Share product ideas and upvotes with our product team
We have recently started using the “Ask for more information before handover” setting within the “Let Fin Handle” workflow step.We’ve noticed an issue where, if a customer asks “Talk to a person” but then doesn’t respond to Fin’s follow-up question, the conversation is automatically handed over to a teammate after 30 seconds.Could you consider adding an option to customise this timeout period, or alternatively disable the timeout entirely?30 seconds can sometimes be too short for a customer to provide a complete explanation of their issue, particularly when they need to gather information, review details, or formulate a more detailed response. Having the ability to increase the timeout, or remove it altogether, would give customers more time to provide the requested information before the conversation is escalated.
⚠️ Problem StatementCurrently, when a Procedure Builder configures a flow to fetch external data (e.g., retrieving a list of recent orders, linked accounts, or purchased products via an API), the experience often breaks continuity. After the data is fetched, Fin has to ask the user to manually type out the specific order number or product name into a blank text input. This manual transcription introduces friction, increases the margin for user error, and degrades the premium "app-like" messenger experience. ✅ Proposed Solution / Feature ConceptIntroduce an automated, temporary messenger app experience generated dynamically from API payloads. Instead of returning raw text or forcing a manual text reply, Fin should leverage an ephemeral UI component (a "Preview App") directly inside the messenger canvas.When an API response returns an array or an object collection, the Procedure Builder should be able to map those objects to interactive UI elements. The customer can then simply click/tap their selection from the visual response to advance the flow. 🛠️ Example Use Case / Flow Step 1: Fin initiates a backend check: "Display all orders for this customer." Current State: Fin lists the text or drops a blank input box: "Please type the order ID you are referring to." Proposed State: Fin renders an interactive list/carousel card component showing the last 3 orders (Order #, Date, Total). Step 2: The customer clicks the specific order card. Step 3: Fin instantly updates the context and moves to the next interactive picker: "Which product from this order are you referring to?" (Renders clickable Product Items).
Product Tours are not automatically multi-language like Help Center articles. All custom translation settings within Intercom don't impact the text we put in the tours. The content of the tour is not translated automatically to the viewer’s browser language. If you want multiple languages, you need to duplicate the tour and manually translate each version. The only part that follows the browser language is interface text (like the “Next” button), and only if that language is enabled in your Messenger languages. If we're not benefiting from the existing translation and language support we've already set up in Intercom, there's no incremental benefit of using Intercom tours versus another provider.
It would be great if the YouTube videos linked to the help centre article related to a topic would be included in the Fin response, particularly for ‘How To’ questions.
'Customers will see this AI-generated title in their Messenger. It's created automatically based on the conversation.' - I need to disable it. It's confusing our customers.
I’d like to see the ability to turn off smart replies (at the admin level) to ensure an entire team is / is not using them as preferred by the org.
When a human agent takes ownership of a case, particularly for complex, multi-step issues like repair bookings or courier collections, and Fin subsequently handles part of that conversation (e.g. answering a status update, running a procedure), there is currently no native way to automatically return the conversation to the original owning agent once Fin's involvement ends or escalates. Our agents handle cases that span multiple touchpoints. An agent may book a scooter collection or repair, making them the case owner with full context. If the customer later contacts us again, Fin may handle the interaction but when Fin escalates or the conversation needs human attention again, it lands in an unassigned team queue. The original agent loses the case unless they find it or it gets manually reassigned to them. The only current workarounds I can think of:Agent name tags + a workflow branch per agent (brittle, requires maintenance as the team changes) Creating a new custom attribute (however, we've exhausted our attribute allowance on reporting-critical fields)None of these are clean solutions, they all require ongoing manual maintenance or sacrifice something else.The ask: A native "Case Owner" concept at the conversation level, separate from the current assignee, that: Can be set by an agent when they take ownership of a case Is preserved throughout Fin's involvement in the conversation Automatically reassigns the conversation back to the case owner when Fin escalates, hands off, or the conversation leaves the bot inbox Alternatively, a workflow trigger condition of "previous human assignee" that can be used to route conversations back to whoever held the conversation before Fin took over. As teams adopt Fin more deeply particularly on email channels where conversations reopen and span days or weeks, case continuity becomes critical. Fin is excellent at triaging and answering, but the handoff back to a human needs to be intelligent, not just "drop it in the team queue." This feature would allow teams to confidently expand Fin's scope without sacrificing the accountability and context that comes with named case ownership. This would affect any teams using Fin on email or async channels with multi-touchpoint cases, repairs, logistics, complaints, or anything requiring follow-up over time. ThanksEd
Allow admins to set a time limit for how long customers can submit a CSAT rating after a conversation handled entirely by FIN.Currently, customers can still rate conversations several months after they took place, which can result in feedback that no longer reflects their original experience. While a time limit is available for the separate CSAT app, this option is not available in the “Let Fin Answer” steps in Workflows.Requested improvement: Allow admins to configure how long FIN CSAT surveys remain open after a conversation. Automatically close the survey once the configured time limit has passed, preventing customers from submitting ratings on older conversations.
We sell a lot of expensive medication. I.e., weight loss jabs (i.e., Mounjaro/Wegovy)It can be over £200 for a month's supply of weight loss medication.We get customers writing in all the time saying they're having issues with their pen. Their pen is broken, and it needs replacing.We are, of course, happy to replace their pens, but really often what is needed are some good troubleshooting instructions.i.e. Have you tried this? First try this, then try this.This could be very smoothly achieved if Fin was able to walk a user through various steps with how-to videos. These pens are a physical medium, and it's quite hard to convey what you need to do to a beginner with just text. I feel a video, or a series of videos following a series of steps in a procedure, would really help bring this to life.
We’d love a way to provide direct feedback on Fin’s category detection, particularly for attributes and list-based categorization. For example, when Fin assigns the wrong category, it would be useful to mark the selection as incorrect and specify the correct category.In cases like this one, the category itself is well defined, but Fin’s reasoning led it to the wrong selection. Having a feedback mechanism for “selected category vs. correct category” that could be used to improve future categorization would be extremely valuable and could significantly improve accuracy over time.
Currently, the only way to use Claude is to connect the Intercom MCP and perform in-article updates or fetch articles, etc., but it feels half-baked and lacks many useful features. For example, Claude can’t edit multiple languages, only the main language can be added. For me, that’s German, so I can’t edit the English version.We also need an HTML/CSS editor within Intercom. When Claude generates an hc article structure and I don’t want to push it via the MCP, I should be able to copy the HTML and paste it into Intercom so everything is interpreted correctly and formatting issues don’t occur. Also editing Snippets is important, which he cannot access.More broadly, give Claude more access to the workspace so he can reference resources and assemble content. Right now, using him beyond fetching articles is painful. This is the functionality I rely on most, but other capabilities are lacking because he doesn’t have full Intercom access.
I would like it if we could have the chat bot view/reference all of the Macros we have set up to use as pre-defined guidance on what to say and do in a given scenario.
It would be nice to use manual reviews submitted by teammates to help train the AI reviewer for similar scorecards
I’d like to report a potential improvement regarding the Audience configuration.Recently, we expanded the operation to other products. Given this scenario, managing the materials now requires more robust audience rules to prevent content related to other products from being incorrectly used in ERP support interactions, and vice versa.After reviewing the current Audience configuration, I noticed that it is not yet possible to use the Sector attributes we created for chat classification as a rule criterion. These attributes are currently essential for correctly routing each interaction to the appropriate Fin configuration and, consequently, to the support team responsible for the respective product.If we could use Sector as a condition in Audience rules, we would be able to segment customers and materials more precisely according to each product. In addition, this configuration would make future maintenance and adjustments much easier, as we could focus corrections and improvements on each sector's specific prompt, without having to revise broader audience rules.Given this need, we would like to request the addition of Sector as an available criterion when creating Audience rules. We believe this improvement would help ensure more accurate material segmentation, prevent inappropriate content from being used across different products, and make it easier to manage and continuously improve Fin.
Our team uses tons of slack channels for communication with customers, so it would be excellent if Procedures triggered on Slack conversations. They have been really impactful to help with troubleshooting and resolution rate, but since they aren’t supported on slack convos, we are hitting a plateau.
If an email comes in and is automatically marked spam that isn’t spam, we mark it not spam. From there, it should follow the workflows that we have setup, but doesn’t. We have emails that are sent to a specific email address that just need routed to that team inbox (things like notifications from a tax agency that a return was processed), Fin never is to get involved. But, if one of those emails arrives marked spam, and is manually unmarked as spam, Fin replies to it and there are no workarounds to prevent that. It either needs to follow the workflows or do nothing so it can manually be handled.
This synchronization looks very neat: https://www.intercom.com/help/en/articles/15943800-sync-content-from-githubUnfortunately, it lacks support for multilingual content, which is a significant drawback.
Currently Fin Recommendations (content gaps, data gaps and action gaps) only have two options - reject or mark as done/approve. Rejecting removes the recommentation and stops Fin from suggesting it again in the future. In our case this is limiting as a recommendation (in particular or data and action gaps) may not be technically possible today, but as technology changes the recommendation may make sense at some future time. It would be amazing if recommendations had a option for a “soft rejection” where it may be rejected now but we’d like Fin to make the recommendation again (if warranted) at some future time, someting like “remind me again in X months”It would also be great if: there were an audit trail, where we could see what recommendations were accepted and rejected. In addition to an audit, being able to specify reason for rejection, this would help with identifying the recommendations that may not make sense today but could in the future. Currently to overcome this limitation in auditability and tracking we need to keep a running list of recommendations in a spreadsheet, noting the recommendation, when it was suggested and the reason for rejection and review it regularly to see what might be feasbile today that perhpas wasn’t a few months ago. This adds extra overhead to our process and means having to bounce between applications vs having one source of truth within Fin.
Already have an account? Login
No account yet? Create an account
Enter your E-mail address. We'll send you an e-mail with instructions to reset your password.