Execution → usage-based. Management → subscription. Outcomes → outcome-based.
If you're a B2B AI-native or SaaS transitioning to AI, you're probably struggling with pricing right now.
Here’s a simple way to think about it, through concrete cases.
Case 1: You sell an AI API to developers (like OpenAI)
Think of your AI as Sam, a junior joining your client’s team. Sam executes tasks: drafting emails, summarizing notes, answering prompts. Each request costs you money. Some clients send 10 requests/day, others 10M.
- Usage is extremely variable
- The value is captured by developers building on top of you
- ROI is hard to measure directly
If you charge a flat fee, heavy users kill your margins. So you charge per usage (tokens, calls). It’s not ideal: predictability is low, and it doesn’t encourage usage. But if you're “just” selling execution, you don’t really have a choice (yet). Usage-based pricing is NOT a choice. It’s a constraint.
Case 2: You build a copilot for employees (like Notion AI or Microsoft Copilot)
Think of your AI as Nicole, a manager joining your client’s team. Nicole coordinates work: routing requests, structuring workflows, adapting execution.
Some employees use her all day, others barely, for now.
- Usage varies today, but should stabilize over time
- The value is productivity (time saved per employee)
- ROI exists, but is hard to measure precisely
Companies don’t want unpredictable bills. They want something simple they can deploy at scale.
And you want usage to grow. So you abstract complexity. You charge per seat (with limits to cover your margins).
Case 3: You automate a business task (e.g. generate qualified leads)
Think of your AI as Raz, a team lead replacing an external vendor or internal team member.
Before you, companies were paying for results. Now you deliver those results.
- Usage matters less than outcomes
- The value is explicit (a lead, a contract, a resolved ticket)
- ROI is directly measurable
You charge per outcome.
The mistake most startups make
They don’t look at these mechanics. Instead, they think:
- What are others doing?
- What worked in SaaS?
But the only question that matters is: Where is the value actually created?
From there, a pattern emerges:
- Execution (Sam) → usage-based
- Management (Nicole) → subscription
- Outcomes (Raz) → outcome-based
AI isn’t changing pricing fundamentals. It’s just making it harder to ignore.
