Notifications
Clear all
Topic starter
10/06/2026 12:09 am
TL;DR: AI companies are changing pricing models up to five times in their first two years, and the Stripe-Metronome acquisition reflects pressure to support usage-based billing, fine-grained metering, and agent-driven purchasing, according to WorkOS. Per-seat pricing weakens when autonomous agents do the work and the buying, so billing becomes an identity and authorisation problem as much as a finance problem.
NHIMG editorial — based on content published by WorkOS: Pricing as product-market fit, Cosmo Wolfe on billing after the Stripe-Metronome acquisition
Questions worth separating out
Q: How should security teams govern AI agents that can initiate purchases?
A: Security teams should treat agent purchasing as a governed non-human identity use case.
Q: Why do per-seat models fail when AI agents do the work?
A: Per-seat models fail because a seat assumes a stable human user, while an AI agent may complete work continuously, on demand, and at machine speed.
Q: What breaks when machine buyers are not tied to identity governance?
A: When machine buyers are not tied to identity governance, teams lose traceability, approval boundaries, and accountability.
Practitioner guidance
- Map agent-triggered transactions to identity events Log which non-human identity initiated each purchase, what scope it had, and whether the transaction was policy-approved before execution.
- Retire seat-based assumptions for agent workloads Review AI features that are still priced as if a human occupies a seat.
- Treat payment permissions as spend entitlements Define explicit limits for machine buyers, including service scope, value ceilings, and approval paths for transactions outside normal patterns.
What's in the full article
WorkOS's full article covers the operational detail this post intentionally leaves for the source:
- The acquisition context behind Metronome's billing expertise and Stripe's payment infrastructure.
- Cosmo Wolfe's full framing for deciding when an agent should be priced as a seat, a task, or a transaction.
- The Machine Payment Protocol discussion and why agents as buyers changes onboarding, checkout, and attribution design.
- The product and business-model implications for teams building AI-native monetisation systems.
👉 Read WorkOS's analysis of AI pricing, agents as buyers, and machine payments →
AI agent buyers and billing models: what should teams change?
Explore further
11/06/2026 1:50 am
AI pricing is becoming an identity governance problem, not just a billing problem. Once software can decide what to buy and when to buy it, the organisation is no longer pricing a user journey. It is governing a non-human actor with execution authority and spending authority. That changes who needs approval, what gets logged, and how accountability is assigned when the buyer is an agent rather than a person. Practitioners should treat agent purchasing as a governed identity behaviour, not a fintech add-on.
A few things that frame the scale:
- AI companies are iterating on pricing five times in the first two years, according to Ultimate Guide to NHIs , 2025 Outlook and Predictions.
- Only 44% of developers are reported to follow security best practices for secrets management, exposing a significant developer behaviour gap.
A question worth separating out:
Q: How do organisations decide whether to use usage-based pricing for AI products?
A: Organisations should use usage-based pricing when the product creates measurable consumption tied to outcomes rather than static access. If the agent’s contribution can be attributed and metered at the action level, usage-based pricing is usually more faithful than seats. If attribution is weak, the model becomes harder to audit and defend.
👉 Read our full editorial: AI agent buyers are reshaping pricing, billing, and product fit
12/06/2026 3:24 am
AI pricing is becoming an identity governance problem, not just a billing problem. Once software can decide what to buy and when to buy it, the organisation is no longer pricing a user journey. It is governing a non-human actor with execution authority and spending authority. That changes who needs approval, what gets logged, and how accountability is assigned when the buyer is an agent rather than a person. Practitioners should treat agent purchasing as a governed identity behaviour, not a fintech add-on.
A few things that frame the scale:
- AI companies are iterating on pricing five times in the first two years, according to Ultimate Guide to NHIs , 2025 Outlook and Predictions.
- Only 44% of developers are reported to follow security best practices for secrets management, exposing a significant developer behaviour gap.
A question worth separating out:
Q: How do organisations decide whether to use usage-based pricing for AI products?
A: Organisations should use usage-based pricing when the product creates measurable consumption tied to outcomes rather than static access. If the agent’s contribution can be attributed and metered at the action level, usage-based pricing is usually more faithful than seats. If attribution is weak, the model becomes harder to audit and defend.
👉 Read our full editorial: AI agent buyers are reshaping pricing, billing, and product fit
