Most SaaS companies have shipped AI. Few are getting paid for it. In six weeks I help you decide what to charge for, how to package it and how to launch it, using your customers’ data and what the market has already learned.
Book a free diagnosticAlmost every software company is shipping AI. The revenue isn’t following yet, and the pricing playbook is being rewritten as we go.
Where most software companies now set their AI revenue targets, down from the 15–20% growth executives had forecast.
CPOs who say fewer than 30% of their customers actively use their AI features.
Software companies now using hybrid pricing, the most common model, up from 25% a year earlier.
AI builders using usage-based pricing, in just six months. Outcome-based rose from 18% to 23%.
Typical gross margin on AI products, against 80–90% for classic SaaS. Inference costs have to be priced in.
Software companies that changed their pricing or packaging in the past year.
GitHub Copilot’s flat $10 plan reportedly lost more than $20 per user on average, with some users consuming about $80 of compute. GitHub later added premium-request limits, a $39 tier and overage fees.
Salesforce’s €2-per-conversation price was reversed because conversations varied too much in value. Intercom refined what counts as a “resolution” after customer feedback.
Intercom Fin charges $0.99 per resolved ticket, against roughly $10 for a human resolution. Fin’s share of resolved tickets rose from 15% to 45% in five months.
76% of AI buyers say they prefer usage-based pricing; only 10% prefer outcome-based. Predictability matters to the person signing.
A platform fee that covers your costs, plus usage or outcome tiers above it, is the safest starting point when you are unsure.
29% of software companies already sell AI credits and another 33% plan to within a year. They buy time while you learn what customers value.
Founders often double their costs and underprice. Price against the value and the alternative, then make sure inference, support and onboarding costs still leave a margin.
If an agent replaces work people did, a per-seat price shrinks as the product succeeds. Agents need a unit tied to the work done.
Clay changed its pricing about twice a year on its way from $2M to $37M in revenue. One or two changes a year tends to work best.
| Model | Share of software companies planning it | Fits when | Watch out for |
|---|---|---|---|
| Bundled in a premium plan | 42% | AI has broad appeal and predictable cost | Margin erodes if heavy users dominate |
| Seat-based add-on | 41% | Value is per person, as with copilots | Shrinks when agents replace work people did |
| Usage or credits | 39% | Usage varies a lot between customers | Buyers fear surprise bills; needs clear limits |
| Outcome-based | 35% | The result is easy to measure and attribute | Hard to define, forecast and contract |
| Hybrid: platform fee plus usage | 45% plan 2+ methods | You are unsure, or serve very different segments | Complexity; keep it simple at 10 and 1,000 customers |
Shares from the Simon-Kucher Global Software Study 2025. Most companies combine more than one: ICONIQ finds AI builders blend 1.7 pricing models on average (ICONIQ 2026).
Usage data by account and feature, a review of customer calls and tickets that mention AI, and short interviews with sales and success.
You get: who uses what, who doesn’t, and the reasons, in customers’ own words.
Two or three packaging and pricing options, each modeled against your current customers: revenue, usage, cost to serve and likely pushback.
You get: a recommended model, the price points and the margin behind them.
The product fixes that unlock usage, the message for customers, migration rules for existing contracts, and the talk track for sales and success.
You get: a launch plan your team can run, and the numbers to track.
Usage by feature, segment and plan, so you know where you start.
The model, the unit, the price points and the fences between plans, with revenue and margin modeled.
How existing customers move, what is grandfathered and how you avoid a renewal backlash.
Product fixes, customer message, sales and support training, and dates.
Accounts using AI each week, repeat use, revenue attached to AI and AI gross margin.
I led product strategy at Zuora, the subscription billing and monetization platform, through its IPO, and have priced and packaged SaaS products for 25 years. This offer applies the same method as the Roadmap Sprint, listening to every customer conversation and ranking by revenue, to one question: what customers will pay for in your AI.
No, you decide. I bring the evidence, the options and the modeled impact so the decision is fast and defensible.
It depends on how broad the appeal is and how variable the cost is. Broad appeal and predictable cost favor bundling into a premium plan; uneven usage and high cost favor an add-on or usage component.
Only if the outcome is measurable, clearly caused by your product and predictable enough for buyers to budget. Most companies start hybrid and move toward outcomes as they learn.
Credits are a good bridge while you learn what customers value. Keep them simple, set clear limits, and plan how you will move beyond them.
Then the plan says so. Often the fix is in the product, onboarding or positioning before it is in the price.
Usage data for the AI features, your AI cost data, access to customer calls and tickets, and an hour each with a few sales and success people.
Tell me about your product and your team. You’ll leave with three things I’d fix first, and an honest answer on whether I can help.