00_featured_image_1200x630 | eFinancialModels

Financial Model, IT Services, Online Businesses, SaaS

Annual Recurring Revenue (ARR), Average Revenue per User (ARPU), Break-Even, Budget, Budgeting, Business Valuation, Calculator, Cap Table, Cash Flow Analysis, Cash Flow Projections, Cost Calculation, Cost Projections, Customer Acquisition Costs (CAC), Excel, Financial Analysis, Financial Feasibility, Financial Model, Financial Modeling, Financial modelling, Financial Planning, Financial Projections, Financial Ratios, Financial Reporting, Financial Statements, Financing, Forecast, Forecasting, Fundraising, Income Statement, Investment Multiple, Investor Cash Flows, Profitability Analysis, Revenue Projections, Scenario Analysis, Startup Financial Models, Three Statement Model | Three Financial Statements, Valuation

What this is

A financial model for AI SaaS companies at pre-seed and seed stage. Three-year
monthly build, 18 sheets, roughly 140 input cells collected on a single
Assumptions sheet. Enter your assumptions once; the P&L, cash flow, balance
sheet, SaaS metrics, valuation, cap table and a one-page investor summary
follow from them.

The problem it solves

A classical SaaS model treats cost of revenue as hosting plus support. Both
scale slowly and predictably, which is why software gross margin sits at 75–85%
and everybody plans around it.

An AI product does not behave that way.

Inference cost moves with usage, not with headcount. A small group of heavy
users consumes several times what the average user does, so blended
assumptions understate the true cost of serving. Humans in the loop — review,
escalation, quality assurance — are cost of delivering the product, yet most
models bury them in OpEx, where they quietly inflate gross margin. And LLM
pricing is not a constant you can hard-code for three years.

Run an AI business through a standard SaaS template and gross margin comes out
too high, LTV comes out too high, and payback comes out too short. The plan
looks fundable right up to the point where growth stops paying for itself.

What is inside

Setup — Cover with model settings, colour conventions and a six-point
health check that must read zero for the model to balance. Map of all sheets.
Defaults sheet comparing suggested starting values across the four pricing
models.

Input — one Assumptions sheet, seven sections, with a Suggested column
showing the industry-typical value for whichever pricing model you selected.

Calculation — Revenue Calcs (per-segment customer and MRR waterfalls,
variable revenue, deferred revenue, renewal cliff), COGS Calcs (traditional
layer, AI layer with variance buffer, human-in-the-loop, services, heavy-user
warning), OpEx Calcs (headcount, loaded compensation, HITL deduction,
R&D/S&M/G&A allocation, stock-based compensation).

Output — P&L, SaaS Metrics dashboard, Cash Flow (indirect method with a
runway flag), Balance Sheet with a balance check, Annual Summary.

Analysis — Scenarios with conservative/base/optimistic plus a renewal-cliff
stress test and inference-cost sensitivity; Valuation via revenue multiples,
VC method and a football-field range; Cap Table with an ownership waterfall and
MOIC/IRR per investor class; Cohort retention with implied LTV evolution.

Deliverable — a one-page A4 investor summary, formatted to drop into the
appendix of a pitch deck.

How it differs from a standard SaaS template

Cost of revenue is split into a traditional layer and an AI layer, and gross
margin is reported for each separately — so you can see the moment the AI layer
starts dragging the blended number down.

Inference cost carries an explicit variance buffer, and the model shows what
happens to margin if LLM cost rises 20% or 50%.

Heavy users are modelled with their own multiplier instead of being averaged
away, and the model raises a warning when heavy-user concentration passes the
point where blended assumptions stop being safe.

Human-in-the-loop cost is deducted from OpEx and placed in COGS, where it
belongs.

Revenue is built from four pricing components — per agent, per activity, per
output, per outcome — which can be switched on independently or combined,
because AI products rarely price on seats alone.

Every benchmark used in the model is listed in a companion file with its
source: a16z, Bessemer Cloud Index, Stripe, High Alpha, Carta, Zuora/Ibbaka,
David Skok, Craft Ventures, Scale Venture Partners and others. No unsourced
numbers.

Who it is for

Founders of AI SaaS companies preparing a pre-seed or seed round who have no
CFO on the team and need numbers that survive a due-diligence conversation.
Also useful to operators pressure-testing pricing before a change, and to
advisors who need a defensible starting structure.

What is not included

Not a DCF model — no WACC, no terminal value. No comparable-company analysis.
No liquidation preferences or anti-dilution modelling. No multi-cohort vintage
tracking. Each of these is named in the sheet where it would sit, with a note
on what to use instead.

Not industry-specific beyond AI SaaS. Not a business plan, and not a template
for a funding application.

What you receive

AI SaaS Financial Model Template — Excel, 18 sheets, all cells unlocked
Populated demo workbook — the same model filled end to end
Financial Glossary — PDF, 36 terms, each with its AI-specific twist
User Manual — PDF, sheet-by-sheet walkthrough
Benchmark Sources — Excel, 87 benchmarks with sources
Disclaimer — PDF

Vanilla Excel throughout: no VBA, no macros, no Power Query, no array formulas.
Tested in Excel 2016 and later, LibreOffice 7 and later, and Google Sheets
import.

Disclaimer

This model is a planning instrument, not a forecast. Every output is a
consequence of the assumptions entered, and the assumptions are yours.
Benchmarks are industry reference points for comparison, not targets and not
guarantees. Nothing in this file constitutes financial, investment, legal or
tax advice. Author: Dmitry Perelygin, Fractional CFO, ACMA/CGMA, MBA
Manchester.

You must log in to submit a review.