Personalize Fin's answers with the Data Hub
Give Fin access to the customer data your team already sees so it can tailor every answer.
Your support team never answers a question without context. They can access an extensive customer profile that includes the customer’s plan, account history, permissions, and more, tailoring their reply in seconds. So how do you give Fin that same context?
This lesson is part of the Optimize Fin course.
Prerequisites
To follow along with this lesson, you’ll need:
- Identity verification enabled: Fin only references customer data in conversations where the customer is identity-verified. Set this up under Settings > Security. Learn more about identity verification.
- Customer data in Intercom: People and company attributes need to be in your workspace already, whether from the Messenger, an integration like Salesforce or Stripe, a CSV import, or the API.
Note: The Data Hub is in open beta and available now under Fin AI Agent > Train > Data. Functionality may change as the product develops.
Video transcript
The more Fin knows, the more it can resolve. Often that comes down to the customer data Fin uses to give personalized answers. You might already have data connectors or procedures set up, which help Fin reach out to your external systems to check data and act on it. But what about all that user data already sitting in your workspace? Well, that’s where the Data Hub comes in.
This is all the data your support agents can see about the customer they’re speaking with. But by default, Fin can’t see all of this unless you enable it to do so. For example, it would be useful for Fin to know seats used so we can answer questions like, why can’t I add a new user? Monthly spend so it knows what perks it should offer the customer and their plan so it can tailor every answer to the features they can actually access.
So picture a customer asking how to set up a specific feature like custom reports. Fin doesn’t know they’re on the Essentials plan, which doesn’t include this feature. So instead, it sends them on a wild goose chase. When the customer realizes they can’t follow along, they get frustrated, ask for a handover, and it takes a teammate to tell them they never had access in the first place. Fin could have delivered the perfect, personalized answer if only it had access to that data.
Let’s enable Fin to handle questions like this in the future. We’ll go to Train and Data, where we manage what data Fin can access. And here’s the attribute that holds the customer’s plan. Here, we can see that it’s disabled for Fin, so Fin doesn’t currently know this valuable context when talking to a customer. We’ll set it to test so that our Fin preview can use it. And let’s ask that question again. Great. Fin tells them right away they don’t have access to that feature. And if we click into Show events here, we can see that Fin used that exact attribute to inform its answer.
Okay, so Fin can now use that data to help customers. But Basic isn’t the name of our entry level plan anymore. Now we call it Essentials. Fin’s using the old name because that’s the value held in the attribute. So let’s tell Fin how to talk about this data with customers. In data transformation, we can map the value basic to the term we want Fin to use instead: Essentials. Now, Fin will talk about our plans just like a teammate would.
Let’s look at a few more key settings here. This checkbox lets Fin use the attribute to shape its answers, while avoiding mentioning it to the customer. One example is monthly spend. It’s useful context for Fin, but it’s best to avoid quoting exact figures back to the customer. Write a clear description of what the attribute contains along with any other useful guidance for Fin here so it knows how and when to use it. We won’t be too prescriptive here. It’s best to give Fin some room to judge when this data can be useful. Hit enable, and this attribute becomes part of the profile Fin holds for each customer.
Ask your support team about the questions they pick up from Fin that they answer in seconds just by glancing at the customer’s profile. And think about which data Fin needs access to, to resolve those by itself. When Fin has the data it needs to provide personalized answers, it resolves more, it’s a better experience for your customers, and it means more time back for your team.
What data you can manage in the Data Hub
The Data Hub (Fin AI Agent > Train > Data) is the single place to see every customer attribute stored in Intercom and decide which ones Fin can use. It currently covers people and company data. For conversation-level or custom object data, use Procedures or data connectors for now.
| Data type | What it describes | Example attributes | In the Data Hub? |
|---|---|---|---|
| People | Users and leads talking to your team | Name, Email, Seats used | Yes |
| Company | The businesses your customers belong to | Company name, Plan, Account region | Yes |
| Conversation | The conversation or ticket being worked on | Rating, Channel | Not yet |
| Custom objects | Objects you define yourself | Orders, Subscriptions | Not yet |
Each attribute shows its source, format, coverage (the share of active contacts with a value), and freshness (when it was last updated). Use these to decide whether the data is reliable enough for Fin to lean on.
Giving Fin access to an attribute
Every attribute has three possible states for Fin: Disabled, Test, or Enabled. Move through them in order so you can check Fin’s behavior before customers see it.
| Fin status | What it means |
|---|---|
| Disabled | Fin can’t reference the attribute. This is the default for most attributes. |
| Test | Fin can use the attribute in preview conversations only. Use this to check how Fin responds before going live. |
| Enabled | Fin can reference the attribute in live customer conversations. |
Fin already has access to five standard attributes by default: name, email, city, country, and browser locale. These appear as Enabled and don’t need to be switched on.
Once you open an attribute, there are additional settings that shape how Fin uses it:
- Description: Helps Fin to know what the data is for.
- Data transformation: Map raw values to the terms you want Fin to use. If your attribute still says “Basic” but the plan is now called “Essentials,” this is where you fix it.
- Audience: Limit an attribute to a specific audience, so Fin only uses it for the customers it applies to, such as region-specific data.
- Hide from customer: Fin can use the value to shape its answer without quoting the value to the customer. Ideal for sensitive figures like monthly spend or account health scores.
- Other guidance: When and how Fin should use this data. Don't try to account for every possibility. Instead, give Fin some freedom to decide when to use this data.
Measuring the impact
You can see exactly where customer data is changing Fin’s answers in two places.
- Report filters: Filter Fin AI Agent reports and monitors by whether Fin referenced personalized attributes in a conversation, so you can compare resolution rate and CX score with and without customer data.
- Conversation events: Open any conversation and check the event log for “Fin referenced personalized attributes” to confirm which conversations used the data.
If Fin isn’t picking up an attribute you expected it to, the usual causes are messy values Fin can’t confidently interpret, a procedure or escalation rule triggering first, or a use case complex enough to need a procedure. Sharpening the attribute’s name, description, and guidance fixes most cases.
