The Price of Intelligence: AI Agent Pricing in 2025
Why your pricing model is already obsolete, the shift to outcome-based pricing, cost commoditization, and the race to unlimited. A comprehensive analysis of 13+ real pricing strategies.
Pricing really just can't be a set it and forget it exercise. Most companies think about pricing like this once a year or maybe twice a year change. Instead you want to think about pricing like this continuous exercise and evolution.
— Chz, Co-founder and CTO, Orb
The New Pricing Paradigm
From Cost-Plus to Value-Based
The most confident AI companies are pricing on outcomes, not usage. This signals product maturity and creates natural ROI conversations with customers.
The insight: When Intercom charges $0.99 per resolution, they're not pricing on tokens—they're pricing on successful outcomes. This aligns incentives and shows confidence in their AI agent's performance.
Watch Intercom pricing analysisOutcome-Based Pricing
Charge for results: $0.99/resolution, percentage of recovered revenue
Continuous Evolution
Pricing changes multiple times per month, not annually
Cost Commoditization
Token costs plummeting—OpenAI costs down drastically in 18 months
Your margin is your opportunity — the famous Jeff Bezos framework applied to AI agents. Companies that understand this are investing in inference optimization and model routing to unlock pricing power.
— Chz, referencing Jeff Bezos
13 Real Pricing Models Analyzed
Intercom Finn
$0.99 per resolution — you only pay for successful customer support resolutions
"You're only paying for the success of fin"
Chargeflow
Percentage per recovered chargeback — no fee until recovery succeeds
"You're not paying anything until chargeback recovery is successful"
Cursor
$20/$40 tiers with hidden complexity: completions vs requests, fast vs slow, premium models cost extra
"Completions, requests, fast requests, slow requests, GPT-4.40, Claude 3.5 Sonet..."
Devon (Cognition)
Monthly + ACUs — compute resources scaling with VM time, inference, networking
"Pretty hard to translate engineering work into acus"
Clay
Credits system — from $100 trials to multi-million dollar enterprise deals
"Prepaid credits generalize across very different use cases"
Jasper
Unlimited credits — model decision engine routes to OpenAI/Anthropic/Cohere
"Built model decision engine to pull from OpenAI, Anthropic, Cohere"
The Four Pricing Principles
1. Consider Your Target Audience
Match pricing to their buying process
SMB: Individual developer, credit card, self-service checkout.
Enterprise: Procurement teams, budgeting against traditional solutions, sales-led process.
Key insight: "It's not just who they are but the whole buying process they're involved in"
2. Simplicity & Predictability (Never Compromise)
Two things that almost never should be compromised
People have easier time making purchasing decisions when pricing is simple. Predictability of spend is crucial with usage-based models.
Quote: "Two things that almost never should be compromised are Simplicity and predictability"
3. Think About Use Cases You're Encouraging/Discouraging
Pricing shapes behavior
What workloads is your product a great fit for? Those are the sorts of things you want to encourage. Pricing should align with your ideal customer workloads.
Strategic insight: Use pricing to guide customers toward your product's strengths
4. Defend Your Margin (But Don't Over-Index on Cost)
Costs will change rapidly
Cost will actually change pretty rapidly — OpenAI's cost per token has decreased drastically over the last 1.5 years. Take the R&D innovation and pass that down in your cost structure.
Action: "Take the R&D, the ultimate technical innovation of your company and pass that down in your cost structure"
Watch cost optimization examplesCase Studies: Technical Innovation → Pricing Power
Character.AI: Inference Optimization for Scale
Character.AI serves 20% of Google search queries with 100M daily active users spending almost an hour per day on the platform.
The bet: In 2022-2023, they invested heavily in inference infrastructure optimization to make their business model viable.
The lesson: For consumer products with heavy usage, R&D in inference cost reduction isn't optional — it's existential.
Jasper: Model Routing Enables Unlimited Pricing
Jasper moved to unlimited credits because marketers couldn't think in "word counts" — they needed to iterate fast on marketing copy.
Technical implementation: Built a "model decision-making engine" that pulls from OpenAI, Anthropic, Cohere to pick the right model for each job.
Strategic insight: Technical innovation (model routing) enabled pricing innovation (unlimited), which reinforced their value proposition.
Devon: The Buyer Cognitive Load Challenge
Devon charges monthly + compute resources (acus scaling with virtual machine time, inference, networking bandwidth).
The problem: "It's pretty hard to translate that [engineering work] into acus... adding these new paradigms does have some cost associated with it because now buyers need to think about workload and translate that into acus"
The lesson: Introducing new pricing metrics creates buyer cognitive load — you're asking them to learn a new language.
2025 Predictions
1. Price Wars Continue
"Race to the bottom on some pricing and some verticals" — as inputs get commoditized and competition increases
2. Push Toward Unlimited Plans
More companies offering "effectively unlimited plans" as inputs commoditize and competition heats up
3. Outcome-Based Pricing Expansion
"We will lean into this outcome or success based pricing but we'll need to get real about what exactly are the guarantees and the SLAs"
4. Increased R&D in Monetization
Underappreciated trend: companies need customer control (throttling, spend caps), visibility into credit burn-down, and usage auditing
The Technical Challenge at Every Stage
Input Stage:
High-volume data infrastructure for metering
Core Stage:
Billing business logic with versioning
Output Stage:
Financial accounting and invoicing
Additional:
Managing customers on legacy price points
Key Takeaways
- Stop pricing on tokens — price on outcomes your customers actually care about
- Build pricing flexibility into your architecture from day one (versioning, migrations)
- Invest in inference optimization — it unlocks pricing power, not just cost savings
- Use prepaid credits as a discounting engine — single conversion rate vs. multiple SKU discounts
- Invest in billing infrastructure — versioning and migrations as first-class features
- Audit capabilities are essential — users need granular visibility into usage
- Support complex business logic — enterprise agreements, discounting, ramps layered on usage
- Plan for legacy price points — customers will stay on old pricing
- Target your buyer's journey — SMB wants self-service, Enterprise wants sales involvement
- Tell a story with your pricing page — logo gardens, metrics, use cases signal who you're for
- Commoditize your inputs — model decision engines, inference optimization unlock new pricing possibilities
- Prepare for margin pressure — competition + VC expectations = need for cost innovation
- Demand usage controls — throttling, spend caps, visibility into credit burn-down
- Require auditing capabilities — granular visibility into where AI spend is going
- Negotiate for flexibility — pricing will evolve; lock in ability to migrate to better models
- Prefer outcome-based pricing — aligns incentives and shows vendor confidence
Companies Analyzed
Intercom Finn
Cursor
Clay
Replit
Jasper
Cognition/Devon
Character.AI
OpenAI
Anthropic
Cohere
MongoDB
Google DeepMind
Video Reference
The Price of Intelligence - AI Agent Pricing in 2025
Chz, Co-founder and CTO, Orb — Usage-based billing infrastructure company
Duration: ~20 min
Event: AI Engineer Conference
Video ID: In7K-4JZKR4
Analysis: Full 48,122 token transcript
Pricing Volatility Disclaimer
All pricing models and examples mentioned were current as of the AI Engineer Conference 2024. Given the rapid evolution of AI agent costs and the "continuous pricing evolution" paradigm discussed, specific pricing details may have changed.
The principles remain valid: Outcome-based pricing, flexibility, and cost innovation are enduring strategies regardless of specific price points.