Summary
- Zendesk calls the automated resolution "the unit of measurement used for calculating and billing your account for AI agent usage", and its price list attaches $1.50 committed and $2.00 pay-as-you-go to that unit.
- A second help-centre article describes the resolution allowance as "a flexible currency pool" of $2, $5 or $10 per agent seat per month, capped at $5,000 a year, and prints no price per resolution anywhere in it.
- Since 18 May 2026 the reported AR% counts contained plus verified resolutions, while the same announcement says contained resolutions do not consume automated resolutions — so the rate a customer watches and the class a customer pays for are two quantities sharing one name.
The sentence is short and it is printed twice. In the AI agents glossary: "Automated resolution | The unit of measurement used for calculating and billing your account for AI agent usage." In the pricing page's own FAQ, the same claim with a capital R: "Automated Resolutions: Are the unit of measurement used for calculating and billing your account for AI agent usage." The main Zendesk glossary uses the plural and adds the trigger: "Paying per automated resolution means that you pay only for customer requests that were successfully resolved by the AI agent, without any escalation to a human agent." The product overview puts it shorter still: "Usage of AI agent functionality is measured through automated resolutions."
Four pages, one name, one payment trigger. What follows is not a naming problem. It is an instrument problem: three documents act on that unit, and they do not act on it in the same unit of account.
The price list does its job
The public pricing page is the only one of the three that prints a currency amount beside the word resolution. Under a row labelled "Included automated resolutions (up to 10k per year)", the retrieved columns read "5 ARs per agent/month" for Support Team, "5 ARs per agent/month" for Suite Team and "10 ARs per agent/month" for Suite Professional. Two rows lower: "Committed automated resolutions" at "$1.50" and "Pay-as-you-go resolutions" at "$2.00", the same figure across those three columns. Seat prices sit in the same table — $19, $55 and $115 per agent per month on annual terms, $25, $69 and $149 monthly — and the surrounding FAQ says pricing "is primarily seat-based (per agent, per month)". All of it is on the price list.
A reader holding that page can price a resolution above the included count. A reader holding that page cannot price an Enterprise resolution, because the extracted comparison returned only those three plan columns; I carry that limit rather than paper over it. The page also prints a ceiling in the same breath as a count: "up to 10k per year".
The same allowance, denominated in dollars
The second instrument prints no rate at all. "Your resolution allowance is a flexible currency pool used to pay for the cost of automated resolutions," the allowance article says, and "You can consume the balance freely across different resolution tiers without pre-selecting a fixed mix of tiers in advance." Then the defaults, per agent seat per month: Suite Team $2, Suite Growth $5, Suite Professional $5, Suite Enterprise $10, Suite Enterprise Plus $10; Support Team $2, Support Professional $2, Support Enterprise $2, with Growth and Enterprise Plus marked "Not applicable" on the Support side. And the ceiling: "The maximum allowance per year for all plans is $5000."
Two worked examples fix the pool's arithmetic — 20 Suite Enterprise seats give $200 a month, 20 Support Enterprise seats give $40 (the same article). Monthly allowances do not carry over. Overage is billed monthly "regardless of your subscription terms", and pay-as-you-go resolutions are charged once the allowance limit is reached.
What the article never does is convert. It prints no dollar figure per verified resolution, so the $5,000 maximum cannot be turned into a count from this set of pages. Its one comparative sentence about cost is qualitative: "If you increase your resolution allowance before overages occur, it provides a better per-resolution cost than overage billing." Better by how much, at what multiple, against which tier — the sentence does not say, and no table supplies it.
Zendesk prints the sentence that dissolves the contradiction
Read the price list against the pool and the temptation is to call one of them wrong. The navigation article forecloses that: "Some accounts have the pricing model for traditional automated resolutions, some accounts have a pricing model for automated resolution tiers managed by a resolution allowance, and some accounts have a pricing model for outcome-based pricing (OBP)." It then tells the reader how to discover which set of articles applies — the automated resolutions dashboard, or the resolution usage and resolution allowance dashboard — and adds a timing rule for anyone moving between them: "If your account is upgrading from a Suite plan with automated resolutions to a new Suite plan with a resolution allowance, Zendesk will give you 30 days notice before your account is upgraded."
That paragraph is why this is a documentary briefing rather than an accusation. The bills do not reconcile across models because they are not the same bill. Three pricing models, three instrument sets, one account apiece.
The billed event is a judgement made after the event
The unit is not a ticket attribute. It is a classification, and there are three of them. An assisted escalation, where "The AI agent contributed to the interaction before a human agent completed the resolution"; a contained resolution, where "The AI agent handled the interaction to completion without the customer requesting further assistance"; and a verified resolution, where the agent "successfully resolved the interaction" and the conversation passed review. Only the third is charged: "You're charged against your resolution allowance only for verified resolutions—requests the AI agent resolves and that pass conversation review. Contained resolutions and assisted escalations don't count against the allowance" (About automated resolution tiers).
The review is a model reading the transcript: "When the conversation ends, a verification process is performed by a large language model (LLM) that evaluates the text of the conversation to confirm that the customer's request was satisfactorily resolved." The ending has a printed clock: "Messaging or email: The conversation ends 72 hours after the first message or email in the conversation. Voice: The conversation ends immediately upon hangup" (the same article). It dates the whole arrangement too: "This article describes the automated resolution tiers that were introduced on May 18, 2026."
So the billable quantity has a trigger and a judge. None of the nine pages I read prints the judge's threshold, its model version, its confidence measure, or a route for a conversation that fails review.
The entrance ticket for the cheaper rate
$1.50 is the lower of the two printed rates, and its conditions are not on the page that prints it. The quantity that admits an account to committed pricing appears in an article titled as the platform before the tiers: "You can add to your automated resolution allotment by purchasing 100 or more automated resolutions ahead of time. Also called committed usage … Committed usage provides a better per-resolution cost than overage billing (or pay-as-you-go)." The same article fixes the counting grain — "Automated resolutions are counted per conversation rather than per user" — and the annual ceiling — "Accounts on all plans have a maximum of 10,000 allocated automated resolutions per year" — and it says of itself: "This article describes the automated resolution platform in place prior to May 18, 2026" (the prior-platform article).
So the pre-purchase threshold and the per-conversation grain travel with a pre-18-May scope attached, and the current price list restates neither. A buyer on the tier model reading $1.50 has to establish which era's counting rule their contract carries. The dashboard test is the only printed way to find out.
The number on the dashboard and the number on the invoice
The reporting article changes a ratio and a charge in the same breath. "You will continue to be billed only for Verified resolutions and Contained resolutions do not consume automated resolutions." Two paragraphs away: "Before May 18, 2026 Verified / (Unassisted conversations + Assisted Escalations + Contained + Verified) x 100" and "After May 18, 2026 (Contained + Verified) / (Unassisted conversations + Assisted Escalations + Contained + Verified) x 100". The article names the visible effect without hedging: "As a result, you may notice an increase in your AR%. This is because your AR% will now include both Contained and Verified resolutions." And a class left the invoice inside a document titled as reporting: "As a part of this reporting change, intelligent triage auto-replies will no longer be charged as automated resolutions" (the same announcement).
Set the sentences side by side. The numerator of the published rate contains a class the published charge excludes. The metric moved up; the billable class did not move with it. Both statements are Zendesk's and both can be true, but the consequence is unavoidable: AR% is not "share of conversations we paid for", and reading it that way overstates the paid share by exactly the contained class.
The arithmetic a stranger can still do — mine, not Zendesk's
Every division below is mine, uses printed numbers, and assumes the pool and the per-resolution rate describe one account. Zendesk prints that different accounts sit on different models and the allowance article prints no rate, so this is a cross-instrument restatement, not anybody's bill.
$5,000 / $2.00 = 2,500 resolutions, if the annual maximum were spent at pay-as-you-go. 10,000 / 2,500 = 4, so the pool at that rate covers one quarter of the count ceiling on the price list. The worked example divides the same way: $200 / $2.00 = 100 resolutions a month, against 20 seats x 10 ARs = 200 — a factor of two between two printed instruments for the same 20 seats. Inside the pool, the same plan name costs five times more on Suite than on Support at Enterprise: $10 / $2 = 5. The cap binds near 42 Suite Enterprise seats, since $5,000 / 12 = $416.67 a month and $416.67 / $10 = 41.67.
And the two printed rates differ by 25 percent: ($2.00 - $1.50) / $2.00 = 0.25.
Those are the only conversions the documents support, and each one requires putting a dollar figure from one model beside a rate from another.
Where the printed surface ends
Six limits, each attached to the claim it bounds. The pricing comparison surfaced three plan columns, so no Enterprise count or Enterprise rate is quoted here. No canonical support.zendesk.com copy of the article the glossary links to — "About automated resolutions for AI agents" — was retrievable, so the per-conversation rule is cited to the prior-platform article with that scope attached. No reviewed page prints the LLM's verification threshold, model version or review route. No reviewed page states how many accounts sit on each of the three models, so no share-of-customers statement is available.
No outcome-based pricing rate card, unit definition or invoice example was obtained; OBP is named and not described. And the pricing page text came back truncated in one pass, with its comparison table read from a separate extraction of the same URL.
Zendesk has printed a unit, a price list, a dollar pool, a judge with a clock, and a ratio. What it has not printed is a sentence that lets an account on one model recompute a line on another. So the reader's first question is not arithmetic. It is which of the three dashboards their own account shows — because that decides which document has authority over the invoice, and Zendesk, to its credit, has published the test. The company's BTW directory entry is where to follow what it prints next.
Member Briefing
Read the full trend briefing
Members read the full briefing with the source notes behind every finding.
Strategic Circle
A daily briefing in your language, built from changes we have verified.
Join for USD 20/monthCancel anytime.
Members also get the full research library, a daily briefing on the networks they follow and the private forum.






