---
title: Personalized Responses
deprecated: false
hidden: false
metadata:
  robots: index
---
<Callout intent="danger">
  Personalized Responses requires that the AI Assistant Reasoner is at GPT-5; GPT-4.1 is not supported.
</Callout>

# Overview

Personalized Responses tailors answers from the AI Assistant to each employee—improving relevance, clarity, and trust - based on the content (Files, KBAs, etc) retrieved. By bringing key user attributes, including any custom attributes, into working memory, the AI Assistant can provide a more tailored response.

<Callout intent="warning">
  **Limitations**

  * Only impacts the Knowledge Search / Enterprise Search Plugin within the AI Assistant
  * Doesn’t affect what content is retrieved—it simply personalizes the response based on the retrieved content, unless attributes are used for query rephrasing.
</Callout>

![](https://files.readme.io/2a75feff50dc81eb58a6b1e45326a8db689689253875fb30fc4e8841d8467be7-CleanShot_2025-10-31_at_14.45.01.png)

# How does it work?

By default, the Moveworks AI Assistant will leverage some of the **native** attributes that are already part of the identity integration to create a Personalized Response:

* Location
* Country code
* Role
* Department

In addition, admins can configure **custom** attributes to also be leveraged as part of Personalized Responses. Here are example categories of **custom** attributes that can be helpful.

<Table align={["left","left","left"]}>
  <thead>
    <tr>
      <th>
        **Category**
      </th>

      <th>
        **Examples**
      </th>

      <th>
        **Primary Use Case in Search**
      </th>
    </tr>
  </thead>

  <tbody>
    <tr>
      <td>
        **Organizational & Role**
      </td>

      <td>
        -Job Grade  
        -Business Unit  
        -Employee Level
      </td>

      <td>
        Content relevance (e.g., sales vs. engineering topics) within the retrieved body of text
      </td>
    </tr>

    <tr>
      <td>
        **Employment & Status**
      </td>

      <td>
        -Salary Basis Code  
        -Years of Service  
        -Employment Type (Full-time, Intern, Contingent)
      </td>

      <td>
        Specific answers for HR, payroll, and benefits (e.g., "Your specific vacation accrual is...") if written within the retrieved body of text
      </td>
    </tr>

    <tr>
      <td>
        **Geo Location**
      </td>

      <td>
        -Subsidiary  
        -Building  
        -Work location (Office or Remote)
      </td>

      <td>
        Location-specific info (e.g., local holidays, office facilities, regional policies) if written within the retrieved body of text
      </td>
    </tr>

    <tr>
      <td>
        **Device Profile**
      </td>

      <td>
        -Laptop OS (Mac or Windows)  
        -MDM Status (Enrolled vs Not Enrolled)
      </td>

      <td>
        Targeted IT support and accurate "how-to" guides (e.g., "Here is the VPN guide for _macOS_...") if written within the retrieved body of text
      </td>
    </tr>
  </tbody>
</Table>

The combination of both **native** and **custom** attributes are injected to the system prompt, which is part of the working memory. When the AI Assistant provides a summary, based on retrieved content, it will tailor the response with these attributes.

# How to Configure it?

1. Ensure that the desired native and custom user attributes are mapped into the user proto (In `Moveworks Setup` > `User Identity` > `Import Users` )

   Ensure user attributes are mapped to the user proto will be automatically used for personalization, as part of GPT-5. See [/docs/user-attribute-reference#/](https://help.moveworks.com/docs/user-attribute-reference) for full list of attributes that _should_ be mapped. If the admin would like to leverage location or any general attribute from a non-primary source, they will have to do an override the native attribute.

   ![](https://files.readme.io/87d87066c1f054b8437e97034563f9cc0a33e9e8a6aa39973df91e904ba4ed57-UPDATE-CleanShot_2025-10-14_at_20.33.29.png)

   Additionally, **custom** attributes can be ingested via Content Gateway. See `/users` within the [API Spec](https://help.moveworks.com/reference/get_users-1).

   ![](https://files.readme.io/92150800a61acb09673225a9f679c95b535fc7bdffb975735cfa64887f2b7f9f-UPLOAD-CleanShot_2025-10-14_at_20.31.41_1.png)

2. Map the _**desired**_ user attribute keys to friendly names and descriptive labels that will be provided to the Assistant (In `Moveworks Setup` > `AI Assistant` > `Advanced Setting` > `User Attribute Personalization`)

   Identify a subset of attributes that will be most beneficial for personalization

   1. ![](https://files.readme.io/83fa1d75c7f6e8388932ca48c56acacfc57e7e3c8011c612d9fb73000b5a8bcb-CleanShot_2025-11-13_at_09.59.19.png)

      <br />

3. To confirm the attributes have been ingested, visit `Moveworks Setup` > `Core Platform` > `Imported Users` to validate that the attribute has been ingested in the “All Users” view.