An AI customer service agent answers customer questions and, increasingly, takes actions such as checking an order, starting a return or updating an address, without a person in the loop. For an ecommerce brand, the first choice is usually the agent built into its help desk: Gorgias AI Agent, Intercom Fin or Zendesk AI agents. All three now charge by results. A custom agent, built on the brand's own data and systems, costs more to start but can do work the help desk agents cannot reach. This guide sets out what each option costs at published prices and when a custom build is worth it. It is from Brand Media Works, an AI consulting and marketing data engineering firm for ecommerce brands.
How do help desk AI agents charge?
- Intercom Fin: $0.99 per outcome, with no seats required, and it can run on top of another help desk, including Salesforce. An outcome counts when a customer confirms the issue is resolved, does not ask for more help after Fin responds, or Fin completes a workflow, including a handoff. Intercom charges once per conversation, even if Fin answers several questions (read 27 October 2026).
- Zendesk AI agents: included in every Suite and Support plan and billed per automated resolution, which Zendesk defines as a request resolved without escalation to a human agent. On its original resolution platform, each resolution is verified by a large language model, and plans include 5 to 15 automated resolutions per agent per month, up to 10,000 a year. Buying 100 or more resolutions in advance costs less per resolution than paying as you go (read 30 October 2026). The pricing page does not list the per-resolution rate.
- Gorgias AI Agent: available on every plan, and the brand pays only when it resolves a conversation. Gorgias prices its help desk by tickets, not agents, and conversations fully resolved by AI Agent do not count as billable tickets (read 30 October 2026). The per-resolution rate was not shown in the pricing page text.
- Salesforce Agentforce: $2 per conversation, or Flex Credits at $500 per 100,000 credits, with each action using 20 credits (read 27 October 2026). That works out to $0.10 per action (our arithmetic).
Per-resolution pricing means the bill grows with success. At Intercom's $0.99 per outcome, 1,000 resolved conversations a month cost $990, and 3,000 cost $2,970 a month, or about $35,640 a year (our arithmetic). That is the number to compare with a custom build.
When does a custom support agent beat the help desk's own agent?
A custom agent is built on the brand's own policies, order data and systems, and it can use a model the brand runs in its own cloud account. Published build prices from AI agent firms give the range:
- One firm publishes $15,000 to $50,000 for a first production agent, typically 2 to 3 months, $50,000 to $100,000 when it acts across several systems, and $5,000 to $10,000 a month to keep improving it. It puts model fees at about $4 to $400 per 1,000 tasks, depending on the model (read 18 October 2026).
- Another firm prices a customer support agent at $40,000 to $120,000 to build (read 18 October 2026).
- A third puts annual maintenance at 15% to 25% of the build cost, and says a task should happen at least 500 to 1,000 times a month for the return to arrive within 24 months (read 18 October 2026).
On those figures, a custom agent tends to win in four situations:
- High volume. Once resolved conversations run into the thousands a month, per-resolution fees approach the cost of a build plus its upkeep.
- Actions outside the help desk. Returns portals, subscription apps, warehouse systems and loyalty programs often sit outside what a help desk agent can change. A custom agent can read and write to each of them, with limits and an approval queue.
- Customer data that has to stay in the brand's cloud. A custom agent can run on a model inside the brand's own AWS account. What a private LLM costs to run sets out those costs.
- Several brands, stores or languages that need one agent with shared rules and one evaluation set.
The help desk's own agent is the better choice when volume is modest, most questions are order status, shipping and returns policy, and the team already works in that help desk. Custom AI tools describes how Brand Media Works scopes and builds support agents, starting from the brand's own ticket history.
Gorgias, Intercom Fin or Zendesk: which fits an ecommerce brand?
| Platform | How the AI agent is billed | Help desk pricing | Fit for an ecommerce brand |
|---|---|---|---|
| Gorgias | Only when AI Agent resolves a conversation. Rate not shown in page text | By tickets: $10 a month for 50 tickets to $900 for 5,000, never per agent | Gorgias says it powers customer conversations for 40% of Shopify brands. Resolved conversations are not billable tickets |
| Intercom Fin | $0.99 per outcome, once per conversation | Fin needs no seats and can run on another help desk | Can be added to an existing help desk without switching |
| Zendesk | Per automated resolution, verified by an LLM. 5 to 15 included per agent per month on the original platform | Per agent: $19 to $115 a month on annual billing, $25 to $149 monthly | Larger teams with several channels and a per-agent budget |
| Salesforce Agentforce | $2 per conversation, or $500 per 100,000 Flex Credits | Priced per conversation or per action | Brands already on Salesforce |
How much is the Gorgias AI agent?
Gorgias bills AI Agent per resolved conversation, on every plan, and the help desk underneath is priced by tickets. Its own plan guide, updated 14 May 2026, lists Starter at $10 a month for 50 billable tickets, Basic at $60 for 300, Pro at $360 for 2,000 and Advanced at $900 for 5,000, with extra tickets at $0.40 on Starter and Basic and $0.25 on Pro. Gorgias says annual billing saves up to 17%. Because conversations that AI Agent fully resolves do not count as tickets, the agent's fee replaces a ticket charge rather than adding to it. The per-resolution rate depends on the plan and was not shown in the pricing page text on 30 October 2026.
What should a brand check before switching on an AI agent?
- What counts as a resolution. Intercom counts a conversation where the customer does not ask for more help after Fin responds. Zendesk ends a messaging conversation 2 hours after the first message and an email conversation after 72 hours before it judges whether it was resolved. Read the definition and check it against real tickets.
- A usage cap. Zendesk lets an account pause AI agents or allow overage when it reaches its limit. Set the cap before launch.
- Which systems the agent can act in, and which actions need a person to approve them, such as refunds over a set amount.
- A test set from the brand's own tickets, covering refunds, damaged orders and questions the agent must not answer.
- Where customer data goes, and whether the vendor uses it to train models.
Frequently asked questions
What is an AI help desk?
A customer service platform where an AI agent answers and resolves customer conversations alongside human agents, with tickets, routing and reporting in one place. Gorgias, Intercom and Zendesk all build an AI agent into their help desks and charge for it by resolution or outcome.
Which AI agents are best for ecommerce support?
For a Shopify brand starting out, the agent built into its help desk is usually the quickest win: Gorgias AI Agent if the brand is on Gorgias, Zendesk AI agents on Zendesk, or Intercom Fin, which can run on top of another help desk. A custom agent is worth considering once volume is high or the agent needs to act in systems the help desk cannot reach.
Do AI-resolved conversations count as tickets in Gorgias?
No. Gorgias says conversations fully resolved by AI Agent do not count as billable tickets. The brand pays the AI Agent fee for those conversations instead.