Contents
Fast answer
Zendesk publishes the unit it bills for and no price for it. Since 18 May 2026 an interaction handled by a Zendesk AI agent lands in one of three resolution tiers. An assisted escalation is where the AI agent contributed and a human completed the resolution. A contained resolution is where the AI agent handled the interaction to completion without the customer requesting further assistance. A verified resolution is where the AI agent successfully resolved the interaction. Only verified resolutions count against your resolution allowance; Zendesk states that assisted escalations and contained resolutions do not. What Zendesk does not publish, anywhere we could find on 1 September 2026, is a dollar figure per resolution. Its own migration documentation converts existing balances 'based on a fixed pricing model' and then tells you the prices shown in its examples are placeholders and to contact Zendesk Sales. Every per-resolution rate circulating in blog posts is third-party, most of it written by companies that sell against Zendesk, and most of it still describes the pre-May-2026 model.
Facts last verified 2026-09
Quick facts
- Resolution tiers introduced
- 18 May 2026 (checked 2026-09-01)
- Assisted escalation
- AI contributed, human completed the resolution. Does not count against allowance
- Contained resolution
- AI handled to completion without further customer request. Does not count against allowance
- Verified resolution
- AI successfully resolved the interaction. The only billable tier
- Price per verified resolution
- Not published by Zendesk
- Balance migration
- Existing balances converted 'based on a fixed pricing model'
- Zendesk's own guidance
- Prices in migration examples are placeholders; contact Zendesk Sales
- Third-party rates
- Circulating figures are not sourced to Zendesk and mostly predate May 2026
On this page
What this page covers
A comparison visitor should understand the tradeoff, the best-fit scenario, and the next diagnostic tool to confirm the choice.
- 01What Zendesk actually publishes
- 02The three resolution tiers
- 03Why no price is published
- 04Forecast in resolutions, not dollars
- 05What to ask your rep
- 06When to look elsewhere
Why does this matter now?
This is a rare case where the vendor has been more precise about the mechanism than about the money, and buyers keep reading it the other way round. If you budget against a rate you found in a competitor's blog post, you are budgeting against a number nobody at Zendesk ever quoted, applied to a billing model that changed in May 2026. The honest forecast is a resolution count, which you can compute accurately, multiplied by a rate only your rep can give you. Getting that order right is the difference between a defensible budget and a renewal argument.
Internal path
Where to go next from this page
These links are part of the A8gent learning and conversion path. Use them to move from concept, to diagnosis, to workflow build, to course.
What you should be able to do after this
- Understand the three resolution tiers introduced in May 2026
- See which outcomes are billable and which are free
- Know why every dollar figure you have read is third-party
- Take the right questions into a Zendesk pricing call
How do you do it, step by step?
1. Learn the three tiers before you count anything
Zendesk's own documentation, updated after the 18 May 2026 change, defines assisted escalation as the AI agent contributing before a human completes the resolution, contained resolution as the AI handling the interaction to completion without the customer asking for more help, and verified resolution as the AI successfully resolving the interaction. These are outcome classifications applied after the fact, not plan features you switch on.
2. Count only what is billable
Zendesk states that only verified resolutions are charged against your resolution allowance, and that assisted escalations and contained resolutions do not count against it. That means your billable volume is a subset, not your total AI-handled conversation count. Any forecast built on total conversations overstates consumption, possibly by a lot, depending on how often your agent escalates.
3. Pull your own tier split from your instance
The only reliable input to a forecast is your own historical mix across the three tiers. Run a representative month and break out how many interactions the AI agent finished cleanly versus how many it handed to a human. That ratio is specific to your content quality, your intent coverage and your escalation rules, and it is the single largest driver of your bill.
4. Do not import a dollar rate from anywhere but Zendesk
There is no published per-resolution price. Zendesk's migration article states explicitly that the prices shown in its examples are placeholders only and directs customers to Zendesk Sales. Any figure you have seen with a dollar sign next to a Zendesk resolution came from somebody other than Zendesk, and frequently from a vendor whose product competes with it.
5. Take the count to your rep and get the rate in writing
Arrive with a verified-resolution count and a tier split rather than a question about pricing in general. Ask for the rate per verified resolution, the allowance included in your plan, the overage rate above it, and whether the rate changes at contracted volume tiers. Those four numbers plus your count are the whole budget.
The unit is documented, the price is not, and that is the whole story
Most pricing research starts from a number and works backwards to what it buys. Zendesk AI agents invert that. The company has published, in its own support documentation, an unusually precise definition of the outcome it charges for, and no dollar figure attached to it anywhere we could find on 1 September 2026. You can determine with confidence whether a given customer conversation will be billable. You cannot determine what it will cost. Both halves of that sentence are true at once, and a buyer who only reads the first half walks into a renewal with a resolution count and no rate.
The change that made this precise landed on 18 May 2026, when Zendesk introduced automated resolution tiers. Before that date the model was simpler and coarser, which matters enormously when you read anything written about Zendesk AI pricing, because a very large share of what is publicly available describes the previous arrangement and is now simply wrong. Publication dates are not a nicety on this topic. They are the difference between a correct description of the billing model and a description of a model that no longer exists.
What Zendesk publishes instead of a price is a mechanism, and it is worth reading the mechanism on its own terms. The documentation classifies each AI-handled interaction into one of three outcomes and then tells you which of the three touches your allowance. That is a genuinely customer-favourable design, because it means the vendor is only charging for the outcomes where the AI actually finished the job. It is also, for a buyer, an invitation to forecast on a number you can measure rather than one you have to guess.
So the useful posture here is not frustration at the missing rate. It is separation of the two questions. Volume is yours to compute, from your own instance, with high accuracy. Rate is your rep's to supply, in writing, with an overage clause attached. Anyone who hands you both numbers in a single blog post is inventing one of them, and it is not the volume.
What actually counts against your allowance
Zendesk defines three tiers. An assisted escalation is where the AI agent contributed to the interaction before a human agent completed the resolution. A contained resolution is where the AI agent handled the interaction to completion without the customer requesting further assistance. A verified resolution is where the AI agent successfully resolved the interaction. The first two are descriptions of AI involvement. The third is a description of AI success, established after the fact rather than asserted at the start of the conversation.
Only verified resolutions count against the resolution allowance. Zendesk states that assisted escalations and contained resolutions do not count against it. Read that carefully, because it has a large and counter-intuitive consequence for forecasting. A conversation where your AI agent did most of the work, answered three questions competently, and then handed a fourth to a human is not billable. Neither is a conversation the AI carried to the end where the outcome does not clear the verification bar. Your billable volume is a subset of your AI-handled volume, and the size of that subset is a property of your content, your intent coverage and your escalation rules rather than of the price list.
This creates a forecasting failure mode worth naming explicitly, because it runs in the expensive direction. Teams pull total AI-handled conversations from a dashboard, multiply by a rate they found somewhere, and produce a number substantially larger than their actual exposure. They then either reject the product on a fictional budget or negotiate against an inflated volume. The correct input is the verified count specifically, taken from your own instance across a representative month that includes at least one busy period.
The other consequence is a product one rather than a commercial one. Because escalation and containment are free, the incentive structure points at letting the agent try. There is no per-attempt charge to ration, which means the interesting question stops being how often to invoke the AI and becomes how good your content is at converting attempts into verified resolutions. That ratio is the metric to instrument first, before you have a contract, because it is the number that will determine both your bill and whether the deployment was worth doing.
How to price it anyway, and what to ask
Zendesk's migration documentation is the closest thing to a public statement about money, and what it says is instructive. Existing automated resolution balances are converted to the new tiers and plan defaults based on a fixed pricing model, and then the article states that the prices shown in its examples are placeholders only, not the actual prices, and directs customers to contact Zendesk Sales for pricing details. That is about as clear a confirmation as a vendor gives that the number is deliberately not public. It is not an oversight you can research your way around.
Which means the third-party figures deserve a specific warning rather than a general one. The search results for Zendesk AI pricing are dominated by posts from companies that sell products competing with Zendesk, several of them publishing repeatedly on the same keyword. Those posts carry dollar figures. Those figures are not sourced to Zendesk, most were written before the May 2026 tier change, and reproducing them would mean publishing a competitor's estimate of a rival's unpublished price as if it were fact. We are not doing that, and neither should your budget.
The practical route is to arrive at the pricing call with volume in hand and four questions. What is the rate per verified resolution on the plan we are considering. How many verified resolutions are included in that plan. What is the overage rate once we exceed it. Does the rate step down at contracted volume, and at what thresholds. Those four answers plus your own verified count are a complete budget, and every one of them has to come from Zendesk.
It is worth comparing this against vendors in the same category who do publish, because the contrast tells you what you are trading. Intercom Fin and HubSpot Breeze both publish per-resolution pricing you can read without a sales call, which makes them straightforward to model and straightforward to compare. Zendesk's model may well be better value for your specific mix, particularly given that escalations and containments are free, but you cannot know that until the rate exists in writing. Do not let the elegance of the tier design substitute for the number.
At a glance
Zendesk AI agent resolution tiers, read from Zendesk support documentation on 1 September 2026
| Item | What Zendesk publishes | Counts against allowance |
|---|---|---|
| Assisted escalation | AI agent contributed to the interaction before a human agent completed the resolution | No |
| Contained resolution | AI agent handled the interaction to completion without the customer requesting further assistance | No |
| Verified resolution | AI agent successfully resolved the interaction | Yes |
| Date introduced | 18 May 2026 | n/a |
| Price per verified resolution | Not published | n/a |
| Resolution allowance per plan | Not published | n/a |
| Overage rate | Not published | n/a |
| Migration of existing balance | Converted to new tiers and plan defaults based on a fixed pricing model; contact Zendesk Sales for details | n/a |
What mistakes should you avoid?
- Budgeting against a per-resolution dollar rate that no Zendesk document publishes
- Reading a blog post written before 18 May 2026 as if it described the current model
- Counting every AI-handled conversation as billable when only verified resolutions are
- Assuming a contained resolution is billed because it sounds like a successful outcome
- Treating a competitor's comparison table as a source for Zendesk's own pricing
- Signing without an agreed overage rate above the included resolution allowance
FAQ
- How much does a Zendesk AI agent resolution cost?
- Zendesk does not publish a price. As of 1 September 2026 we could find no per-resolution dollar figure in Zendesk's own pricing or support documentation, and its migration article states that the prices in its examples are placeholders only and directs customers to Zendesk Sales. Any specific rate you have read came from a third party rather than from Zendesk.
- What are the three Zendesk resolution tiers?
- Introduced on 18 May 2026. Assisted escalation: the AI agent contributed to the interaction before a human agent completed the resolution. Contained resolution: the AI agent handled the interaction to completion without the customer requesting further assistance. Verified resolution: the AI agent successfully resolved the interaction.
- Which Zendesk resolution tiers are billable?
- Only verified resolutions count against your resolution allowance. Zendesk states that assisted escalations and contained resolutions do not count against it. Your billable volume is therefore smaller than your AI-handled conversation volume, and the gap depends on your own escalation and verification rates.
- Why do other sites quote a per-resolution price for Zendesk?
- Because a comparison post needs a number in the headline. The figures circulating are third-party estimates, most of them published by companies selling an alternative to Zendesk, and most of them describe the billing model that existed before 18 May 2026. We do not reproduce them, because we cannot source them to Zendesk.
- What should I ask Zendesk Sales about AI agent pricing?
- Four things: the dollar rate per verified resolution, the resolution allowance included in your plan, the overage rate once that allowance is exhausted, and whether the rate steps down at contracted volume. Bring your own verified-resolution count so the conversation is about your workload rather than a list price.
Sources & further reading
Done for you
Want this built for your business?
Tell us the workflow. We scope, build, and ship the agent with guardrails and the numbers to prove it worked. The scoping call is free.
Free scoping call. You own the code.
Was this page helpful?

