# Research: how AI support vendors define and bill a resolution

> What 14 AI customer-support vendors' own documents say a resolution is, when it is decided, what voids it and how it is billed, read on 8 November 2026.

Published 2026-11-12 · Reviewed 2026-11-12. Canonical: https://attify.app/research/support-resolution-definitions

Of 14 AI customer-support vendors whose own documents we read on 8 November 2026, the six that define a resolution count a conversation the customer did not escalate, and where they state a window it is 72 hours or less. They bill on six different units (resolution, automated interaction, session, action, conversation, and outcomes set per contract), and three publish a price per resolved outcome: Intercom, Zendesk and Help Scout. Gorgias prices its automated interactions by plan, but the page read showed conflicting figures.

Every fact below comes from the vendor's own public documentation; a date in a source link is the date the page itself shows. Figures that appeared only in third-party writing are left out, and the page makes no claim about any vendor's results.

## Definitions

Read on 8 November 2026.

| Vendor | What counts | When it is decided | What voids it | Customer asked? | Source |
|---|---|---|---|---|---|
| Intercom (Fin) | Two kinds of resolution: confirmed (the customer replies that the answer helped) and assumed (the customer leaves after Fin's last answer without asking for more help). A reply to a greeting alone does not count | No window stated | A request for a human or more help, or a failed Procedure. A resolved conversation the customer returns to is deducted, even in a later billing period. A teammate joining after Fin answered can still leave it resolved | Implicitly; confirmed and assumed count alike | [Intercom, updated 23 Sep 2026](https://www.intercom.com/help/en/articles/8205718-fin-ai-agent-outcomes) |
| Zendesk (AI agents) | Since 18 May 2026, three tiers. Zendesk's "Verified resolution": the AI answered, the customer did not ask for further help, and Zendesk's LLM check after the conversation ended confirmed it was resolved. "Contained resolution": handled to completion but failed that check. "Assisted escalation": a human completed it | When the conversation ends: 72 hours after the first email; 2 hours after the first message on messaging by default, configurable up to 72; at hangup on voice | A request for further help; a human completing the conversation | No; Zendesk's own LLM check | [Zendesk](https://support.zendesk.com/hc/en-us/articles/9570369117338-About-automated-resolution-tiers) |
| HubSpot (Customer Agent) | The agent replied from a content source or took an action, with no qualifying handoff in the window. A lead marked qualified, partly qualified or not qualified also counts as resolved | 72 hours after the visitor's last response; a lead qualification at once | Only a handoff the visitor starts (asking for a human, a configured phrase); a manual assignment, including by workflow, does not void it. Once set, the status does not change after a live-agent message, a transfer request or negative feedback | No; the share of visitors who said the agent helped is reported as a separate figure | [HubSpot, updated 25 Sep 2026](https://knowledge.hubspot.com/customer-agent/analyze-your-customer-agents-performance) |
| Gorgias (AI Agent) | An automated interaction: a request fully handled by AI Agent (or Flows, Order Management, Article Recommendations) where the customer does not need a human | Within 72 hours | A human reply within 72 hours turns the billing into a ticket only; a handover bills the ticket fee only | No | [Gorgias](https://docs.gorgias.com/en-US/automate-pricing-289528) |
| Ada | An automated resolution: a conversation that is relevant, accurate, safe and contained (no transfer to a human), assessed after it ends | When the conversation ends: on End Chat or hang-up, when a human ends an escalated one, or after 24 hours idle on web chat and social, 72 hours on email | A transfer to a human, marked Not Resolved at once | No; reviewers can give feedback on a label in the dashboard, which does not change it | [Ada docs](https://docs.ada.cx/docs/generative/measure-success/understand-and-improve-your-ai-agent-s-automated-resolution-rate), [Ada, 17 Jun 2026](https://www.ada.cx/resources/guide/what-is-automated-resolution-in-ai-customer-service/) |
| Help Scout | A single AI chat session resolved without human help: the AI gave a full answer and the customer did not ask for more | No window stated | A request for more help | No | [Help Scout, updated 5 Nov 2026](https://docs.helpscout.com/article/1746-ai-resolutions-pricing) |
| Freshworks (Freddy AI Agent) | A session, the billed unit: a unique interaction between a customer and the AI Agent. No resolution is defined | Chat: interactions within 24 hours of the session's start are one session. Email: one 72-hour window from the first email | Not stated | Not stated | [Freshworks](https://www.freshworks.com/freshdesk/omni/pricing/) |
| Sierra | An outcome with criteria agreed in each contract, such as a resolved support conversation, a saved cancellation or an upsell | Not published | Unresolved conversations and escalations are generally not charged | Not stated | [Sierra](https://sierra.ai/blog/outcome-based-pricing-for-ai-agents) |
| Decagon | A fully resolved conversation, not defined further. The page says outcomes such as a customer who stops responding can be unclear, and recommends per-conversation billing to avoid that dispute | Not stated | Escalations are not charged under resolution billing | Not stated | [Decagon, 10 Dec 2024](https://decagon.ai/resources/pricing-ai-agents) |
| Gladly | Its glossary defines the autonomous resolution rate in general (conversations closed by AI without escalation, over conversations AI takes on) and says there is no standard industry definition. It does not say how Gladly counts | Not stated | Escalation, in the general definition | Not stated | [Gladly](https://www.gladly.ai/glossary/autonomous-resolution-rate/) |

Salesforce, Kustomer, Forethought and Crescendo define no resolution on the pages read; their billing is below.

## Billing

Read on 8 November 2026. Prices change; the linked page is the reference.

| Vendor | Billing unit | Published list price | Source |
|---|---|---|---|
| Intercom (Fin) | Per outcome, once per conversation; confirmed and assumed alike. A Procedure that ends in a handoff is its own billable outcome. Qualifications are billed separately. Fin on another help desk has a monthly minimum of 50 outcomes | $0.99 per outcome; $9.99 per qualification | [Fin pricing](https://fin.ai/pricing) |
| Zendesk | Per automated resolution in Zendesk's "Verified resolution" tier, beyond an allowance included per agent per month by plan; contained resolutions and assisted escalations are not billed | $1.50 each when 100 or more are bought ahead; $2.00 pay as you go | [Zendesk pricing](https://www.zendesk.com/pricing/), [tiers](https://support.zendesk.com/hc/en-us/articles/9570369117338-About-automated-resolution-tiers) |
| Help Scout | Per AI resolution, once per session, billed monthly in arrears; no charge when the customer asks for more help | $0.75 per resolution | [Help Scout](https://docs.helpscout.com/article/1746-ai-resolutions-pricing) |
| Gorgias (AI Agent) | Per automated interaction, on top of the ticket fee; a handover bills the ticket fee only. Each plan includes an allowance | On the pricing page, by plan | [Automate pricing](https://docs.gorgias.com/en-US/automate-pricing-289528), [Gorgias pricing](https://www.gorgias.com/pricing) |
| HubSpot (Customer Agent) | HubSpot Credits | $0.010 per credit; the pages read do not say how many credits a resolved conversation uses | [HubSpot pricing](https://www.hubspot.com/pricing/service/customer-agent) |
| Freshworks (Freddy AI Agent) | Per session | $49 per 100 additional sessions on Growth, Pro and Enterprise; 500 free sessions once per account; packs expire with the payment cycle | [Freshworks](https://www.freshworks.com/freshdesk/omni/pricing/) |
| Salesforce (Agentforce) | Per action (Flex Credits) or per conversation for customer-facing agents; no resolution is billed, and "conversation" is not defined on the page | $500 per 100,000 Flex Credits, an action being 20 credits ($0.10); $2 per conversation; user licenses on top | [Salesforce](https://www.salesforce.com/agentforce/pricing/) |
| Sierra | Per outcome, criteria agreed in each contract; routing or greeter interactions may be billed by conversation count | Not published | [Sierra](https://sierra.ai/blog/outcome-based-pricing-for-ai-agents) |
| Decagon | Per incoming conversation, or a higher rate per fully resolved conversation; never per seat | Not published | [Decagon, 10 Dec 2024](https://decagon.ai/resources/pricing-ai-agents) |
| Forethought | A platform fee plus an outcome-based cost, by quote | Not published | [Forethought](https://forethought.ai/pricing) |
| Ada | No billing unit on the pages read | Not published | [Ada pricing](https://www.ada.cx/pricing/) |
| Kustomer | No AI billing unit on the pages read | Not published | [Kustomer](https://www.kustomer.com/pricing/) |
| Gladly | No billing unit on the pages read | Not published | [Gladly pricing](https://www.gladly.ai/pricing/) |
| Crescendo | No billing unit on the pages read | Not published | [Crescendo](https://www.crescendo.ai/pricing) |

## Where the definitions agree

- An escalation to a person stops the outcome from counting, though each vendor draws the line at a different event.
- The unit is the conversation, charged once however many questions it holds (Intercom, Zendesk, Help Scout; Freshworks counts sessions the same way).
- An answer is required: a reply to a greeting alone does not count at Intercom, and HubSpot needs a reply from a content source or an action.
- Every vendor that states a window counts silence after the answer toward success; Zendesk and Ada add their own model's check.

## Where they differ

- **Window.** 72 hours at HubSpot and Gorgias and for Zendesk email; 2 hours by default, up to 72, for Zendesk messaging; idle timers of 24 hours on chat and 72 on email at Ada; none stated at Intercom, Help Scout, Sierra and Decagon.
- **Handoff.** Zendesk files a conversation a human completed as an assisted escalation. HubSpot ignores manual assignment and live-agent messages after the status is set. Intercom can still count a conversation a teammate joined after Fin answered.
- **Later contact.** Intercom deducts a resolution the customer returns to. HubSpot freezes the status once set. Gorgias turns the billing into a ticket only if a human replies within 72 hours. The Zendesk and Ada pages read do not address a customer who comes back after the conversation ended.
- **Confirmation.** Intercom names a confirmed tier and bills it like an assumed one. Zendesk and Ada use their own model's assessment. None of the definitions read requires an explicit answer from the customer that the issue was resolved.
- **What else counts.** HubSpot counts a lead marked qualified, partly qualified or not qualified as a resolution; Intercom bills qualifications as a separate outcome.

## Against Attify's standard

The [Attify Audited Resolution Standard](https://attify.app/standards/support-resolution) tests each vendor's own resolved flag exactly as set, then asks for no repeat contact within 7 days and no handoff of any kind, with a judge confirming a sample. Every window stated above is shorter than 7 days, and no vendor documents a handoff rule that strict. That is a reading of definitions and says nothing about any vendor's results; the check-by-check comparison is in [how AI support agents define and bill a resolution](https://attify.app/compare/support-ai-agent-outcome-definitions).

## Not known

- How long Intercom waits before counting an assumed resolution. Its article states no window; the figure in circulation comes from third-party writing only.
- How many HubSpot Credits a resolved conversation uses. HubSpot's pages read do not say.
- Gorgias's current price per automated interaction. The pricing page read showed conflicting plan prices, so none is given here.
- How Salesforce defines a billed "conversation", and where the Agentforce session outcome is reported.
- How Zendesk and Ada treat a customer who comes back after the conversation ended.
- Prices and resolution windows for Ada, Sierra, Decagon, Kustomer, Forethought, Gladly and Crescendo. Third-party figures exist and are not used.
- Dates for the Intercom, Zendesk, Salesforce, Freshworks, Ada, Gladly and Crescendo pricing pages, which show none.
- What vendors' contracts say. Sierra and Decagon both say outcome criteria are agreed per contract, and this research read public pages only.

## Vendors

A vendor described here can send corrections with a link to its own documentation under the [right of reply](https://attify.app/right-of-reply). Attify updates the page and the date it was read.

## Questions

### What counts as a Fin resolution, and does it count a conversation where the customer just stops replying?

Intercom counts two kinds: confirmed, when the customer replies that the answer helped, and assumed, when the customer leaves after Fin's last answer without asking for more help. Both are billed alike, so a customer who stops replying after an answer can be counted.

### A human agent finished the conversation after Fin replied. Does Fin still count it as resolved?

It can. Intercom's article says a teammate joining after Fin answered can still leave the conversation resolved; a customer asking for a human is what stops it from being billed.

### How does Zendesk decide a conversation was an automated resolution I get billed for?

Since 18 May 2026, Zendesk bills its "Verified resolution" tier: the AI answered, the customer did not ask for further help, and Zendesk's own LLM check after the conversation ended confirmed the request was resolved. Conversations that fail the check, or that a human completed, are not billed.

### If Fin answers and the customer goes quiet for 24 hours, am I billed for a resolution?

Intercom's own article gives no waiting period for an assumed resolution; the 24-hour figure appears only in third-party writing. What the article does say is that an assumed resolution is billed the same as a confirmed one.

### What window should I use to count a customer coming back as a failed AI resolution, 3 days or 7?

The vendors that state a window use 72 hours or less, and some state none. Attify's standard uses 7 days.

## Sources

- Intercom: [Fin AI Agent outcomes](https://www.intercom.com/help/en/articles/8205718-fin-ai-agent-outcomes), read 2026-11-08
- Intercom: [Fin pricing](https://fin.ai/pricing), read 2026-11-08
- Zendesk: [About automated resolution tiers](https://support.zendesk.com/hc/en-us/articles/9570369117338-About-automated-resolution-tiers), read 2026-11-08
- Zendesk: [Zendesk pricing](https://www.zendesk.com/pricing/), read 2026-11-08
- Salesforce: [Agentforce pricing](https://www.salesforce.com/agentforce/pricing/), read 2026-11-08
- HubSpot: [Analyze your customer agent's performance](https://knowledge.hubspot.com/customer-agent/analyze-your-customer-agents-performance), read 2026-11-08
- HubSpot: [Customer Agent pricing](https://www.hubspot.com/pricing/service/customer-agent), read 2026-11-08
- Gorgias: [Automate pricing](https://docs.gorgias.com/en-US/automate-pricing-289528), read 2026-11-08
- Gorgias: [Gorgias pricing](https://www.gorgias.com/pricing), read 2026-11-08
- Freshworks: [Freshdesk Omni pricing](https://www.freshworks.com/freshdesk/omni/pricing/), read 2026-11-08
- Ada: [Understand and improve your AI Agent's automated resolution rate](https://docs.ada.cx/docs/generative/measure-success/understand-and-improve-your-ai-agent-s-automated-resolution-rate), read 2026-11-08
- Ada: [What is automated resolution in AI customer service](https://www.ada.cx/resources/guide/what-is-automated-resolution-in-ai-customer-service/), read 2026-11-08
- Ada: [Ada pricing](https://www.ada.cx/pricing/), read 2026-11-12
- Sierra: [Outcome-based pricing for AI agents](https://sierra.ai/blog/outcome-based-pricing-for-ai-agents), read 2026-11-08
- Decagon: [Pricing AI agents](https://decagon.ai/resources/pricing-ai-agents), read 2026-11-08
- Kustomer: [Kustomer pricing](https://www.kustomer.com/pricing/), read 2026-11-08
- Forethought: [Forethought pricing](https://forethought.ai/pricing), read 2026-11-08
- Help Scout: [AI Resolutions Pricing](https://docs.helpscout.com/article/1746-ai-resolutions-pricing), read 2026-11-08
- Gladly: [What is autonomous resolution rate?](https://www.gladly.ai/glossary/autonomous-resolution-rate/), read 2026-11-08
- Gladly: [Gladly pricing](https://www.gladly.ai/pricing/), read 2026-11-08
- Crescendo: [Crescendo pricing](https://www.crescendo.ai/pricing), read 2026-11-08
