Per-Resolution Pricing: What One AI Answer Really Costs Your Business

What does a business actually need from AI support pricing? A number it can budget. Per-resolution pricing – paying a fixed fee every time the AI resolves a customer conversation – sounds like the fairest deal in software: you pay only for success. The published rates look small. Zendesk charges $1.50 per automated resolution on a committed plan and $2.00 pay-as-you-go. Salesforce charges $2 per Agentforce conversation, and $2 per Help Agent resolution. Gorgias charges $0.90-$1.00 per AI interaction – and also counts that interaction as a billable ticket.

Small rates, serious bills. At 1,000 resolutions a month, the AI layer alone costs about $1,500-$2,000 a month at Zendesk and $2,000 a month at Salesforce – $18,000 to $24,000 a year, before the helpdesk seats underneath it. The same workload on Oscar Chat, priced on published rates, costs about €43 a month. And per-resolution bills have a property no flat plan has: they spike in your best month, in direct proportion to how well the AI works. This article shows the math, the definitions hiding inside the word “resolution”, and the questions to ask before you sign up for success-based billing.

Written by:

Matt Maloney, Oliver Cole

In Publication:

ON November 13 2026

AI chatbot
Per-resolution AI pricing explained: what one automated answer really costs a business

First, the honest conclusion

Rates are the vendors’ published figures, checked on their official pricing pages on 12 November 2026. Worked examples state their assumptions: a resolved conversation averages about 5 AI messages.

Budget AI support without a per-resolution toll

1,000 AI resolutions a month cost $1,500-$2,000 at Zendesk and about €43 on Oscar Chat’s published rates.

Start free with Oscar Chat

Free plan to start · No credit card required

The rate card: who charges what per resolution

VendorPublished rateWhat the meter counts
Zendesk AI agents$1.50 per automated resolution (committed) / $2.00 pay-as-you-goAn automated resolution; only 5-15 included per agent per month depending on suite tier, the rest is overage
Salesforce Agentforce$2 per conversation / $2 per Help Agent resolution; or Flex Credits at $0.10 per standard actionA conversation (billed resolved or not on the Conversations meter), a resolution (Help Agent), or each action the agent takes (Flex Credits)
Gorgias AI Agent$0.90 (annual) / $1.00 (monthly) per automated interactionEvery AI interaction – and the same interaction also counts as a billable ticket on the helpdesk plan
Oscar ChatFlat: Basic €33/mo includes 1,000 AI messages, extra messages €2.4 per 1,000; Pro €65/mo includes 2,000Messages inside a flat platform subscription – no per-resolution meter at all

Three different units hide in that table: a resolution, a conversation and an interaction are not the same event, and an “action” is smaller than all three. Comparing headline rates without comparing definitions is how stores end up surprised.

The monthly math at three volumes

400 resolutions a month (a small store having a decent month):

  • Zendesk AI layer: $600-$800
  • Salesforce: $800
  • Gorgias: $360-$400 for the interactions, plus 400 billable tickets consumed on the helpdesk plan
  • Oscar Chat: about €35 – the roughly 2,000 AI messages fit the Basic plan’s included volume plus a few euros of published extra messages

1,000 resolutions a month:

  • Zendesk AI layer: $1,500-$2,000 ($18,000-$24,000 a year)
  • Salesforce: $2,000 ($24,000 a year)
  • Gorgias: $900-$1,000, plus 1,000 billable tickets on top
  • Oscar Chat: about €43 (roughly 5,000 messages: 1,000 included, the rest at €2.4 per 1,000)

3,000 resolutions a month (a good store in a peak month):

  • Zendesk AI layer: $4,500-$6,000
  • Salesforce: $6,000
  • Gorgias: $2,700-$3,000, plus 3,000 billable tickets
  • Oscar Chat: about €67 (roughly 15,000 messages on the same published rates – at this point most stores sit on the flat €65 Pro plan anyway)

Read the Zendesk and Salesforce columns again. These are not enterprise edge cases; 1,000 resolutions a month is a mid-size Shopify store answering order-status, returns and product questions. The per-resolution model takes a five-figure annual line item out of a workload a flat platform prices in double digits.

The definition game: what counts as a “resolution”?

The rate is only half the contract. The other half is the vendor’s definition of the billable event, and definitions differ in ways that all lean the same direction:

  • Resolved vs touched. Salesforce’s Conversations meter bills every conversation at $2, resolved or not; its Help Agent meter bills only resolutions. Zendesk’s automated resolutions are counted by its own resolution logic. Gorgias bills per interaction – a lower unit than a resolution – and bills the ticket as well, so one customer problem can generate two charges.
  • Included volumes are small. Zendesk includes 5-15 automated resolutions per agent per month. A single busy agent-day can consume a month’s inclusion; effectively the whole workload runs at overage rates.
  • Actions multiply. On Salesforce Flex Credits, the meter is the action – about $0.10 per standard action – so a conversation in which the agent looks up an order, checks stock and issues a label is several billable units, not one.
  • You cannot audit the counter easily. The vendor’s system decides what counted. Your dashboard shows the total; the definition lives in their documentation.

None of this makes per-resolution pricing dishonest. It makes the headline rate nearly meaningless without the definition – and the definition is the part nobody puts on the pricing page’s hero banner.

The spike problem: success is the billing event

Every other software line on your P&L is at its most predictable in your best month. Per-resolution AI is the exception: the bill is a direct function of how much your customers needed help and how well the AI handled it. A Black Friday month at 3x volume is a 3x AI bill (see the Black Friday support playbook for how the volumes move), arriving alongside peak ad spend and peak shipping costs. Finance teams hate variance; per-resolution pricing is variance by design.

There is a subtler effect. When each resolution costs $1.50-$2.00, teams start rationing the AI: narrower deflection targets, more conversations forced to humans “to save money”, automation switched off in peak weeks. The pricing model quietly taxes the exact behaviour it was bought for. Flat pricing has the opposite incentive – automate everything automatable, because the marginal answer is already paid for.

When per-resolution pricing is genuinely fair

Honesty requires the other side. Per-resolution pricing can be the right deal when:

  • Your volume is low and spiky in the other direction – a store with 50 resolutions a month pays about $75-$100 and buys enterprise-grade AI it could never justify as a platform. (Though check the flat free and entry tiers first – Oscar’s free Lite plan and €33 Basic exist precisely in this zone.)
  • You are an enterprise replacing headcount, not software. At thousands of resolutions, a $1.50 resolution that replaces a $6-8 human-handled ticket is a clear win against staffing cost. The comparison that matters there is cost per resolution vs cost per human resolution – not vs a flat chatbot plan.
  • The vendor guarantees the outcome definition. If “resolution” is contractually tied to verified customer satisfaction rather than the vendor’s internal logic, the alignment is real.

The model fails stores in the middle: too much volume for the per-unit math to stay small, too little for enterprise negotiation to fix the definitions.

How to budget before you sign: five questions

1. What exactly is the billable event? Get the definition in writing – resolution, conversation, interaction or action – and what triggers it. 2. What is included before overage? Count it against your real monthly volume, not your hopeful one. 3. Can the same conversation bill twice? If AI interactions also consume helpdesk tickets or credits, model the stack, not the rate. 4. What does my peak month cost? Multiply by your November volume, not your average month. 5. What is the flat alternative for the same workload? Price it on published rates – if a vendor cannot give you a number for your volume, that is the answer.

questions & answers

FAQ

  • What is per-resolution pricing?

    A billing model where the vendor charges a fixed fee each time its AI resolves a customer conversation: $1.50-$2.00 at Zendesk, $2 per resolution or conversation at Salesforce. The fee applies per event, on top of the platform or seat subscription underneath.
  • Is per-resolution pricing cheaper than a flat subscription?

    At very low volumes it can be. At 1,000 resolutions a month it costs $1,500-$2,000 a month for the AI layer alone, while a flat platform like Oscar Chat prices the same workload at roughly €43 a month on published rates. The crossover arrives early – usually within the first few hundred resolutions.
  • What is the difference between a resolution, a conversation and an interaction?

    A conversation is one customer dialogue; an interaction is a single AI exchange inside it; a resolution is a conversation the vendor’s logic counts as solved. They are different units atdifferent prices, and vendors choose the unit that their pricing page does not always make obvious.
  • Why does per-resolution pricing spike on Black Friday?

    Because the bill is proportional to resolved volume, and BFCM multiplies support volume. A store doing 1,000 resolutions in a normal month may do 3,000 in the peak month – and pay for all of them, at peak-season rates, in its most expensive month.
  • Does Oscar Chat charge per resolution?

    No. Oscar Chat is a flat subscription (€33-€65 a month) with AI messages included – 1,000 on Basic, 2,000 on Pro – and a published rate of €2.4 per extra 1,000 messages on Basic. There is no per-resolution, per-conversation or per-outcome meter.

Pay for the platform, not for every answer

Per-resolution pricing reframes your support from a cost you control into a toll on every happy customer. The rates look like rounding errors – $1.50 here, $2 there – until a real month of real questions turns them into five figures a year. Price your actual volume both ways before you commit: the math in this article takes ten minutes with your own ticket count. Oscar Chat’s answer is a flat subscription where the ten-thousandth answer costs the same as the first – nothing extra.

Price your volume both ways before you sign

A flat subscription with AI messages included and a published €2.4 per extra 1,000 messages. Price your volume both ways, then start free.

Start free with Oscar Chat

Free plan to start · No credit card required

Data sources: official pricing pages and documentation of zendesk.com, salesforce.com (Agentforce pricing), gorgias.com (AI pricing FAQ) and oscarchat.ai – checked 12 November 2026. Worked examples assume about 5 AI messages per resolved conversation and Oscar’s published included volumes and extra-message rate.