Yes, you can’t use Fin without accepting the Additional Product Terms for AI Products/Features.

For detailed information about Fin AI Agent data privacy and storage, see our legal and security guide for AI Products/Features (you will need to be logged in to view this article).

Your Intercom public articles are already imported and ready to use with Fin.

Importing external content for Fin usually takes 10 minutes. However this can take up to 10 hours depending on the source content you are using.

Yes, you can add Fin to multiple Workflows which target different audiences.

Yes, this can be customized from Fin AI Agent settings. It will carry over any branding that you already had for Operator.

Fin AI Agent is currently supported and optimized to work in multiple languages, including:

Arabic, Bengali, Bosnian, Brazilian Portuguese, Bulgarian, Catalan, Croatian, Czech, Danish, Dutch, English, Estonian, Finnish, French, German, German (Formal), Greek, Hebrew, Hindi, Hungarian, Indonesian, Italian, Japanese, Korean, Latvian, Lithuanian, Malay, Mongolian, Norwegian, Persian, Polish, Portuguese, Romanian, Russian, Serbian, Simplified Chinese, Slovenian, Spanish, Swahili, Swedish, Thai, Traditional Chinese, Turkish, Ukrainian, and Vietnamese.

Learn how to use Multilingual Fin AI Agent.

However, using Custom Answers with Fin can be set up to support any languages you’ve enabled in your Messenger settings.

Fin AI Agent is available on EU and AU hosted workspaces:

Fin AI Agent is currently supported across Intercom Messenger (on web, iOS, and Android), email, WhatsApp, SMS, Facebook, and Instagram.

No, Fin AI Agent is currently not available for conversations created via the API or email, Twitter, and Voice.

Fin AI Agent will not be supported for [TEST] workspaces on launch. You can test Fin via the teammate preview and you’ll have a free one-time amount of resolutions to test live with customers.

When you’re given access to Fin AI Agent, you'll no longer see Resolution Bot in the navigation. This is because it has evolved into a feature of Fin called Custom Answers. These Custom Answers will need to be created manually, the same way you were creating Resolution Bot answers.

Your existing Resolution Bot answers will continue to work when Fin AI Agent is added to your workspace, but they'll now be found in Fin AI Agent > Custom Answers.

Resolution Bot behaviors still exist for Fin AI Agent, but these are now set up using Workflows.

Customers on our Starter or Essential plans who don't have access to Workflows can learn how to configure behaviors and Fin AI Agent setup here.

Yes, Fin AI Agent now integrates directly with your existing support platform on Zendesk or Salesforce.

Simply import your knowledge base and connect Fin to your existing ticketing and messaging channels. When Fin isn’t able to resolve, it'll seamlessly hand off to your team in Zendesk or Salesforce.

(More platform integrations coming soon).

Any updates to your Intercom public articles will update immediately with Fin. However, Fin AI Agent updates and syncs with external content weekly. If you want to update it sooner, you can remove and re-sync the content.

You can import up to 10 external sources, and for each source, we can sync up to 3000 URLs; however, this can be less based on a number of factors like your site structure or duplicate pages.

Yes, you can control which content is available for AI Agent and apply audience rules from the Knowledge Hub.

Fin AI Agent can use PDF files, Snippets, Custom Answers, and Conversation content [beta], as well as external content sources from a public URL. You can manage content for Fin AI Agent and add new content from Knowledge Hub > Sources under the For AI Agent tab.

No, Fin currently can't create or send images during a conversation.

No, Fin can only gather information from text in the body of an article.

No, Fin can’t read ALT text in images.

If Anthropic is down, AI-generated answers from Fin AI Agent will not send (any Custom Answers will still work as normal). Customers will see a message from Fin in the Messenger: "Oops, something went wrong. Could you repeat your question?".

Inbound and ongoing conversations will be routed to the team (respecting any handover flow), with the message "Sorry, I'm still having trouble. I'll hand you over to the team."

Fin AI Agent usage is measured in resolutions. This ensures that you only pay when Fin does what you care about most: resolving a customer’s question.

Everyone is able to trial Fin AI Agent for 14 days before paying.

You can use Fin AI Agent reporting to see how Fin AI Agent is impacting your deflection and resolution rates. For a closer look at conversations and Fin’s responses, you have an Inbox view for any conversations where Fin AI Agent was involved.

Visit AI & Automation > Workflows and filter by Type is Using Fin to see all Workflows where Fin is being used.

There isn't a way to provide feedback to Intercom directly within the UI about an incorrect answer. Fin's answers are derived from your enabled support content, so optimizing this content is the best way to improve Fin's performance.

However, if you believe Fin's performance is not related to your content, you can notify our Support team so we can report this feedback (simply start a conversation in the Messenger and request to "Talk to a person"). We’re always eager to improve Fin and appreciate any feedback on irregularities or unexpected behavior.

No, Fin will formulate the answer and there's currently no way to provide any feedback outside of creating more relevant and specific content.

Clarifying questions from Fin AI Agent are built into the AI conversation experience and cannot be edited.

Assignment rules

When Fin provides a Custom Answer to a customer's question, it will immediately route conversations based on your assignment rules. This means the conversation will be visible in a team or teammate’s inbox. If you’d like to keep bot conversations out of the way, you should enable the dedicated bot inbox. Learn how to enable the bot inbox.

With the bot inbox enabled, assignment rules will only be applied after the customer has finished interacting with Fin AI Agent. This happens:

Enabling the dedicated bot inbox also lets you exclude bot time from your SLAs and certain Inbox reports. Read on to find out more.

Even when Fin AI Agent successfully closes a conversation, it will still be assigned to the right team or teammate’s closed conversation list.

SLAs

By default the time Fin AI Agent spends handling a conversation is included in your SLAs, but you can exclude this time from your SLAs with the bot inbox enabled. Just tick this box when enabling the bot inbox in AI & Automation > Settings:

Find out more about SLA rules here.

Intercom Reports

In your Intercom reports, the ‘Time to close’ and ‘First response time’ metrics will include the time Fin AI Agent and Workflows spend handling a conversation by default. You can see these metrics without bot interactions by enabling the dedicated bot inbox.

With the bot inbox enabled, you’ll see a new metric for Bot inbox time included and Bot inbox time excluded.

Choosing one of these metrics will change the values in the report to include or exclude the time a conversation spends in the bot inbox.

The important thing to know about Custom Answers is that they will trigger first if they match the customer’s question (before any AI answers with Snippets/Articles etc.) So if you have a specific answer or solution to a particular question your customers ask, you would create a Custom Answer to ensure Fin prioritizes this response.

Custom Answers are built into Workflows so they can take a little longer to set up, but this allows you to use conditional rules such as “If the customer is X, give Y answer, else give Z answer”. They can also include Custom Actions and Objects which allow Fin to fetch external data and take actions like finding a customer’s order and processing a refund. This gives you greater control over how the conversation flows and what happens at each stage.

Example:

Customer: “Can I reschedule a delivery?”

Custom Answer: “You can reschedule a delivery up to 24 hours before the delivery time. Would you like to reschedule your delivery?”

Customer: “Yes” (reply button)

Custom Action: “Fetching your orders now… Please pick a new delivery slot: [10:00] [12:00] [15:00].” (reply buttons)

*Customer selects [15:00]*

Fin: “Your delivery has been rescheduled!”

Snippets, on the other hand, are used to generate answers with AI (the same as your Help Center articles). However, unlike your Help Center, snippets are not public and only supplement Fin’s answer quality and coverage.

For example, you might have a known bug which is generating volume for your support team this week. Instead of passing these conversations onto the team to give the same response, you could quickly create a Snippet with this information in a simple Q&A format that Fin will reuse when customers run into this issue. In this instance, Snippets are a great low-lift solution to a temporary problem because they don’t take long to create and Fin can instantly start using them to reduce volume. They’re also quick to delete when you no longer need them and won’t break anything 😉

So where Custom Answers allow you to decide exactly when and how they’re used, Snippets are basically saying “here’s some information, use it where you think it’s appropriate”.

Example:

Question: “Why is it giving a 500 error message when I make a booking?”

Answer: “Our engineers are looking into this issue. In the meantime, here’s a temporary workaround…”

If you ask the same question twice to the Fin AI Agent and receive different answers, it's because the system uses a variety of AI models and language learning technology that may call on different sources at different times. This is designed to reduce latency and improve accuracy, but can result in varying answers due to the dynamic nature of the AI's decision-making process.

When Fin AI Agent is configured to use both Custom Answers and AI answers, if an end-user sends multiple messages that would trigger the same custom answer, the custom answer is only sent the first time it matches. Subsequent times, the system will try with AI answers instead, to avoid end-users being stuck getting the same custom answer for their problem after asking multiple times. This behavior is based on the assumption that if the user asks a similar question again, the initial answer might not have been what they were looking for, so the system searches for a different answer.

The Fin AI Agent can learn from interactions that have been successfully resolved without the need to escalate to a human teammate.

If customers are generally satisfied with an AI-provided answer, Fin is likely to use that answer in future conversations, as it employs machine learning to improve over time.

However, it's important to note that Fin doesn't "remember" previous answers it has given within the same conversation. If a customer asks a repeated question, Fin will attempt to answer it again using the available sources, without considering its previous responses.

Fin AI Agent is more than a chatbot: using the most sophisticated AI language models, it instantly generates answers to resolve customer issues and questions across multiple channels (including chat and email). Fin can engage in customer conversations and offer support through conversational answers based solely on your support content.

Yes, Fin AI Agent answers questions based on the articles that the customer on Messenger has access to see using audience targeting. So it won't answer questions using private or restricted articles that a customer has no access to.

When Fin AI Agent doesn’t find a clear or confident answer using your available knowledge sources, Fin AI Agent will deliver a disambiguation response to the customer which will typically:

Currently, Fin AI Agent is not able to present links in nicely formatted hyperlinks. However, you can use a Custom Answer to format the reply in any way you like.
​

​Learn more about creating Custom Answers.

When something goes wrong on our end trying to get a Fin AI Agent response in a conversation, we will send a message in the conversation that says "Oops, something went wrong. I'll route you to the team." Right after that Fin will proceed with the handover process (just like when end-users ask to get more help or talk to a person).

Fin is both an AI agent and an AI copilot. The difference between them is that Fin AI Agent is for supporting your customers, while Fin AI Copilot is for supporting your teammates on the Help Desk.

Fin AI Agent uses Intercom’s patented AI Engine to understand the customer question, search for relevant answers in your support content or data and actions, and provide accurate answers to customers based on those sources.

While we aim to provide accurate and helpful answers to your questions, it's important to remember that Fin AI Agent is powered by machine learning models, and these are not perfect. They are built on statistical patterns, and as such, occasionally they may return an incorrect response.

Here are the safeguards that we put in place to ensure that your AI Agent provides the most accurate answers:

We work hard to make sure our model is as accurate as possible, however, the nature of current AI models means they may occasionally give an incorrect answer.