AI Pricing Strategy
AI pricing strategy is the discipline of mapping unstable, usage-driven cost-of-goods (tokens, GPU minutes, embedding calls) onto a price plan customers will accept. Three dominant patterns exist: (1) Bundled — AI features included in existing seat prices, marketed as 'AI-powered,' margin absorbed by the rest of the SKU. (2) Add-on / power user — a separate AI add-on (e.g., $20/seat/month) that mirrors how Microsoft Copilot, GitHub Copilot, and Notion AI shipped. (3) Usage / metered — pay per call, per token, per generation, common in API products and emerging in apps for heavy-action features (image generation, video, agents). The right pattern depends on COGS volatility, customer sophistication, and whether AI is a feature or a product.
The Trap
The trap is bundling AI into an existing seat price without modeling COGS volatility. When usage spikes 5x because customers love it, gross margin collapses from 75% to 30% overnight and finance forces an emergency repricing — which damages trust more than charging correctly from day one. The opposite trap is putting every AI feature on a metered meter, scaring customers with unpredictable bills, and watching adoption stall. The KnowMBA POV: bundle the feature, meter the agent — anything that takes autonomous action with non-trivial compute should be metered, anything that augments a human within a session should be bundled.
What to Do
Run this 5-step process. (1) Estimate COGS per active user per month at three usage tiers (P50, P90, P99). (2) Decide if your typical heavy user is profitable at the proposed price; if not, you need a usage cap or a metered tier. (3) Choose the pattern: feature → bundle, capability → add-on, autonomous action → meter. (4) Build a soft cap (warning at 80%, throttle at 100%) before a hard cap or overage. (5) Re-price quarterly for the first year — AI COGS are still moving 30-60% per year and your initial pricing is wrong.
Formula
In Practice
Microsoft 365 Copilot launched at $30/seat/month in late 2023 — a deliberate add-on price roughly equal to the underlying M365 E3 seat itself, signaling AI as a premium capability rather than a free feature. GitHub Copilot priced at $10/seat (individual) and $19/seat (business) — low enough to drive adoption, high enough to cover GPU costs at typical usage. Cursor and similar AI-native IDEs priced at $20/seat/month with metered overages on advanced models. The pattern: anchor the AI price near the value of an additional human-hour saved per month, not the cost of compute.
Pro Tips
- 01
Publish a 'fair use' policy with explicit numerical thresholds before customers find them. Surprise throttling generates more support tickets than transparent caps. Anthropic, OpenAI, and Cursor all publish per-tier limits — copy that pattern.
- 02
Price the AGENT, not the call. As you ship agents that take multi-step autonomous action (research, code, refactor), shift toward outcome-based or task-based pricing. A successful 'PR review completed' is worth $5-50 to the customer; the 12 LLM calls that produced it are not the unit they value.
- 03
Watch your top 1% of users obsessively. The P99 usage curve is where margins die. If your top 1% are using 30x the median, you have a structural pricing problem masked by averages.
Myth vs Reality
Myth
“AI pricing should be cost-plus on tokens”
Reality
Token costs are dropping 40-70% per year (Anthropic, OpenAI, Google, open models). If you cost-plus today, you are forced into awkward price cuts every quarter, eroding the perceived value of your product. Price on customer value (hours saved, decisions accelerated), not your inputs.
Myth
“Customers prefer usage-based pricing because it feels fair”
Reality
Customers say they prefer it; they actually hate the bill anxiety. Adoption data from Snowflake, AWS, and OpenAI all show flat-rate or capped tiers driving more usage and less churn than pure metered pricing. Use metered pricing only where the customer can directly tie usage to value (e.g., API products).
Try it
Run the numbers.
Pressure-test the concept against your own knowledge — answer the challenge or try the live scenario.
Knowledge Check
You're launching an AI writing assistant in your existing $25/seat/month SaaS. Heavy users will burn $8 of tokens/month, average users $2, light users $0.50. What's the smartest pricing move?
Industry benchmarks
Is your number good?
Calibrate against real-world tiers. Use these ranges as targets — not absolutes.
AI Feature Add-On Price (per seat/month)
B2B SaaS AI add-ons (2024-2026 market data: Microsoft Copilot, GitHub Copilot, Notion AI, Cursor)Premium / Copilot Tier
$20-30
Mainstream Add-On
$10-20
Adoption-Driven
$5-10
Bundled / Free
$0
Source: Hypothetical: synthesized from public pricing pages of Microsoft 365 Copilot, GitHub Copilot, Notion AI, Cursor, Linear AI
Real-world cases
Companies that lived this.
Verified narratives with the numbers that prove (or break) the concept.
Microsoft 365 Copilot
2023-2026
Microsoft launched 365 Copilot at $30/seat/month — roughly the price of an entire M365 E3 license — making it one of the boldest AI pricing moves in enterprise software history. The price signaled premium capability, anchored AI as worth a full additional seat, and gave Microsoft margin to absorb GPU cost volatility. Adoption was slower than the marketing suggested, but the per-seat economics held up, and competitors (Google, Salesforce) followed Microsoft's pricing anchor rather than racing to the bottom.
Launch Price
$30/seat/month
Anchor
≈ price of M365 E3 itself
Pricing Pattern
Premium add-on (not bundled)
Pricing AI as a premium add-on protects the base SKU and gives you margin headroom. Bundling AI for free into existing seats may drive adoption but destroys your ability to ever charge for it later — and AI COGS demand pricing flexibility.
Cursor (Anysphere)
2024-2026
Cursor priced its AI-native IDE at $20/seat/month with included usage of frontier models, then a metered overage for premium models (Claude Opus, GPT-4 Turbo). This hybrid pattern — flat included quota plus pay-as-you-go overage — became the dominant pricing pattern for AI developer tools. The included quota covered ~90% of users with predictable bills; the metered overage captured value from heavy users without scaring everyone else.
Base Price
$20/seat/month
Pattern
Flat + metered overage
Coverage
~90% of users stay in flat tier
Hybrid pricing (flat base + metered overage) gives you the predictability customers want with the upside metering customers' usage justifies. It is the safest default pattern for any AI product where COGS varies more than 10x between light and heavy users.
Related concepts
Keep connecting.
The concepts that orbit this one — each one sharpens the others.
Beyond the concept
Turn AI Pricing Strategy into a live operating decision.
Use this concept as the framing layer, then move into a diagnostic if it maps directly to a current bottleneck.
Typical response time: 24h · No retainer required
Turn AI Pricing Strategy into a live operating decision.
Use AI Pricing Strategy as the framing layer, then move into diagnostics or advisory if this maps directly to a current business bottleneck.