AI Customer Service Pricing Models Compared: Per-Resolution vs. Per-Seat vs. Per-Conversation

A framework for evaluating the five pricing models used by AI customer service platforms, with cost benchmarks.

AI customer service platforms use five distinct pricing models: per-seat, per-conversation, per-resolution, per-session, and platform fee plus usage. Each model changes how costs scale, how predictable budgets are, and whether you pay for effort or outcomes.

This guide breaks down every model, compares real vendor pricing, and provides a framework to calculate your total cost of ownership.

The Five AI Customer Service Pricing Models

Every AI customer service platform falls into one of five pricing categories. The model you choose directly affects cost predictability, alignment between spend and value, and your ability to forecast budgets as automation scales.

Per-Seat Pricing

Per-seat pricing charges a fixed monthly fee for each human agent using the platform. This is the legacy model inherited from traditional helpdesk software. It scales with team size, not with the volume of customer interactions your AI handles.

Advantages: predictable monthly costs, simple to budget, familiar to finance teams.

Limitations: costs stay flat even as AI handles more volume and agents handle less. You pay the same whether your team resolves 100 or 10,000 conversations. This model rewards adding headcount, which works against the efficiency gains AI provides.

Vendors using per-seat pricing: Zendesk Suite plans start at $55 to $169 per agent per month. Freshdesk ranges from $15 to $79 per agent per month. HubSpot Service Hub starts at $25 per user per month.

Per-Conversation Pricing

Per-conversation pricing charges for each conversation the AI agent handles, regardless of whether the issue is resolved. A conversation typically starts when the AI engages and ends after a defined inactivity window (often 24 hours) or when the customer closes the interaction.

Advantages: costs scale directly with usage volume.

Limitations: you pay even when the AI fails to resolve the issue. A conversation that ends with the customer abandoning in frustration costs the same as one that solves their problem. This creates a misalignment between what you pay for and the value you receive.

Salesforce Agentforce launched with $2 per conversation pricing. Salesforce has since introduced an alternative Flex Credits model at $0.10 per action due to customer pushback on conversation-based billing, though the $2 per conversation option remains available.

Per-Resolution Pricing

Per-resolution pricing charges only when the AI agent successfully resolves a customer issue without human intervention. You pay for outcomes, not activity.

Advantages: direct alignment between cost and value delivered. If the AI does not resolve the issue, you are not charged. This creates strong incentive alignment between the vendor and the customer.

Limitations: total monthly costs fluctuate with resolution volume, which requires monitoring. As AI performance improves and resolves more conversations, spend increases in absolute terms (though cost per interaction drops relative to human agent alternatives).

The critical question with per-resolution pricing is how "resolution" is defined. Some vendors define a resolution as any conversation where the customer does not request a human agent. Others use an LLM verification step to confirm the issue was addressed. Fin counts only genuine, positive resolutions where the customer's issue was actually solved, which produces a more honest metric.

Fin is priced at $0.99 per resolution. Zendesk charges $1.50 per automated resolution (committed volume) or $2.00 pay-as-you-go for overages. Gorgias charges approximately $0.60 to $1.27 per automated resolution depending on tier.

Per-Session Pricing

Per-session pricing charges for each session the AI agent initiates, regardless of how many messages are exchanged or whether the query is resolved. A session typically covers one customer interaction window.

Advantages: simple to understand and track.

Limitations: like per-conversation pricing, you pay for engagement rather than outcomes. Multiple sessions may be needed to resolve a single issue, inflating the true cost per resolution.

Freshworks' Freddy AI Agent uses per-session pricing at $100 per 1,000 sessions ($0.10 per session). Sessions are charged regardless of resolution.

Platform Fee Plus Usage

Some vendors combine a recurring platform fee (monthly or annual) with usage-based charges. The platform fee covers access to the software, infrastructure, and base capabilities. Usage charges apply for AI interactions on top.

Advantages: the platform fee creates cost predictability for base access, while usage charges scale with volume.

Limitations: the combined model can obscure true costs. Teams often underestimate usage charges during initial budgeting. Platform fees may cover features you do not use, and bundled pricing makes it difficult to compare the AI agent component against standalone alternatives.

Salesforce Agentforce requires a base Service Cloud or Sales Cloud subscription plus either $2 per conversation or Flex Credits. Zendesk requires a Suite subscription ($55 to $169 per agent per month) plus per-resolution charges for AI agent activity.

Pricing Comparison: Major AI Customer Service Platforms

How to Calculate Total Cost of Ownership

Comparing AI customer service platforms on sticker price alone is misleading. A $0.99 per-resolution price and a $2.00 per-conversation price are measuring different things. Calculating total cost of ownership (TCO) requires accounting for four cost layers:

Layer 1: Platform subscription. Some AI agents require a separate helpdesk platform. If the AI agent does not include a helpdesk, inbox, knowledge base, and reporting tools, add those costs. A platform like Zendesk or Salesforce Service Cloud can add $55 to $300+ per agent per month before any AI charges.

Layer 2: AI agent usage charges. Calculate your expected monthly volume. Multiply by the per-unit cost (resolution, conversation, session, or action). Factor in whether you pay for resolved interactions only or all interactions.

Layer 3: Add-on costs. Many platforms charge separately for copilot tools, QA modules, workforce management, and advanced analytics. Zendesk charges $50 per agent per month for its Copilot add-on and $35 per agent per month for QA. These costs compound quickly for large teams.

Layer 4: Implementation and ongoing management. Some platforms require engineering resources and professional services for setup. Enterprise AI agent deployments can take 3 to 6 months with vendors like Sierra, which uses a TypeScript-based SDK requiring developer involvement. Self-managed platforms that non-technical teams can configure reduce this cost layer significantly.

TCO Example: 50-Agent Team, 20,000 Monthly AI Resolutions

This example illustrates how per-resolution pricing without a separate platform subscription can result in substantially lower TCO. The exact numbers will vary based on your specific vendor negotiations, team size, and volume, but the structural cost differences between models persist at every scale.

What Makes Per-Resolution Pricing Different

Per-resolution pricing stands apart because it ties cost to the outcome the customer and the business actually care about: the issue being resolved. Three characteristics distinguish genuine per-resolution pricing from other usage models:

You do not pay for failed interactions. If the AI cannot resolve the issue and the conversation escalates to a human, no charge applies. This contrasts with per-conversation and per-session models where every interaction incurs cost.

Resolution definitions matter. Ask any vendor using per-resolution pricing: how do you define a resolution? Is it the absence of escalation? A customer inactivity timeout? Or a verified, positive outcome? The definition directly affects what you are actually paying for.

Costs decrease per interaction as AI improves. When an AI agent achieves a higher resolution rate, more conversations are resolved without human involvement. The cost per customer interaction drops even as the absolute AI spend increases, because you are replacing more expensive human agent time.

Five Questions to Ask Before Choosing a Pricing Model

1. How is a "resolution" or "conversation" defined in your billing? Ask for the specific technical definition. A resolution that counts after five minutes of customer inactivity is a different product than one verified by an AI quality check.
2. What is included in the base price? Does the AI agent price include a helpdesk, knowledge base, reporting, and copilot tools? Or are those separate subscriptions?
3. What engineering resources are needed? Can your CX team configure and optimize the AI agent, or does setup require developers, professional services, or the vendor's own engineering staff?
4. What happens when I exceed my committed volume? Overage rates can be significantly higher than committed pricing. Zendesk charges $2.00 per resolution for overages versus $1.50 for committed volume.
5. Do I retain ownership of my data and training? With managed service models or vendor-dependent configurations, switching providers may mean losing your AI training, knowledge management work, and conversation history.

Why Fin Uses Per-Resolution Pricing at $0.99

Fin is priced at $0.99 per resolution because outcome-based pricing creates the strongest alignment between what businesses pay and the value they receive. When Fin resolves a customer conversation end-to-end, the business pays $0.99. When Fin cannot resolve the conversation, the business pays nothing.

This model works because Fin is built to resolve, not deflect. Fin's average resolution rate across 7,000+ customers is 67%, improving roughly 1% every month.

Ecommerce brands regularly achieve 70% to 84% resolution rates. The Fin AI Engine, a proprietary architecture with purpose-built retrieval and reranking models, is specifically engineered to maximize genuine resolutions.

Fin's per-resolution pricing includes capabilities that other platforms charge extra for:

A complete AI agent system, not a point solution. Fin includes the AI agent, procedures for multi-step workflows, omnichannel deployment across chat, email, voice, social, Slack, and SMS, and 45+ language support. There is no separate platform fee required to run Fin.

A native helpdesk option. Fin is the only AI agent that includes an optional native helpdesk for seamless AI-to-human handoff. Businesses can also connect Fin to existing helpdesks including Zendesk, Salesforce, and others through native integrations.

Self-managed configuration. Teams can train, test, deploy, and optimize Fin without engineering resources. The Fin Flywheel (Train, Test, Deploy, Analyze) gives CX teams direct control over AI performance improvement.

The Fin Million Dollar Guarantee. Fin backs its performance with the Fin Million Dollar Guarantee: new customers who are not satisfied within 90 days receive a full refund of Fin spend up to $1,000,000. For high-volume enterprises, Fin guarantees a 65% resolution rate or pays $1,000,000.

"We're in the millions of dollars of cost savings from leveraging Fin." - Simon Millichip, SVP Customer & Risk Operations, ZayZoon

"Fin fundamentally changed our support strategy. It helped us scale instantly, resolve over 50% of conversations, and save more than 1,700 hours in the first month." - Isabel Larrow, Product Support Operations, Anthropic

Use the Fin ROI Calculator to estimate your expected savings based on your current support volume and costs.

Frequently Asked Questions

How much does AI customer service cost per month?

Monthly costs vary widely based on the pricing model, team size, and conversation volume. A mid-size team handling 20,000 AI resolutions per month could pay approximately $19,800 with Fin ($0.99/resolution), $30,000+ with Zendesk ($1.50/resolution plus per-seat platform fees), or $40,000+ with Salesforce Agentforce ($2/conversation plus Service Cloud licensing). The total depends on which platform components are included versus charged separately.

What is per-resolution pricing for AI agents?

Per-resolution pricing charges businesses only when the AI agent successfully resolves a customer issue without human intervention. This model ties cost to outcomes rather than activity. Fin uses per-resolution pricing at $0.99 per resolution, counting only genuine positive resolutions. Other vendors may define resolution differently, so it is important to ask how each vendor measures and verifies successful resolution.

How do AI agent pricing models compare?

The five main models are: per-seat (fixed cost per agent, regardless of AI volume), per-conversation (charged per AI interaction regardless of outcome), per-resolution (charged only for successful resolutions), per-session (charged per session window), and platform fee plus usage (base subscription plus variable AI charges). Per-resolution pricing provides the strongest alignment between cost and value because you only pay when the AI solves the problem.

Is per-resolution pricing unpredictable?

Per-resolution costs scale with resolution volume, which means monthly spend changes as AI performance improves. However, this is a positive cost dynamic: higher resolution rates mean the AI is handling more work that would otherwise require human agents. Businesses can forecast costs using their conversation volume and expected resolution rate. With Fin, usage controls and spending limits are built into the platform. The alternative, per-seat pricing that stays flat regardless of outcomes, is predictable but disconnected from value.

What hidden costs should I watch for?

Look for: separate helpdesk subscriptions required to run the AI agent, copilot and QA add-ons charged per agent per month, professional services and implementation fees, overage rates for exceeding committed volume, and engineering resources needed for setup and configuration. A $0.99 per-resolution price that includes the helpdesk and copilot tools is a different value proposition than a $1.50 per-resolution price that requires a separate $115/agent/month platform subscription plus $50/agent/month for a copilot.