15 AI

SaaS AI exposure test

How much of this software company's revenue could AI agents wear away over the next three years, and what protects it?

For investors, boards and SaaS CEOs. Takes about four minutes. Opens with an example.

The company, the pressure and its defences

1The company

What the product does

Picking a category fills starting assumptions for steps 1 and 2. They're my estimates, not measured data. Replace them with the company's own numbers.

Annual recurring revenue

Currency

Revenue priced per seat or user

Priced on usage, transactions or outcomes

Net revenue retention today

Most revenue comes from

2AI pressure over three years

Work done in the product that agents could take over

Share of that freed work that turns into fewer paid seats

Customers often move freed time to other work and keep the seats for a while.

Seat price change at renewal

Revenue from customers who could build their own with AI tools

Revenue an AI agent or an AI-native rival could take

3Defences

System of recordCustomers' core data lives here Owns a regulated or approval workflowAudit trails, sign-offs, compliance Proprietary data or network effectsGets better as more customers use it Deep integrationsReplacing it means rewiring other systems Already sells usage, agent or outcome pricing Agents can work through itAn open API or MCP server customers' agents use Ships its own agents that customers pay for

The answer

High

Today's revenue from today's customers, three years on

Today's revenue as 100 squares, three years on

Revenue from today's customers, year by year

At today's retentionWith this AI pressure

Questions for management

Take the result with you

A PDF report with your inputs, results, charts and method. Or the same as plain text for an email.

How this tool works

What it does

Most software is sold per seat. If an agent does part of a user's work, customers may buy fewer seats. Some may build their own tool with AI. Some may let an agent or an AI-native rival do the job.

This tool puts a number on each of those over three years. It then credits what the company wins back through usage, outcome or agent pricing, and through its defences.

What you enter

The company. Category, revenue, how it's priced, retention and customer size.
The pressure. Work agents could take over, how much of that becomes fewer seats, price pressure, building in-house, and agents or rivals taking the job.
Defences. System of record, regulated workflows, data, integrations, and new pricing or agents.

What you get

Revenue at risk over three years, as a bridge from today's revenue.
A 100-square map of today's revenue and what happens to each part.
The drag on retention and questions for management, in a PDF.

Try it: untick "Already sells usage, agent or outcome pricing". The seats lost stay lost.

The maths, assumptions and sources

Customers who leave: built in-house = revenue share × size factor (small businesses 0.3, mid-sized 0.8, enterprises 1.2) × defences. Taken by agents or rivals = share × defences. Among customers who stay: seats lost = seat revenue × work agents take over × share that becomes fewer seats × size factor (1.0, 0.9, 0.8). Seat prices then move by the renewal change. Won back = seats lost × the capture from usage pricing (35%), agent access (15%) and its own agents (20%), at most 80%. The yearly drag = 1 − (revenue kept)^(1/3). It doesn't include new customers or price rises.

Defences multiply: a system of record halves building in-house and cuts agent takeover by 30%. Regulated workflows cut takeover by 30% and seats lost by 10%. Data and network effects cut both by 30%. Integrations cut building in-house by 20% and takeover by 10%. Agent access cuts takeover by 20%; its own agents by 15%. The category presets are my judgment, not measured data.

Gartner, cited by Deloitte (2026 predictions): by 2030, at least 40% of enterprise SaaS spend will shift to usage-, agent- or outcome-based pricing.
Gartner, reported July 2026: up to $234 billion, about 20% of enterprise application spend, exposed to "agentic arbitrage" by 2030. Gartner calls it "less an apocalypse and more of a metamorphosis".
Long Angle, Software vs AI, Q1 2026: the SEG SaaS Index fell 25.7% in the first quarter of 2026; Bain's median net revenue retention for public SaaS fell from 117% (2021) to 108% (2025).