Fin Product Area

We've all spent time improving Fin's knowledge, testing answers and looking at resolution rates.But I'm curious about what happens after Fin actually sends the answer.Say Fin tells a customer:“Yes, you can get a full refund.”…but your actual policy doesn't allow it.Or it gives an answer that sounds completely normal, but contains something your team shouldn't be promising.How are you catching those cases today?Manual QA? Monitors? Sampling conversations? Customer complaints? Or do they mostly get missed?I'm particularly interested in teams running Fin at meaningful volume. How are you handling this in practice?

The search bar on the modern Help Center template doesn't return results for queries under 3 characters. For example, searching "플랜" (plan) or "가격" (price) returns no results, while "플랜이" or "가격은" (the same root word with a particle attached) works fine.This minimum length may work reasonably for English, where 2-letter inputs are often low-signal, but it's overly restrictive for Korean, where 2-character words are frequently complete, high-value search terms. A language-aware minimum, or a lower threshold for CJK queries, would better serve non-English Help Centers.Could you let us know if this minimum length is configurable, or if there are plans to make it language-aware?

Hello folks! I’m trying to workout how to (if possible) create a follow-up Ticket when a NPS Survey is below a certain threshold.  The NPS Survey is currently being sent up to once every 90 days, and my intention is potentially follow-up with user’s based on their response.  I want agents to be able to review the NPS Survey response, and optionally contact the user based on the feedback given in the response.  Therefore, I was thinking the best way to go about this would be to create a ticket containing the NPS Score and NPS comment, and then the agent will be prompted to review (and optionally contact) the user.  However, I can’t work out how to get this to work via a Series; I can trigger a webhook (but that doesn’t seem capable of creating a ticket). I feel like I could probably use an Action, but I can’t work out how to trigger an action via Series.  Has anyone come across anything simila

Hello! Since view columns are stored locally in the browser’s cache, and clearing the cache resets the view columns' data to default, I developed a Tampermonkey userscript for my team to preserve selected/ordered table columns per view. Yesterday and today, it seems tests (or changes) have been conducted that caused the userscript to stop working, and the columns keep getting reset to default.Yesterday morning, the localStorage key was split into multiple different keys, causing our userscript to stop working. After a few hours into our day shift, the original localStorage key came back, and the userscript was working again. Today, the same symptoms were witnessedIs it expected that these changes made it to our environments? I’m more than happy to share more information if needed.

We have unique URLs for every page for a user. For example, User A home page URL is: toddleapp.com/platform/32476/unitPlansUser B home page URL is: toddleapp.com/platform/99172/unitPlans Will product tour work here?

ive added the intercom chat widgte to my website, but it isnt appearing on the page. the website loads normally and ive checked that the intercom code in included. when i open the website link, the chat icon doent appear, is there a setting or url-related issue i should check?

Fin doesn't read your knowledge base the way a person does. It reads structure, tags, and outcomes. Get those wrong during migration, and resolution rate suffers from day one — even if every article technically "moved."Across Intercom's customer base, Fin averages a 67% resolution rate. Top-performing teams hit 80% or higher. The gap between those two numbers usually comes down to migration quality, not the model.How Fin actually reads your knowledge baseFin doesn't treat your help center as a flat pile of articles. It uses hierarchy — Collection, Section, Article — to understand scope and context before it answers.Most help desks only support a three-level structure. Intercom allows a simplified Collection > Article path with no Section in between. That's fine natively. It's a problem in migration: if your source platform's categories don't map cleanly onto that structure, a migration tool has to guess, usually by dropping orphaned articles into a default folder. That default bucke

self‑only performance view in Workforce  (So an agent will only see their performance an not the whole team) - I believe this is in progress as per conversations with the Workforce team but just looking for an update.

In the last few weeks, it appears Intercom’s behavior around handling copy/pasted email addresses have changed and I really don’t like it. I’m trying to quickly fill out an email template and want to replace a few PLACEHOLDERS with a plain text email address like test@example.com. But now the current behavior instead of replacing the text with email is just adding a mailto: link. So instead of “test@example.com” it becomes “PLACEHOLDER” .I don’t think there’s any universe where adding a mailto link to my intercom message would ever be desired. This is causing me about 2 extra keystrokes to deal with and in aggregate it’s quite annoying behavior and unwanted.

After a lot of testing, iterations, and working through the Intercom review process, I’m happy to share that ChatShield is now officially listed on the Intercom App Store. 🎉 ChatShield monitors customer support conversations across the workspace and helps teams surface conversations that may deserve human review whether the response came from an AI agent or a human support agent. It looks for things like: Unsupported promises or commitments Potential privacy/data issues Incorrect policy claims AI representation issues Potentially risky financial or compliance-related responses The goal isn't to replace human review or legal/compliance teams. It's simply to help teams find the conversations that are worth looking at instead of manually checking everything. This is still an early-stage product, so I'm especially interested in hearing from Intercom teams using AI or automation in customer support: How are you currently checking whether your AI/support conversations contain risky response

We've had multiple instances now that calls come in (get added to a team) but no-one can pick-up the call. Is configured, and we see a button in the conversation, but when we click on it nothing happens. Has anyone else experience this?

Getting intermittent hard bounces from Outlook/Microsoft recipients:```550 5.7.515 Access denied, sending domain HELP.HUEL.COM doesn't meet the required authentication level.SPF=Fail, DKIM=Pass, DMARC=Pass``` Setup: domain SPF managed via Sendmarc, mail sent through Intercom → Sparkpost EU. DMARC passes via DKIM alignment, so by DMARC's own rules this should be fine — but Microsoft is rejecting on SPF specifically.I’m being told by Intercom that this is just an accepted loss with not being able to respond to these customers.  Anyone else experience this?  

We connected a POST API to SendGrid within our email workflow to capture FIN's response to users and store it as audit evidence in forwarded emails. However, we haven't found any parameter or attribute that exposes FIN's response text.Is there a way to create a custom attribute to pass this value, or do you recommend an alternative method to capture it? ​@Nathan Sudds 

Hey everyone 👋As Fin handles more customer conversations, I'm curious how teams are checking the actual responses being sent to customers.Not just whether Fin resolved the conversation but whether the response itself could create a problem.For example:promising a refund or exception that isn't actually allowedgiving incorrect pricing or policy informationmaking a commitment the business can't honorexposing sensitive customer informationgiving inappropriate financial or medical advicemissing an important AI disclosureThe difficult cases aren't necessarily obvious. A response can look perfectly reasonable on its own but become risky when you look at the customer's situation and the company's policies.How are you handling this today?Manual sampling? QA? Fin's monitoring tools? Internal rules? Something else?I'm working on this problem myself and would genuinely like to understand how other Intercom teams are approaching it.

I have great difficulties managing the Brand appearance once chat goes to email, due to the fact that I can only sign up with one email adress.  When I turn off using my own email Intercom will still show my workspace name in the generated email Even if it says "Viktor from Brand 1" or "Viktor from Brand 2" The email adress is still limited to one domain or Intercoms own. I cant be the only one working with multibrand setup. How do I grow if I cant customise my brands. Any advice / tips /workarounds?  Did not get lucky in Customer Support (BTW I have the possibility to edit different brand styles)

Hi everyone,We’re using Fin in Hebrew, but we keep seeing cases where Fin adds random English words inside Hebrew replies, even when those words do not appear in our Help Center articles or snippets.We also had a more serious case where Fin inserted Japanese/Chinese-looking characters into a Hebrew customer-facing reply.We already added clear Guidance telling Fin to reply fully in Hebrew and avoid English unless it is a brand name, URL, product name, or a required term. The issue still happens.Intercom Support suggested using Guidance and mentioned the multilingual glossary, but also clarified that the glossary does not apply to AI-generated replies.Has anyone using Fin in Hebrew or another non-English language found a reliable way to prevent this?Is this something Guidance can actually solve, or is this a current limitation of Fin?This is affecting customer trust, so any practical advice would be appreciated.

Hi All, I’m wondering if there’s a report we can generate to see how many unique users sent us messages?I want to compare adoption between 2022 and 2023. Your help putting this together would be appreciated. Thanks! Michael

Hi Intercom team and community,I’m paginating through a company segment using the /companies endpoint and I’m seeing exact duplicate company records appearing on multiple pages.For example:GET https://api.intercom.io/companies?segment_id=xyz&per_page=60GET https://api.intercom.io/companies?segment_id=xyz&per_page=60&page=2Some of the same company records show up in both the page 1 and page 2 responses. Items are not updated between paginated queries.This makes it difficult to reliably process the full dataset without extra deduplication logic.Is this expected behavior for the Companies API? And is there a recommended way to retrieve a segment’s companies without duplicates across pages?

Hi all,I'm building a labeling/tagging system for a fintech support team in Intercom and want to start simple, then expand.Context:Small support team handling inbound fintech queries (payments mostly) Using tags only for now (no custom attributes yet) Eventual goal: pipe tagged conversation data into a data warehouse for reporting/analysisWhat I'm looking for:Examples of tag taxonomies others have used for support (fintech) Is it ok to use inly tags and automated topics?  Pitfalls to avoid — tag sprawl, inconsistent application by agents, etc.Happy to share what I land on once it's live, in case it helps others in the same boat. Thanks in advance!

I'm evaluating Intercom because I'm thinking of switching from Freshworks (Freshdesk+Freshchat).My WhatsApp number is linked to Freshworks. If I switch to Intercom, can I unlink it and link it to Intercom instead? Gustavo

Hey,  Saw your profile, I was wondering if you can help me build operational flows and escalation flows between teams. Also, help me set up priority and ticket status rules.  

Hi,I’m trying to connect Intercom to Marketo through the Marketo app from the Intercom app store.I have the Client, Secret ID and the API subdomain.I have this error message : “We couldn't save this app's settings. Try again in a minute”Is anyone have this issue?

Hey everyone 👋I'm curious how Intercom teams are handling something that seems to get harder as support volume grows.How do you currently make sure that every message sent to a customer is safe and accurate whether it's written by a human agent or Fin?A support agent might accidentally:promise a refund they aren't authorized to givegive incorrect pricing or policy informationmake a commitment the company can't honorexpose sensitive customer informationgive inappropriate financial/medical advicesend something that creates a potential liability issueAnd with AI handling more conversations, there can be hundreds or thousands of responses that nobody manually reviews.Do you manually sample conversations, rely on CSAT/customer complaints, have a QA process, or are you monitoring them automatically?I've been building ChatShield specifically for this: it connects to Intercom and monitors workspace conversations not just AI conversations to flag potentially risky messages so teams can review

I conducted a site audit of my Help Centre and it returned a lot of redirects and 403s. The 403s I was able to fix. But the redirects and canonicals, I don’t know how to fix.Any help is appreciated. 

Hey everyone For those using Intercom with AI/automation, how are you currently monitoring the conversations happening in your workspace for things like:AI making an incorrect refund or pricing promiseIncorrect policy information being given to customersSensitive financial/medical advicePII or other sensitive information being exposedAI responding without the appropriate disclosureI'm curious whether you're manually reviewing conversations, relying on Intercom's existing tools, or using another compliance/QA solution.I’ve been building ChatShield around this exact problem it connects to Intercom and monitors workspace conversations for potentially risky interactions, rather than only checking the AI bot itself.If anyone here is dealing with this problem, I'd genuinely love to hear how you're handling it today.ChatShield: https://chatshield.cloudNot trying to sell anything here mainly interested in understanding how other Intercom teams approach this.