BLOGDate: Nov 11, 2026Reading Time: 09 Minutes

Should I Build a Custom AI Support Agent or Use Intercom Fin?

AuthorPriyanka Malve

Fin is quicker to launch. A custom agent gives you more control over cost, data and actions. Here is how to tell which one your team needs.

  1. Tag tickets Sort 200 recent tickets by type
  2. Count actions How many need a change in another system?
  3. Price both Fin fees vs build and running cost
  4. Pilot one Test on real tickets before you commit
Team comparing a custom AI support agent with Intercom Fin at a planning meeting

1. The Short Answer

Use Intercom Fin if your questions are mostly answered by help articles and you want to go live in weeks. Build a custom AI support agent if it must take actions in your own systems, follow strict data rules, or handle high volume at a lower cost per conversation.

Price is often the deciding factor. Intercom’s published pricing charges $0.99 for each conversation Fin resolves, which it calls an “outcome”. With a custom agent, you pay to build it, then for model usage, hosting and upkeep.

Key takeaways

  • Fin charges $0.99 per resolved conversation, so cost rises with volume. It works with Zendesk, Salesforce and other helpdesks.
  • Fin suits support that help articles can answer. It is the faster option to launch.
  • A custom AI support agent suits work that needs actions in your systems, strict data rules or high volume.
  • In our view, most teams should start with the option that fits their ticket mix, not the one with the lowest first bill.

2. What Is Intercom Fin?

Fin is Intercom’s AI agent. It reads your help centre, past conversations and other content you connect, then answers customers in chat, email and other channels. When it can’t answer, it passes the conversation to a person.

You don’t have to use Intercom’s helpdesk to use Fin. Intercom’s help centre says Fin also works with Zendesk, Salesforce and other helpdesks. On those, you pay only the per-outcome fee. If you use Intercom’s own helpdesk, seats start at $29 per agent a month on top.

Intercom says Fin resolves an average of 76% of customer queries. That is the vendor’s own figure, not an independent benchmark, so test it on your own tickets.

Note: The company behind Intercom renamed itself Fin in May 2026, and Salesforce completed its purchase of Fin on September 10, 2026. The helpdesk is still called Intercom. Salesforce says Fin will keep serving its customers, but check contract terms before you sign a long deal.

3. What Is a Custom AI Support Agent?

A custom AI support agent is software your team, or a partner, builds for your business. It usually combines a large language model (LLM) with your own knowledge base and connections to your internal systems.

Because you own the code, you decide which model it uses, where the data is stored, and which actions it is allowed to take. You also own the upkeep: testing, monitoring and fixing it when answers go wrong.

Two developers writing code at their desks, the kind of team that builds and maintains a custom AI support agent

4. Custom AI Support Agent vs Intercom Fin: Side by Side

  • Time to launch — Intercom Fin: Intercom says under an hour on an existing helpdesk; allow days to weeks to prepare content and test. Custom AI agent: Usually several weeks to a few months.
  • Pricing — Intercom Fin: $0.99 per outcome; Intercom helpdesk seats from $29 a month if you use them. Custom AI agent: Build cost, then model usage, hosting and upkeep.
  • Actions in your systems — Intercom Fin: Through data connectors you set up. Custom AI agent: Any action you build and allow.
  • Data and hosting — Intercom Fin: Runs on Intercom’s platform. Custom AI agent: You choose the model, region and storage.
  • Helpdesk — Intercom Fin: Intercom, or on top of Zendesk and others. Custom AI agent: Connects to whatever you use.
  • Upkeep — Intercom Fin: Intercom maintains the product. Custom AI agent: Your team or partner maintains it.
  • Best fit — Intercom Fin: Answer-heavy support with good help content. Custom AI agent: Action-heavy support, strict data rules, high volume.

Fin pricing and launch time from Intercom’s own pages, checked November 2026. Custom agent rows are our assessment of typical projects.

5. What Does Each Option Cost?

Fin’s cost grows with every resolved conversation. A custom agent costs more up front, but the cost per conversation is usually lower once it runs.

Fin resolution fees at $0.99 each

  • 1,000 resolutions a month: $990
  • 5,000 resolutions a month: $4,950
  • 20,000 resolutions a month: $19,800

Resolution fees only. Intercom seat fees are extra if you use its helpdesk. Volumes are examples, not benchmarks.

A custom agent’s running cost depends on the model, the length of each conversation and how many lookups it makes. Ask any partner for a cost per conversation estimate based on your own transcripts, not a general figure.

Warning: Don’t compare only the monthly bill. A custom agent also needs someone to test, monitor and update it. Count that time in your estimate.

6. When Should You Build a Custom AI Support Agent?

In our view, a custom build makes sense when at least one of these is true. If none are, start with Fin.

  • Most tickets need an action: Refunds, plan changes or account fixes across several internal systems.
  • Strict data rules: Health, finance or government data that must stay in your cloud or region.
  • High, steady volume: At tens of thousands of resolutions a month, per-resolution fees add up fast.
  • Unusual workflows or languages: Processes or languages a general product doesn’t handle well in your tests.

Not sure which side you fall on? We can review a sample of your tickets and tell you.

Book a free AI consultation

7. What Do Real Deployments Show?

Klarna built its own AI assistant. In its February 2024 announcement, Klarna said it handled two-thirds of customer service chats in its first month. In 2025, the company said customers would always be able to reach a person.

Customers also care about access to people. A 2024 Gartner survey of 5,728 customers found 64% would prefer companies didn’t use AI for service. Their main worry was losing access to a human. Whichever option you choose, keep a clear hand-off.

8. How to Decide in 6 Steps

  1. Tag 200 recent tickets. Mark each one as answerable from help content, or needing an action.
  2. Check your data rules. Confirm where customer data may be stored and processed.
  3. Estimate monthly resolutions. Multiply your volume by a realistic share the agent could close.
  4. Price both options. Fin fees and seats against build, running and upkeep costs.
  5. Run a short pilot. Try Fin’s trial or a small custom prototype on real tickets.
  6. Compare results. Look at correct resolutions, reopen rates and customer feedback, not only volume handled.

Tip: You can start with Fin and build later. Keep your help content clean and well tagged, so it can feed a custom agent if you switch.

9. Frequently Asked Questions

How much does Intercom Fin cost?

Intercom lists Fin at $0.99 per outcome, meaning each conversation it resolves. On another helpdesk, that is the only fee. On Intercom’s own helpdesk, seats start at $29 per agent a month on top. Check current pricing before you buy.

Can I use Intercom Fin with Zendesk?

Yes. Intercom’s help centre says Fin can connect to Zendesk tickets and messaging, import your Zendesk knowledge base, and hand unresolved conversations to your Zendesk team. It also works with Salesforce, HubSpot and Freshdesk.

Is it cheaper to build a custom AI support agent?

At low volume, usually not, because you pay to build and maintain it. At high volume, a custom agent can cost less per conversation than per-resolution fees. Estimate both using your own ticket volume and transcripts.

How long does it take to build a custom AI support agent?

It depends on scope. As a rough guide, a simple agent that answers from help content can take a few weeks. One that takes actions across several systems takes longer, because each connection needs to be built and tested.

Should customers still be able to reach a person?

Yes. Gartner’s 2024 survey of 5,728 customers found 64% would prefer companies didn’t use AI for service, mainly from fear of losing access to a human. Whichever option you choose, make the hand-off to a person easy.

10. Conclusion

Choose Fin for speed when your questions are answered by help content. Choose a custom AI support agent when you need actions, data control or a lower cost at scale. This week, tag 200 recent tickets and count how many need an action in another system. That number will point you to the right option.

If a custom agent fits, our AI customer support automation service can help you plan, build and run it alongside your current helpdesk.

See examples of our work in the AI Voice Assistant and Zenva case studies.

Author

Priyanka Malve
Priyanka Malve
AI BLOG

Related Articles

Have a project in mind?

Bring a spec, a half-formed idea, or a vision. Book a quick call with our team and we’ll map out what it takes to ship.

“Talk it through with the people who’ll build it.”