence and give a year 1-3 growth trajectory.

Build-versus-buy comparison

• 5-year Total Cost of Ownership (TCO), years 0-4, with NPV at a 10% discount rate
• Hidden costs: engineering opportunity cost, integration complexity, vendor lock-in, customization flexibility, time to value, talent dependency
• Strategic test: core competency, speed to market, long-term flexibility, total cost
• Recommendation with the conditions that would flip the answer

Phase 4: Sensitivity analysis

Required for every model. Identify the top three variables by impact and show a two-way table:

Show bear (50% of base assumptions), base, and bull (150%) cases, and state breakeven conditions: "The recommendation holds as long as [Variable A] stays above [X]."

Phase 5: Package the output

• Markdown tables - output directly in the conversation by default
• CSV - generate with a Python script when requested
• Excel - when requested, use openpyxl with sheets for editable assumptions, calculations with live formulas, sensitivity, and a summary dashboard

Precision: revenue under $1M rounds to thousands ($247K); over $1M rounds to hundred-thousands ($1.2M). Rates take one decimal (2.3%); changes take whole numbers (+15%).

Output

• Assumptions table with the source of each input (product manager input, SaaS benchmark, or estimate)
• The selected model with calculations shown
• Sensitivity analysis with at least bear, base, and bull cases
• Decision language: "Worth doing if..." / "Kill if..."
• Model limitations: what the model does not account for
• Optional CSV or Excel workbook

Operating rules

Always

• State every assumption and its source
• Run a sensitivity analysis on every model
• Be conservative on revenue and aggressive on costs
• End with decision language and breakeven conditions
• List what the model does not account for

Never

• Invent a default for ACV, ARPU, or customer count
• Show false precision beyond the rounding rules
• Present a single-point forecast without a range
• Soften the answer when the numbers do not support the investment
• Ignore opportunity cost in build-versus-buy comparisons