Koshima Satija
Co-founder & COO, Flexprice
Updated February 2026: When we first published this in May 2025, the big story was why usage-based pricing was growing. A year on, the landscape has moved further and faster than expected. Agentic AI has introduced an entirely new pricing category outcome-based billing and the data on seat-based pricing's decline has sharpened considerably. We've updated this post throughout with 2025/2026 research, new company examples, and a new section on outcome-based pricing.
Usage-based pricing is transforming how software and AI services are monetizing their product. Unlike seat-based pricing where companies used to charge a set fee per user account regardless of actual activity, this model charges customers according to actual usage of a product.
Adoption of usage-based pricing among SaaS companies has accelerated rapidly in recent years. According to a survey, the percentage of SaaS companies using any form of UBP rose from 30% in 2019 to about 85% by 2024.
But what has driven this surge in adoption? Let’s break it down:
Usage is Tied to Infrastructure, Not User Count
AI models, especially those using deep learning, are computationally intensive. The primary cost drivers are not the number of users, but the GPU cycles, API calls, and tokens processed.
Unlike a traditional SaaS tool where user count and active sessions are the main cost factors, AI companies face steep infrastructure bills that can scale unpredictably.
In fact, infrastructure costs for some AI businesses have risen from just 10% of their cost to as much as 35-40% as they scale, making efficient cost recovery through usage-based pricing essential.
AI Value Doesn’t Scale Per Seat:
The value derived from an AI product is often not tied to the number of users but rather the volume of work being processed.
For example, one customer might run a few hundred model inferences per month, while another could be processing millions of requests.
In both cases, the number of “seats” might be identical, but the compute load and associated costs can be vastly different. This makes seat-based pricing a poor fit for AI workloads.
Fair and Transparent
Usage-based pricing directly ties cost to value. Customers pay in proportion to the benefit they receive. This alignment builds trust and reduces friction.
In contrast to flat subscription fees, where customers might pay for features or capacity they never use, UBP ensures that if they consume less, they automatically pay less. This perceived fairness encourages customers to start small and scale as needed.
Lower Barrier to Entry for New Customers
Startups and small teams can start with minimal spend and scale naturally, meaning usage-based pricing removes the need to commit to a large upfront payment. This lowers friction and drives faster adoption, particularly in developer-first markets.
Best-in-Class Retention and Upsell Opportunity
Usage naturally increases as customers deepen their engagement. Unlike seat-based models where upsell requires contract renegotiation, usage-based models grow revenue organically, without friction.
The Agentic Shift: Pricing by Work Done, Not by Access
A new force is now accelerating the move away from seat-based pricing: AI agents. Agents don't log in, don't hold licenses, and can complete entire workflows autonomously sometimes thousands of tasks in the time a human completes one. Charging per-seat for an agent is like charging per-parking-space for a self-driving car fleet.
Gartner projects that 40% of enterprise applications will feature AI agents by end of 2026, up from less than 5% today.
As agents take over discrete units of work resolving a support ticket, drafting a contract, booking a meeting the natural pricing unit becomes the outcome achieved, not the number of people with access.
This is the most important pricing frontier of 2025–2026, and it has given rise to an entirely new category: outcome-based pricing.
But there's hesitancy around predictability
Enterprise buyers, especially CIOs and CFOs, want to allocate set budgets for technology. If teams can't determine a fixed spend number, adoption can wane.
The scale of this concern is significant: 78% of IT leaders report unexpected charges from consumption-based or AI pricing models, and 90% of CIOs cite cost forecasting as their top challenge in AI deployment. This is where AI companies have had to be creative:
Cost alert systems notify teams when they approach usage thresholds
Hard usage caps prevent overspend
Prepaid plans like credit bundles or minimum commit allow customers to budget upfront
Pricing models AI companies are using
Since we first published this, the range of pricing models in the wild has expanded meaningfully.
The models below are organised into four categories: freemium & trial-led, add-on & credit-based, tiered usage, and new in 2025 outcome/resolution-based. Existing examples are kept as published; updated and new examples are highlighted.
Freemium & Trial-Led Models
Freemium Model
Buffer
Buffer gives 10 AI credits per month for free with an unlimited paid plan at $10/month. This model reduces the barrier to adoption by giving users a taste of AI features before committing to paid tiers.
Free Trial / Freemium-to-Paid
Intercom Fin has since fully embraced outcome-based pricing at $0.99/resolution now the most-cited AI pricing case study in the industry. Within one year of launch, Fin generated tens of millions in revenue. The per-resolution model is covered in full in the new Outcome-Based Pricing section below.
Add-On & Credit-Based Models
Personal Add-On
Notion: Notion offers an AI add-on for $8/month per member for additional AI features such as writing assistance, summarisation, and Q&A on workspace content.
Workspace Add-On
Slack: Slack charges $10/month additional for the Slack AI feature pack, which adds AI-powered search, channel recaps, and thread summarization across the workspace.
Credit Model
Airtable: Airtable offers additional AI credits for $6 per 100,000 tokens within the platform, enabling teams to scale AI feature usage beyond their plan's included allocation.
Fixed Rate & Tiered Usage Models
Fixed Rate Usage
Algolia: Algolia charges $0.50 per 1,000 search requests with an unlimited searches plan at $1.20 per 1,000 providing a clear, predictable per-unit cost that scales directly with customer usage volume.
Premium Tier
Loom: Loom uses AI features as justification to charge a premium: $16/month vs $12/month for the Business tier. AI-powered summaries, transcripts, and clips are bundled exclusively into the higher tier.
Tiered Usage Model
Miro: Miro charges AI credits that reset monthly based on the plan tier — teams on higher plans receive more AI credits for features like diagram generation and content summarisation.
Controlled Rollout
Asana: Asana has gradually introduced AI features only for Enterprise+ tiers with custom pricing, using a controlled rollout to manage demand and gather feedback before broader release.
Outcome / Resolution-Based Models (New in 2026)
This category did not exist when we first published this post. Outcome-based pricing where customers are charged only when the AI successfully completes a defined task has emerged as the most innovative and alignment-focused model in the market.
Under 10% of AI companies use it today, but it is growing fast and is widely expected to become the dominant model for agentic AI products.
$0.99 per resolution, Intercom Fin AI Agent
The defining case study of outcome-based pricing. Intercom charges $0.99 only when the AI fully resolves a customer conversation no charge for failed attempts.
A resolution is counted when the customer confirms it, or doesn't follow up. Generated tens of millions in revenue within its first year.
The model directly solves the AI paradox: a company deploying Fin may eventually need 80% fewer human support agents so per-seat pricing would have penalised Intercom's own success.
$2 per conversation, Salesforce Agentforce
Salesforce launched Agentforce in late 2024 at $2 per AI conversation explicitly framed as a replacement for human agent cost, which typically runs $30–$50 per interaction.
Charged per conversation regardless of whether the issue is fully resolved, differentiating it slightly from Intercom's model. Hundreds of enterprise customers onboarded within months of launch.
Pay per fully-resolved ticket only, Zendesk AI
Zendesk adopted the most aggressive form of outcome-based pricing: customers are charged only when a ticket is fully resolved by AI, with zero charge for failed attempts.
This transfers all performance risk to Zendesk, making it a powerful enterprise sales differentiator customers have no downside to trialling AI features. The challenge: defining 'resolved' required extensive documentation agreed upfront with each customer.
But implementing these pricing models is hard
Most conventional billing tools like Stripe and Chargebee, while powerful for subscription-based billing, fall short when it comes to handling the dynamic requirements of AI usage-based pricing.
92% of AI companies that started with usage-based pricing have subsequently adjusted their model at least once (Metronome, 2025) pricing is a continuous experiment, and the tooling needs to keep up.
Metronome's 2025 field research estimates it takes 3–6 months of engineering work to build robust usage metering infrastructure from scratch at exactly the moment when founders should be focused on product-market fit. We've seen three core challenges consistently across AI companies:
Real-time metering: Tracking and processing usage data in real-time is computationally intensive and requires specialised infrastructure. A billing event that arrives 30 seconds after the usage creates disputes and under-billing.
Dynamic pricing adjustments: Supporting mid-cycle pricing changes, volume discounts, and custom tiers requires significant engineering effort. Outcome-based models add another layer: what counts as a billable event needs to be defined, tracked, and defended.
Multi-dimensional billing: Billing across multiple usage dimensions (tokens, API calls, compute time, or resolved outcomes) simultaneously is something conventional tools were not built for.
This is the exact problem Flexprice was built to solve
Flexprice is an open-source, developer-friendly, real-time usage-based billing infrastructure that handles all these complexities out of the box so your team can focus on building your product, not your billing system.
Why are AI companies moving away from seat-based pricing to usage-based pricing?
What is outcome-based pricing in AI, and which companies use it?
What is the biggest challenge enterprise buyers have with usage-based pricing?
How long does it take to build usage-based billing infrastructure from scratch?
What pricing models are AI SaaS companies using in 2026?
