Share product ideas and upvotes with our product team
The Problem:We have bugs that arise, just like any other tech company. We’re constantly innovating and pushing updates, and therefore encounter unexpected issues. We typically fix widespread issues and outages very quickly. And we link all customers to a tracker ticket. Fin spends way too much time blindly leading a customer through troubleshooting steps, when there’s a deeper issue Fin is unaware of. How we solve today:It takes a lot of effort to create a snippet, and create guidance for the snippet, to quickly train Fin on a known issue. And then that guidance and snippet has to be manually deleted hours or days later when the issue is resolved. It’s constantly evolving, requiring multiple steps. We create tracker tickets to inform customers and human agents of the progress of issues like this. So I’m training my agents and Fin separately, in two different places. The Dream Scenario: Awareness: Fin is aware of all active Tracker Tickets logged by human agents. Mitigation: Fin automatically offers a workaround to the customer if one is documented. Proactive Subscription: Fin asks the customer if they would like to be notified of updates regarding the bug. Automation: Upon confirmation, Fin automatically links the conversation to the Tracker Ticket and converts the conversation into a customer ticket. Persistence: Fin remains the primary point of contact unless the issue requires human escalation. Resolution Loop: When the Tracker Ticket is marked "Resolved," Fin proactively follows up with the customer to confirm the fix is working for them. Proposed Implementation in IntercomTracker Ticket Settings Add a "Sync to Fin AI" toggle on Tracker Tickets (or enable this by default for specific Ticket Types). When enabled, provide specific AI-optimized fields: Symptoms: Describe what a customer would experience (to help Fin match the issue). Workarounds: Step-by-step instructions for Fin to provide. Target Audience: Specific conditions or user segments impacted. Auto-Link: Toggle option to allow Fin to automatically link users to a Tracker Ticket. Auto-Ticket Creation: Toggle option for Fin to generate a linked conversation ticket automatically upon bug identification.
What I'm requesting:Expose the data from the legacy Help Center Articles report (found under Reports > Proactive Support > Articles) through the Intercom Reporting API, so it can be surfaced and used within AI-assisted analytics tools like Intercom's own Operator/Copilot assistant.The data I'm referring to includes:Article viewer counts (per article, per time period) Reactions (happy / neutral / sad) Conversations triggered per article Last updated dateWhy this matters:We use our Help Center extensively as a first line of support. This data already exists in Intercom — it's visible in the legacy report — but it's siloed from the reporting API, which means AI tools can't access or reason about it.Right now, when I ask Intercom's AI assistant to analyze help center content performance, it can tell me which articles Fin cited in conversations or what resolution rates look like by topic — but it can't cross-reference that with how many customers actually viewed those articles or how they reacted to them. That's a significant blind spot.The use case is straightforward: combining article view and reaction data with Fin's resolution rates, topic breakdowns, and conversation outcomes would give support teams a complete picture of how their knowledge base is performing — not just from Fin's perspective, but from the customer's.What good looks like:Being able to ask the AI assistant: "Which articles have high views but low positive reactions or high follow-up conversation rates?" — and get a data-backed answer without having to manually cross-reference two separate reports.Who benefits:Any team using the Help Center seriously — especially those using Fin as their AI agent — would immediately benefit from this connection being made. It closes the loop between content publishing and support outcomes.
Problem:There's nowhere on an article to leave a note for other editors. Anything an editor needs to know before changing an article has to live outside Intercom, where the person making the edit won't see it.Examples of what a note would carry:Linked from the product. "This article is linked from inside our app, do not delete it” Used by Fin configuration. "A procedure or guidance rule depends on this article. Check it before restructuring or removing sections." Owner or reviewer. "Technical content owned by the deployment team. Ask them before changing the steps." Change coming. "This behaviour changes in the next product release. Update it then, not before." Approved wording. "Legal approved the security and data handling wording. Don't reword it without sign-off." Deliberate choices. "Kept separate from the related troubleshooting article on purpose. Don't merge them." Or: "Restricted to a specific audience because it covers a paid feature." Known gaps. "Doesn't cover the sandbox case yet. A customer conversation shows the gap."Why it matters:More people are editing articles now, including through Fin's content recommendations, and Fin answers from these articles directly. An edit made without this context can break a link in the product or a procedure, or undo a decision someone made on purpose. Nobody finds out until a customer gets the wrong answer.What we'd like:An internal notes field on each article that teammates see in the editor and that is never published or used by Fin to answer customers. It would help further if the notes showed as a warning when someone edits a protected article, and if the API could read and write them so our own tooling can keep them current.
It would be incredibly helpful if Fin AI’s Content & Source Guidance allowed us to reference entire Help Center collections (or even tags), instead of requiring individual article links.Right now, our Help Center is structured so that multiple articles live under the same topic within a collection. When configuring Fin AI guidance, we have to manually add each article one by one. This creates a few challenges:There’s a character limit in the guidance field, which makes it difficult to include all relevant articles It’s time-consuming to maintain, especially as content grows It’s not scalable, every time we add or remove an article, we need to manually update the guidanceBeing able to point Fin AI to a collection (or dynamically include articles via tags) would make this much more efficient and future-proof. It would ensure guidance stays up to date without constant manual intervention.This would be a big quality-of-life improvement for teams managing larger, structured Help Centers.
Please allow us the ability to toggle off Fin replies for a specific ticket where we do not want Fin to be a part of the conversation.
New Support Stack episode with @Dawn, AI Knowledge Manager at Fin. She's turned that vague instruction into a concrete 14-factor checklist you can run against any help article — the things that break AI retrieval and human comprehension at the same time (ambiguous references, missing alt text, steps buried in prose, headings that don't match how customers actually ask). Then she runs the whole check live as a single Operator prompt: what passes, what fails, and exactly what to rewrite.Watch Dawn’s second episode on Support Stack here!
Actualmente Fin responde en cuanto detecta un mensaje, incluso si el usuario todavía está escribiendo varios mensajes seguidos. Esto puede causar respuestas incompletas o confusas cuando el usuario aún no ha terminado su pregunta. Sería útil que Fin detecte cuando el usuario sigue escribiendo y espere unos segundos (por ejemplo 30–60 segundos) antes de responder. Esto mejoraría la experiencia en conversaciones donde los usuarios envían mensajes en varias partes.
Our team would benefit from a simple integration between Intercom and RescueMe/LogMeIn (and other remote tools) that allows support agents to initiate remote assistance directly from a conversation. The goal is to generate a one‑click session link for the customer within Intercom, so they can start a remote session without leaving the Messenger or conversation view. This would streamline troubleshooting and reduce friction for customers who need guided assistance. A lightweight API-based integration that generates and returns a session URL to Intercom would enable this workflow.
Product idea: "Always Surface" markers for AI-highlighted article sectionsThe problemWhen AI assistants answer questions from articles, they pick which parts of the content to use. Authors have no control over this. A critical section (a warning, a prerequisite, a pricing caveat, a key step) can get skipped or buried, even though the author knows it must always reach the reader.The ideaLet authors mark a specific section of an article as "always highlight." Whenever AI draws on that article, it must include or prominently surface the marked section in its answer, instead of treating it like any other text.How it could workIn the editor: the author selects a section and applies a flag, such as a "Highlight for AI" toggle, a callout block, or a tag. Behind the scenes: the marker is stored as metadata on that section, so retrieval and answer generation can see it. In the AI's answer: if the article is used as a source, the flagged section is always quoted, summarized, or linked, and visually distinguished (for example, "Important:"). Optional controls: a priority level (must include vs. should include), and a trigger scope (always, or only for certain topics or questions).Who benefitsAuthors and content teams get control over what's non-negotiable. End users get critical information instead of a partial answer. Support and compliance teams get fewer errors caused by missed warnings or caveats.ExampleA help article on connecting a data source includes the section "You must be an admin to complete this step." Flagged as always-highlight, the AI includes that line whenever it answers a question about connecting a data source, so users don't hit a dead end.
Right now, Fin AI usage limits (alerts and hard limits) are configured at the workspace level. This means all users — regardless of plan or segment — share the same usage budget. For teams rolling Fin out to different audiences (for example free trial users vs paying customers), it would be very helpful to set separate usage limits per audience or segment.Example use case:Free trial users: lower monthly Fin outcome limit Paid customers: higher monthly Fin outcome limitThis would allow teams to safely enable Fin for larger audiences without all users competing for the same usage budget.
Currently, any ticket created through the Fin Slack agent will receive a “created_at” timestamp corresponding to when the first message was sent in a thread, instead of when the actual ticket was created in Fin.For example, our customer sent a message on 24 September, but we didn’t initiate a support ticket for it until 1 October. Now it looks like we didn’t respond to them for 8 days, and it skews our metrics (time to first response, resolution time, ...)Can you change this behaviour to set the “created_at” timestamp to when a ticket is actually created in Fin?We can’t be the only customers who have a problem with this.
Today Intercom allow users to sync up the public and the internal articles with a Notion database. The issue is that most Notion database contains outdated articles. This leads Copilot and Fin AI to provide wrong advises.Notion provide a native feature to encourage the verification of articles: https://www.notion.com/help/guides/verify-knowledge-your-teammates-can-trust-with-page-verification Then verified articles holds a badge ‘Verified’. This is then very easy using Notion API to filter verified articles only. Having the possibility to automate Intercom articles on verified Notion page would be a great feature to improve the answers of Intercom AI.Note that this is possible to select manually articles from a Notion database but the maintenance of the page to synchronize is way to costly for medium / large companies.
Configurable retention periods for Fin Memory (currently fixed at 45 days for messages/summaries, 180 days for facts), requested by a customer
Over time I would find it helpful if there could be a separate function that allows us to tailor our email preferences more specifically. For example, it would be useful to have the option to distinguish between tickets/conversations and the more general Finn AI updates when using the platform. Screenshot below:
Auto-flag Help Center articles that go stale when we ship product updates, and suggest the edits needed. This would keep Fin's answers accurate and the knowledge base updated without manual audits/new fin guidance every release.How it would work: we input release notes into Intercom; Intercom surfaces impacted articles and highlights recommendations for sections needing edits and/or net new content.Example Scenario: A feature is changed or removed in our product UI. We paste/upload release notes. Intercom surfaces the exact articles impacted and highlights the recommended edits or new articles. It also ensures that Fin is not surfacing old conversation answers that used to be true in the past. This reduces the risk of outdated content and keeps self-serve accurate.Why it matters: Reduces outdated self-serve content, improves Fin resolution rates (meaning more $ across all customers using fin), and scales KB maintenance as release velocity increases. I imagine even Intercom could leverage this internally for their own knowledge base.
We’re currently building an internal wiki/information hub in Superhuman Docs. It’s build to be a knowledge hub and team wiki, so it would be very useful to give FIN access to a specific Doc that we create for FIN. A useful feature in Superhuman docs is the ability to sync pages between Docs. This would allow us to sync pages from our private knowledge wiki into a wiki Doc that we give FIN access to.
We use Fin in the Inbox, but we repeatedly run into the same problem: Fin sometimes sends very long messages and occasionally incorrect or irrelevant information.At the moment, Fin only reliably stops once an agent actively replies in the conversation.This is a real issue in day-to-day operations because when I’m on the go, I often can’t reply immediately — but I still need to stop Fin instantly as soon as I see that it’s going in the wrong direction.Feature request: “Stop Fin without replying”I’m requesting a way to immediately pause or stop Fin without requiring an agent to send a message.Clear trigger examplesFin should automatically stop when any of the following happens: The conversation is assigned to an agent An agent takes over / opens the conversation A manual toggle in the conversation header: “Pause Fin” / “Stop AI replies” (one click) Controllable on mobile (e.g. via the Intercom mobile app while on the go) Why this matters Prevents incorrect information from being sent to customers (quality & brand risk) Avoids unnecessary long AI messages in the chat (customer experience) Allows a safe handover to a human even when mobile, without having to send a “placeholder” reply just to stop Fin Acceptance criteria When a conversation is assigned to an agent, Fin stops immediately (no further messages or follow-ups). Agents can pause/stop Fin without sending a reply (UI control). The feature is available on mobile (at least stop/resume). Optional: audit log (e.g. “Fin paused by Gustav via assignment/manual toggle”).
Right now, ticket attributes can only be edited in the web Inbox sidebar before closing a ticket. The mobile Conversations app doesn’t show this section, so attributes can’t be filled in on mobile.This creates a gap for teams that work primarily from their phones. It would be very helpful if the mobile app allowed teammates to view and set ticket attributes directly before resolving a conversation.This would make mobile workflows much smoother and reduce the need to switch to desktop just to complete required ticket data.
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.