General Assistant (Next-Gen)

A Next-Gen General Assistant is an advanced tool-based assistant, designed to let teams experiment with new tools, features, and infrastructure in a safe, isolated way. It’s essentially a sandboxed assistant system: you can attach experimental tools, run workflows, and test integrations without risking disruption to the assistants or flows already running in production.

You can create as many Next-Gen bots as needed for your teams and configure their settings, tools, and data sources so that the responses they receive are tailored to the team’s specific needs.

Note: This feature is still in beta. Before getting started, please ensure that your workspace admin has given permission to users to create next-gen assistants.

Step 1 | Creating a General Assistant (Next-Gen)

To create a general assistant:

• Click on  Control Center.

• Click on  Assistants.

• Click +New Assistant.

• Click General Assistant (Next-Gen).

Note: If you're creating a next-gen assistant to support the chat page, you'll need to share it with the relevant users you want it to assist. Learn more about sharing assistants.

Step 2 | Configuring the Assistant Model

Set your assistant’s LLM. Currently, the Next-Gen assistant supports GPT-5 with a default reasoning level of medium.

The Reasoning Level can be changed to:

• Low

• Medium

• High

Step 3 | Configuring the Assistant Tools

Add tools such as web search or enterprise search to enable the assistant to find and compile answers from across the web and within your company’s documents.

The available tools are:

• Web Search

• File Search

• Enterprise Search

• Custom Tool (via MCP/API)

CustomTools (via MCP/API) are coming soon. If you would like this feature already added to your Unleash workspace please contact your Unleash rep!

Web Search

When asking a question that requires up-to-date, time-sensitive information not included in the LLM's built-in knowledge (e.g., today's weather or the current exchange rate), the LLM can provide web-based results.

File Search

This feature enables you to upload files directly into your LLM Assistant’s private knowledge base, allowing it to generate more accurate, relevant, and organization-specific answers, which go beyond its general training data. When you select file search, a section will open allowing you to drag & drop or upload files.

What happens when you upload files?

• Once uploaded, your documents become part of the Assistant’s private knowledge base.

• The Assistant can reference, summarize, and answer questions directly based on your uploaded content.

• You can preview or delete uploaded documents at any time.

How does the Assistant use uploaded files?

• When you ask a question, the Assistant first searches your uploaded knowledge base.

• If it finds a match, the AI will answer only using that content.

• If no match is found, it will fall back to:

  • Web search (if enabled)

  • Its general training data

File upload requirements

• Upload Limit: Up to 20 files per Assistant.

• Maximum File Size: 50 MB per file and 20MB for images.

• Permissions: Only users with Full Access to the Assistant can upload documents.

Enterprise Search

You can configure the assistant to draw on your company’s knowledge when answering questions.

To configure the assistant’s enterprise search sources:

• Click on + Add App.

• Select the relevant app you want to connect.

  • If you have multiple links connected for an app, a popup of the links you have access to will appear. Click on the relevant account.

• To add resources from Unleash, select Unleash.

• Under Wikis, choose the wiki/s you would like to include.

• You can add Unleash again as an app and select additional wikis.

• Narrow the scope of the resources the bot will use by clicking on +More Filters (optional).

• As you apply filters, the system dynamically updates the available options, ensuring that new filters remain within the scope of your current selection. Three constants are: Specific Resource, Last Edited and Type.

  Specific resource:

  • Search for a resource and click the checkbox next to the relevant resource to check specific resources as data sources.

  • You can select as many as you want.

  • Once done, click Select.

  Last edited: Establish specific timeframes for the bot to draw answers from when responding to a question.

  • Choose a time filter or scroll to the bottom and select Custom.

  • Select a time from Before, After, or Between.

  • Select the dates.

  • To save, click out.

  Type:

  • Select the type of resource available for that app.

• To add another app, click +Add App (optional).

  • You can activate knowledge prioritization when you have more than two app sources configured. The assistant will seek answers within the prioritized resources you set. If no answer is found, it will expand its search to other connected resources.

  To activate knowledge prioritization:

  • Click Knowledge Prioritization on the sources to prioritize (you can prioritize more than one).

• Click Save.

Link Indications

• If any card contains a file, the Assistant will scan and utilize those files to provide more accurate answers

• Private wikis and links that are set as a data source will be visible as answers delivered by the bot in channels. Private wikis and links are identifed by this icon: 

• Links that are still syncing to Unleash are displayed with this syncing icon: 

• If the selected link fails to sync, an indication will appear below the app, prompting you to try reconnecting the link on the Unleash app connection page.

• If the link was removed, an indication will appear below the app.

• When a link is selected within 24 hours of its initial connection, a message informing you that full AI functionality will be available within 24 hours of the connection time will be displayed.

Step 4 | Configuring the Assistant Instructions

Assistant instructions let you personalize how your assistant behaves, responds, and retrieves information. By entering custom free-text instructions (up to 32,000 characters), you can define tone, behavior, data sources, and other response preferences to align the assistant with your team’s needs. For best practices, please see this article.

You can provide instructions to tailor:

• Trigger - Specify which tools the assistant should use when responding, for example:

  • Enterprise Search: Use for internal company data such as Confluence, Google Drive, or Slack.

  • File Search: Use for uploaded or shared files.

  • Web Search: Use when information is not available in your enterprise or uploaded sources.

  • Code Interpreter: Use for calculations, data analysis, or file-based problem solving.

• Tone (e.g., formal, casual, assertive, confident).

• Response Length (e.g., concise, moderate, detailed).

• Role and Behavior Guidelines (e.g., acting as a professional customer support representative or an expert in a specific field).

Example: You are a helpful assistant for our internal support team. Always respond in a concise, professional tone. For numbers or calculations, use the Code Interpreter (within the File Search tool). For uploaded files, search them first and cite the source. For company knowledge (e.g., Confluence, Drive, Slack), use the Enterprise Search tool. If no relevant information is found, state that clearly instead of guessing.

Step 5 | Configuring the Assistant Prompt Templates

Prompt Templates

Add prompt templates to your assistant to help workspace members understand the types of questions they can ask the assistant in the chat. These prompts serve as straightforward, practical examples that guide users and encourage engagement. For example:

• What is our parental leave policy and how does it compare to EU standards?

When a member clicks a prompt template card from the chat’s empty state, the prompt text is inserted into the input field for easy editing before sending. Learn more about using prompt templates in the chat.

To create a prompt template:

• Click inside the prompt field and type your message.

• Use brackets to mark editable variables (e.g., Ask about [topic] or [project name]?).

• Prompts are saved automatically when you click outside the field.

• To edit a prompt, simply click on it.

• Click + Add to add another prompt.

Note: Each assistant can have up to four prompt templates only.

Step 6 | Configuring the Assistant Display Preferences

There are two essential toggles to optimize how users interact with the next-gen General Assistant. By default, these toggles are off.

Pin to Chat Sidebar: Displays the assistant in the chat sidebar for users with interact permissions, allowing them to start a new chat with the assistant. Only users with Full Access can unpin this assistant.

Display in Quick Search: Integrates the assistant into the quick search experience. When typing a query into the homepage search bar or launch bar, a dropdown will appear, enabling users to select the chat assistant to answer their questions or receive answers from the Unleash Workspace Assistant. Learn more here.