Fin Tips and Best Practices

📊  When Your Data Doesn’t Tell the TruthHow Fin’s AI Categorization Turned Messy Data into Clarity Every CX or ops leader knows the feeling. You pull up a report, and at first glance the numbers look fine. But something feels off. The story in the charts doesn’t line up with the reality on the floor. The problem isn’t the report—it’s the tags feeding it. When tagging is inconsistent, the data turns unreliable. And unreliable data makes reports lie. That was the bind this team found itself in. They needed to spot product issues early—like which models were most likely to crack, which parts kept showing up damaged or missing, or which shippers were the least reliable. With that kind of clarity, they could protect revenue, get ahead of problems, and spare customers unnecessary pain. But with inconsistent tags, they weren’t seeing the full picture. The pivot: Stop Wrestling with Tags. Start Tracking Reality 🔁 Here’s the move: instead of drafting yet another tagging framework with sprawli

Hey everyone — curious to learn from other teams here!We’re building out a Case Closure Data Model as part of our Intercom + Fin AI rollout, with the goal of turning every resolved ticket into AI-ready data for automation, prevention, and knowledge training.Our current thinking includes adding structured closure fields like: Technology / Sub-Technology Problem Code / Resolution Action / Root Cause Repeatability / Escalation / Customer Effort KB Linked The intent is to improve data quality, reduce “Other” categories, and help Fin identify repeatable automation candidates over time.I’d love to hear how others are approaching case hygiene at closure: Which closure fields have been most valuable for your AI or reporting goals? How do you balance structured data collection vs. agent friction? Have you seen any creative ways to tie closure data to Fin training or KB improvement? Any best practices, examples, or lessons learned would be hugely appreciated.— Lance 

This Fin Task was designed to help customers quickly understand whether an issue they’re experiencing is related to a known incident, without needing to contact support directly. The goal was to make it feel fast, human, and reassuring while surfacing relevant technical information from our infrastructure. 🌍 Step 1: Detect the Customer’s Hosting RegionWe start by detecting the customer’s data hosting region, which is stored as a company-level attribute in Intercom. This attribute is critical — it allows the task to tailor its response and status checks to where the customer's workspace is hosted.Regions we support include:US   Europe   Australia  If the region attribute is missing or unrecognized, the task defaults to US, and clearly explains this fallback behavior in the customer message (without speculating or misleading). 🔌 Step 2: Query the Incidents API via Region-Specific Data ConnectorsFor each supported region, we’ve set up a dedicated Fin Data Connector that calls our intern

Recently we made a small change with a big impact in our Intercom workspace. We’ve now enabled Fin to re-engage customers if their conversation hasn’t been picked up yet. After a set wait time, Fin checks back in: “While you're waiting, would you like to chat with me (Fin AI Agent)?” If the customer says yes, Fin will attempt to resolve the customers issue. If not, the conversation stays right where it is in the team queue. Why this matters?Because it’s a game-changer for speed and efficiency. We’re seeing more conversations resolved automatically, fewer customers waiting around, and less work piling up for the team. How to enable it in your workspace: Go to your Workflows page Create a new workflow using the If teammate has been unresponsive trigger In your audience rules select the team inbox(es) you want this to fire in, and the amount of time the Workflow should wait before firing (e.g. teammate has been unresponsive for 1 hour, this means the conversation has been waiting an h

When using collapsible sections in your articles, consider repeating the core question of your FAQ (Frequently Asked Question) within the answer itself.  This significantly helps AI agents like Fin more accurately find and utilize the relevant content for their answers. It's a handy trick to optimize your FAQ sections for better AI performance!

You can use AI tools like Claude or ChatGPT to: Repurpose internal content: Quickly convert internal documents and product updates into customer-friendly Help Center articles. Review and rewrite: Rephrase content to match customer language. Ensure proper formatting (headers, bullet points, etc.). Clarify ambiguous terms.  Using AI tools works really well if you create a style guide with the structure, formatting, tone of voice, and template you'd like AI to use when generating articles for your Help Center. Once you’ve got a style guide, try using it with these AI prompts:New content creation You are an expert technical writer. Goal: Transform the attached “[name of internal document]” into a polished, customer-friendly Help Center article that strictly follows the attached “[name of Style Guide]”. The finished piece must be easy for humans to read and for AI systems to parse and extract structured information. Use only facts present in the internal doc. Do not inve

One of the easiest ways to improve how Fin handles conversations is to use if–then logic in your Guidance. It helps Fin understand exactly when to apply a response, and what that response should look like.Here’s a simple example:"If a customer asks how to change their password, tell them to click 'Forgot password?' on the login screen and follow the steps."That’s much clearer than:"Explain the password reset process."By adding a condition (“if a customer asks…”) and a specific action, you give Fin a clear playbook to follow. It also helps prevent Fin from jumping in with the wrong response in unrelated situations.To take it further, you can include a direct link to your help center or a relevant support article. This makes it even easier for Fin to guide customers to the right place, without needing to explain every detail.

On the surface, it can seem that Copilot is a simple yet powerful tool that helps your team answer customers questions. But, it’s so much more than that!We've been using Intercom’s Copilot across our team and it's been a game-changer for us. If you're not using it to its full potential yet, here’s a quick breakdown of some ways to use Copilot you may not have thought of that help streamline work and boost efficiency:Faster Onboarding for New AgentsNew joiners no longer need to ping senior teammates every five minutes. Copilot acts like a super-smart buddy, guiding them with internal resources and examples from previous chats. It’s drastically shortened our onboarding ramp time.Internal Process GuidanceWhenever someone has a “how do we do this again?” moment, Copilot jumps in with answers sourced from our internal articles and documentation. It’s perfect for navigating complex or frequently changing internal workflows.Major Time SavingsBy aggregating info from external knowledge bases (

Since it was in early beta we have been using the AI Summaries feature in our Intercom workspace and it’s proven to be a massive time-saver for our support team. Before this, whenever a teammate picked up a conversation after Fin handled it, they had to scroll through everything to understand what had already been covered. Now, Intercom generates a quick, clean summary at the point of handoff from Fin in an internal note.It’s a small change with big impact—no more duplicated steps, no more “Let me check what’s already been said,” and a much smoother handoff between Fin and our human teammates.🛠 How to enable it in your workspace: Go to your Workflows page Go to one of your existing triage workflows or create a new workflow In the path (or before) you are assigning to an inbox, add the “Add summary note” action Done! The summaries will automatically appear as internal notes at this point in the workflow so when a teammate picks up the conversation, they have all the context they ne

Here’s something that’s easy to overlook: Fin isn’t reading between the lines. It’s taking your guidance literally, so the simpler and more specific your language, the better the results.We’ve all written something like:"Be friendly and professional."It sounds good, but it’s open to interpretation.A clearer version might be:"Use a professional yet warm tone. Keep responses short and avoid jargon. When a customer is frustrated, acknowledge the issue and say something like, ‘I understand how that can be frustrating. Here's what we can do.’"This type of Guidance gives Fin a clear tone to follow and reduces guesswork. It also helps keep responses consistent across your support experience.Think of it like coaching a new team member. If you wouldn’t rely on vague tone tips when onboarding a new agent, don’t use them with Fin either.

We noticed a recurring issue in some of our Fin-involved conversations. Fin was surfacing Help Center content that included phrases such as "please contact support" or "reach out to us through the Messenger."While this wording makes sense when someone is reading an article independently, it can cause confusion during a live support interaction. When Fin presents this type of content mid-conversation, it can give the impression that the customer is not already receiving support. This creates friction and can lead to drop-off.To improve the experience, we provided more specific guidance for Fin to follow. We instructed it not to include phrasing that directs customers to contact support, since they are already in a support conversation. Instead, Fin should treat the interaction as part of the support experience by either providing the information the customer needs or asking for clarification to help resolve the issue.This change reduced unnecessary loops and made Fin feel more integrate

A lot of us naturally default to vague guidance like,"Make sure Fin understands our product types."But Fin thrives on clarity. A stronger version might be:"If a customer asks about the 'search' feature, first ask which product they’re using. Then give product-specific instructions."See the difference? That second version gives Fin clear steps and a condition under which to act. It's specific, it's actionable, and it’s something Fin can consistently follow.When you're writing guidance, ask yourself: what should Fin say first? What should happen next? That simple shift in thinking can make your guidance 10x more effective.How do you typically plan out what Fin should do in different scenarios?

Tip: One of my new weekly rituals is using the AI-powered Suggestions feature to improve our knowledge base - and it's incredibly satisfying.Before this feature, most content improvements happened when a teammate noticed a problem, took the time to submit a Back-office ticket, and someone else followed up to make changes. That worked... when it happened. But it often didn’t.Now, Suggestions does the heavy lifting. It automatically looks at where Fin may have stumbled in a real conversation and recommends a fix (like editing an article or adding a snippet). I review them once a week while waking up with a coffee, and approve about 70% with just a few clicks.After you’ve deployed Fin, head to Fin AI Agent > Train > Suggestions. You’ll see AI-generated recommendations for improving your content, with links to the original conversations so you can double check the context. It’s honestly one of the fastest ways to boost Fin’s performance without chasing down feedback manually!

Currently when you add the “Notify Slack Channel” action to a Workflow step, it doesn’t allow you to customize the notification which isn’t to great when you want to send a specific message or include data pertinent to the user. Well... with the use a Data Connector, you can create your own custom Slack notification!  1. Set Up a Webhook in SlackFirst, create an incoming webhook in Slack:Go to your Slack workspace and click on your workspace name Select "Settings & administration" → "Manage apps" Search for "Incoming WebHooks" and click "Add to Slack" Select the channel where you want messages to appear Click "Add Incoming WebHooks integration" Copy the Webhook URL (you'll need this later) Optionally customize the name and icon for your webhook2. Create the Data Connector in IntercomNow set up the connector in Intercom:In Intercom, go to "Settings" → "Data Connectors" Click "Create new data connector" Choose "Custom" as the connector type Name your connector (e.g., "Slack Notificat

Tip: Product releases can significantly affect what Fin needs to know to respond accurately to customers. Here’s a simple process to keep your content aligned and Fin informed: Create a tag: Start by creating a tag like "[Feature name] content updates". Search & review: Search the Knowledge Hub for content related to the release. Open top results and assess if they need updates. Apply the tag to anything that does. Filter & collaborate: Filter by your tag to create a list of all impacted content. Share this view with teammates to divide and conquer updates. Prepare drafts: For public articles, save updates as drafts ahead of the release. Publish in one go: When it’s time to ship, filter by your tag + status “Published with pending draft”, then bulk publish to go live in one click ⚡️ Fin will instantly have the most accurate, up-to-date content to handle any inbound conversations confidently and correctly.

Tip: Use simple, specific language that mirrors the words your customers actually use. Avoid technical jargon or internal terms—Fin works best when articles match how users naturally phrase their questions. For example, say “log in” instead of “authenticate” to improve understanding and AI searchability.You can red more about it here.

Tip: Enhance the clarity of your help articles by restating user questions within the answers. Instead of a simple “Yes” or “No,” provide full-sentence responses that mirror the user’s query. This approach not only aids AI comprehension but also ensures users receive precise and context-rich information. For instance, if a user asks, “Do I need to use the original shipping box to return an item?” Rather than replying with just “No,” elaborate with, “You can use your own packaging when sending an item back for a return or exchange; there’s no need to save the original shipping box.”  By adopting this method, you make your support content more accessible and effective for both AI agents and human readers. You can read more about it here.

Cross-post from other group, but…...Loads of new features, some AI capabilities complimenting/replacing legacy features and many terminology updates. The Intercom glossary is being updated daily, but after some recent Intercom partner training, felt it worthwhile to post some of our notes on the Intercom Lexicon for those that are new to the platform or about to explore Fin. Check it out here: https://www.journeymapper.net/blog/intercom-glossary-101