Zendesk Outcome-Based Pricing and Automated Resolutions: what does outcome-based AI-agent pricing mean for the buyer?
Published
Read time
7 minutes
Article summary
Zendesk uses an outcome-based pricing model for its AI agents, based on successfully resolved customer contacts (Automated Resolutions). While the model isn't entirely new, all of Zendesk's AI agents now run under this single unified model, since the previously separate advanced AI features were opened to all Suite and Support tiers. This article explains how an AI-resolved contact is measured and verified, and how to manage consumption directly from the Zendesk Admin Center.
The AI-agent pricing model: the shift to unified outcome-based pricing
For its AI agents, Zendesk has moved entirely to an outcome-based pricing model, where you pay only for successfully resolved customer interactions (_Automated Resolutions_).
The model itself isn't entirely new, but its significance is now more pronounced than ever. In May 2026, Zendesk opened its previously separate advanced AI-agent features (_AI agents – Advanced_) to all its Suite and Support customers, while the old base tier (_Essential_) moved into the legacy category.
In practice this means that all of Zendesk's AI agents now run fully under this single, unified outcome-based model.
What is an Automated Resolution (AR) and how is it measured?
An _Automated Resolution_ is the unit of measurement by which an AI agent's consumption and value are measured. In line with Zendesk's latest policies, the old flat-rate model has been replaced by four new measurement categories.
You pay only for interactions where the AI agent genuinely resolves the customer's problem. The categories split into free and paid as follows:
Free categories:
Unassisted conversation (free): No automations or back-end processes were performed in the conversation. This category covers situations where:
It's small talk or a welcome message: The conversation contained only chit-chat or purely automatic system responses.
The AI agent didn't understand the request: If the bot doesn't recognise the question and can't provide an answer.
It's spam: Empty openers or spam are automatically filtered out.
It's an internal action: Internal comments or notes made by the AI agent don't consume resolutions.
It's testing: Testing the AI agent in Sandbox environments is always free.
Assisted escalation (free): The AI agent took part in the conversation and did some of the work, but the matter was handed to a human agent, who completed the resolution. As soon as a conversation or ticket is handed to a human at any stage, it becomes free.
Paid resolutions:
Contained Resolution (paid): The AI agent handled the customer's request from start to finish without human intervention. The resolution is interpreted as successful when the customer hasn't reacted or returned to the matter within a channel-specific time limit (by default 72 hours for inactive conversations and 2 days for conversations already resolved).
Verified Resolution (paid): The AI agent resolved the request, and an independent language model working in the background (LLM Verification) independently analyses and audits the entire conversation to confirm that the customer's problem was genuinely and properly resolved. This independent quality check ensures that incorrect or incomplete bot answers aren't billed as successful resolutions, even if the customer didn't manually confirm the resolution.
How do automated resolutions work across different channels?
The AI agent's capabilities and how it's measured depend on the channel:
1. Messengers and chat channels (Messaging, WhatsApp, etc.)
A conversation is considered resolved (Contained or Verified) if:
The status is Handled: The AI agent took responsibility for the conversation and carried it to the end.
No escalation: If the conversation is handed to a human at any stage, it becomes an _Assisted Escalation_ (free).
Session time limit: If the bot offers a resolution and the customer leaves the conversation without requesting an escalation, the assessment happens when the session closes. The inactive-session time limit is by default 3 days from the first passive moment, but it's adjustable in the Admin settings from 2 hours up to 72 hours. If, on the other hand, the AI agent recognises the conversation as already resolved, it closes the session after 2 days.
2. Email and web forms (Agentic Email)
With the 2026 updates, the email channel works fully agentically (Agentic Email) without rigid pre-set paths. The AI agent reads a free-form email, understands the context, fetches information and can perform API integrations (e.g. order changes). A resolution is billed if:
The AI agent's reply: The AI agent has sent the customer a reply (a generative email reply) or performed an automated process in the back-end system and sent confirmation of it.
No human intervention: No human agent replies to or makes changes to the ticket within 3 days of the AI agent's reply.
Monitoring and managing consumption from the Admin Center
Zendesk offers excellent tools for tracking AI-agent usage and preventing surprise cost spikes directly from the admin panel (_Account > Usage > Automated resolutions_).
1. Overage management (Overage Settings)
One of the things that worries decision-makers most is costs getting out of hand. In the Admin Center, a company can choose how the system reacts when the monthly automated-resolutions limit is reached:
Maintain functionality and allow overage (default): The AI agent continues working normally after the limit is exceeded, and the extra resolutions are billed at an overage rate. This ensures uninterrupted customer service.
Pause functionality and don't allow overage: The AI agent's operation across all channels automatically pauses as soon as the limit is reached. All new contacts are then routed directly to human agents, and no extra costs arise. The system sends automatic warnings by email when 80% and 100% of the usage limit is reached.
2. Analysing and forecasting usage data (Usage Details)
The Admin Center dashboard offers a real-time view of consumption:
Current view: Shows accumulated resolutions broken down by type.
Forecast view: Based on the current daily average, the AI calculates an estimate of future consumption up to the next renewal date, which helps with budgeting.
Contributing automated resolutions: A log showing the ID number, type, brand and exact timestamp of every individual billed resolution. This guarantees 100% transparency and enables auditing of individual events.
3. Visual reporting: Explore and Sankey charts (new)
The reporting section of the Advanced AI dashboard includes new Sankey-based conversation journey reports. These charts visualise the entire customer flow very clearly:
You see at a glance where customers drop out of the chatbot or email thread.
They break down successful, assisted and unsuccessful resolutions.
They show the exact points where conversations end up escalating to the human-agent queue, which helps with the bot's ongoing development.
How do you calculate the ROI of your AI investment?
Because outcome-based pricing ties costs to genuine resolutions, making ROI calculations becomes extremely transparent. You can compare the price of an AI resolution directly to the cost of a ticket handled by a human agent.
Because the price of an AI resolution is a fraction of the cost of human work, every resolution the AI agent makes directly saves the company's operating costs and frees up agents' time for more complex cases.
Calculate the ROI of the investment here: https://www.zendesk.com/service/ai/ai-agents/#ai-roi-calculator
How do you achieve a high resolution rate?
AI's outcome-based pricing makes experimenting financially risk-free, but it places high demands on the implementation. If the AI agent isn't configured correctly, it can't resolve customers' requests, and the allocated free resolutions go to waste while customers end up with humans.
An AI agent needs two key factors to work effectively:
A knowledge base optimised for AI (Help Center): Generative AI needs clear, structured and up-to-date articles to answer questions reliably. (Read tips here on how to create an AI-optimised Help Center.)
Functional integrations (API connections): The automation rate is raised high by connecting the AI agent to back-end systems. The AI agent can then independently perform actions on the customer's behalf, such as order cancellations or data lookups, which lead to genuine, approved resolutions.
Get in touch and let's spar together on your customer service's AI readiness and design a customer service system that works for you into the future.
