Skip to content
Resolvara

A resolution definition you can sign

Per-resolution pricing is only as fair as its definition. This clause is built from the Strict Resolution Standard. Adjust it, then paste it into your order form, an MSA addendum or a renewal.

How long a conversation is watched before it can be billed. The Standard uses 7.

A score below this fails. The Standard uses 0.4, which is below a rating of 1 or 2 on a 1 to 5 scale.

The first matches the Standard, which counts silence as a pass but reports confirmed passes apart.

A human reply after the AI always counts as a failure. The Standard has no exception for courtesy replies, so this page does not offer one.

Definition of a Billable Resolution

This clause applies between the Customer and the Vendor. It defines which AI conversations the Vendor may charge for as resolutions, and what data the Customer receives to check that.

1. Definitions

1.1"AI Conversation" means a conversation between an end user and the Customer's support channel in which an AI agent provided by the Vendor took part.

1.2"Judgement Window" means the period of 7 days after the time at which an AI Conversation was created.

1.3"Fail Signal" means any one of the following:

(a)Vendor escalation: the Vendor's own recorded state for the conversation is escalated, handoff or negative feedback.

(b)Human reply: a human agent or administrator replied in the conversation after the AI agent took part, including a courtesy reply.

(c)Reopened: the conversation was reopened within the Judgement Window. Where the export gives no time for a reopen, a reopen count above zero counts as a reopen within the Judgement Window.

(d)Repeat contact: the same end user opened another conversation after this one was created and no later than the end of the Judgement Window.

(e)Negative rating: the end user gave a satisfaction score below 0.4 on a scale of 0 to 1 (a rating of 1 or 2 on a scale of 1 to 5), after converting the score to the scale of 0 to 1 by the scale of the rating.

1.4"Positive Evidence" means any one of the following:

(a)the Vendor's own recorded state for the conversation is "confirmed";

(b)the end user gave an explicit positive rating, meaning a satisfaction score of 0.75 or more on a scale of 0 to 1 (4 or 5 on a scale of 1 to 5);

(c)a System Receipt exists for the conversation.

1.5"System Receipt" means a record created by an action the AI agent took in that conversation, such as a refund identifier or an order update. A System Receipt counts as confirmation that the conversation had an effect, and is Positive Evidence.

1.6"Billable Resolution" means an AI Conversation that meets every condition in clause 2.

2. When a resolution becomes billable

2.1The Vendor may charge a fee for an AI Conversation only if it is a Billable Resolution.

2.2An AI Conversation becomes billable only after its Judgement Window has ended. The Vendor shall not include it in an invoice before then. A fee for a conversation still inside its Judgement Window is invoiced in a later period, once the window has ended and clause 2 is met.

2.3An AI Conversation that shows any Fail Signal is not a Billable Resolution, whatever the Vendor's own classification of it.

2.4An AI Conversation with no Fail Signal is a Billable Resolution even if the end user gave no confirmation. The Vendor shall report separately the number of Billable Resolutions that have Positive Evidence and the number that have none.

2.5If the Vendor does not supply the customer identifier for a conversation, the repeat contact test cannot be applied to it and it is not a Billable Resolution.

3. Reporting

3.1With each invoice, and in any case monthly, the Vendor shall give the Customer an export, in CSV or JSON, with one row for every conversation in the period. The export includes conversations that were not billed, and conversations in which no AI agent took part, marked as such.

3.2Each row shall contain at least:

(a)conversation identifier;

(b)customer identifier, hashed with the same key throughout, so that repeat contact can be tested without sharing personal data;

(c)time the conversation was created, with its time zone;

(d)whether an AI agent took part;

(e)the Vendor's recorded resolution state;

(f)whether the conversation was billed, and the fee charged;

(g)the time of the last AI message and the time of any human reply after it;

(h)the reopen count and the time of any reopen;

(i)the satisfaction score, with the scale it was given on;

(j)the reference of any System Receipt.

3.3The Vendor shall not remove or alter rows after delivery, except to correct an error, and shall tell the Customer of each correction.

4. Audit and reconciliation

4.1The Customer may check the export against this clause, by hand or with any tool, including by applying the Strict Resolution Standard version 0.2 with a Judgement Window of 7 days and a rating threshold of 0.4. The Customer is not required to send the export to any third party.

4.2Within 30 days of receiving an export, the Customer may tell the Vendor in writing which conversations it believes are not Billable Resolutions, giving the conversation identifier and the Fail Signal for each.

4.3The Vendor shall review each disputed conversation against this clause and answer in writing, before the next invoice is issued.

4.4Each disputed conversation that shows a Fail Signal, or that is not a Billable Resolution for another reason under clause 2, shall be credited in full on the next invoice.

5. Changes to the definition

5.1During the term, the Vendor shall not change the definition of a Billable Resolution, the Fail Signals, the Judgement Window or the fields in clause 3 without the Customer's written consent and at least 60 days' written notice.

6. Attribution

6.1Definitions adapted from The Strict Resolution Standard (CC BY 4.0), resolvara.com/standard.

This is not legal advice. Have counsel review any clause before you sign it. The Standard is a measurement, not a verdict, and your vendor will apply its own contract definition unless you agree otherwise.

Questions to ask your vendor before signing

  1. Is a customer who stops replying billed? After how long?

    6 of the 13 vendors on the definitions page are listed as billing assumed resolutions, where silence counts. The silence window differs: HubSpot Customer Agent: 72 hours after the visitor's last message; status is set after the window. Gorgias AI Agent: 72 hours without human involvement. Front Autopilot: 24 hours for chat, SMS, WhatsApp and custom channels; 72 hours for email and portal.

  2. Can we get a per-conversation export that shows which conversations were billed?

    Not every vendor documents one. Help Scout: Review billable conversations under Reports > Scout (view only). Export of those conversations is not stated (unconfirmed). Generic Reports export (CSV/XLSX) exists but is reported to exclude Scout; the Inbox API exposes conversation data.

  3. Is a conversation still billed if a human replies after the AI?

    Practice differs. HubSpot's published rule treats manual assignment as not a qualifying handoff, so a human can handle the thread and it still counts as resolved.

  4. If the customer comes back about the same problem, is the first conversation credited?

    The data usually exists. Intercom's documented fields include a reopen count (statistics.count_reopens).

  5. Are test and internal conversations left out of the billed count?

    Ask for the flag. Ada's documented export includes an is_test_user field.

  6. Can the billing definition change during our term, and who decides?

    Vendors do change units and prices. Zendesk bills per Verified resolution (since 2026-05-18), drawn from a dollar allowance. HubSpot: $0.50 per resolved conversation (50 credits), effective 2026-04-14.

  7. Can we check the count ourselves, and what happens to a conversation we dispute?

Facts come from each vendor's published documentation, with sources and check dates on the definitions page.