← back
The Price of Intelligence - AI Agent Pricing in 2025
Takeaway
Pricing should match audience and use case: outcome-based for confidence, simple/predictable always, and acknowledge that AI usage scales over time so encourage the workloads you want.
Summary
- CJ (co-founder/CTO of Orb, usage-based billing infra) surveys 2025 AI agent pricing patterns across customer support, sales, enterprise search, and dev tools.
- Outcome-based pricing examples: Intercom Fin charges $0.99 per resolution on top of tiered plans; ChargeFlow takes a percent per recovered chargeback (Roi-guaranteed).
- Tiered + credits + axes hybrid: Unify and Cursor mix seat-based prices with credit usage; Cursor distinguishes completions vs requests, fast vs slow, premium models (GPT-4o, Claude 3.5 Sonnet).
- Audience signaling: Clay's logo gardens (OpenAI, Airbnb, Anthropic) plus people-search counts target high-growth GTM teams; Replit shows free + technical addons (Autoscale, checkpoints); Hebbia and ServiceNow push 'book a demo' because they're cross-selling to enterprise.
- Devin (Cognition) charges per month plus ACUs (compute units) — adds buyer-side friction translating engineering hires into ACUs.
pricingai-businessagents
Original description
AI agents are reshaping industries, but pricing them is one of the hardest decisions startups face. Get it wrong, and you’ll either leave money on the table or burn cash subsidizing your heaviest users. This talk dives into how AI agent companies are evolving their pricing models—from OpenAI’s API price cuts to Jasper’s unlimited plans to Clay's prospecting credits. If you’re building an AI agent—or thinking about how to price one—you need to understand not just what to charge, but why pricing must evolve as your product, costs, and market mature. This talk gives you a clearer framework for pricing AI agents sustainably while staying competitive in an industry where costs, value, and expectations are constantly shifting.