Introduction
The AI assistant is Docebo's conversational AI experience. It brings together multiple features that are enabled and configured independently:
- Search and content discovery— allows all platform users to ask questions, discover content, and ask about their learning status in a unified panel.
- Tutoring (Beta) — allows learners to use the AI assistant as a tutor inside courses and learning programs, with answers grounded on the learning content.
- Platform guidance— provides contextual platform guidance, step-by-step instructions, and administrative support for Superadmins and Power Users.
- Conversational insights— allows Superadmins to ask natural-language questions about learning data and receive answers powered by platform analytics.
You can also personalize how the assistant presents itself and how it responds, by setting a custom name, custom instructions and a communication style. See the chapter Customization.
The AI assistant is available in all the platform-supported languages and is continuously evolving. Keep an eye on the Product updates page for newly supported capabilities.
This article explains how to enable and configure AI assistant features. For information on how users interact with the AI assistant once it is enabled, see AI assistant.
Accessing the AI assistant configuration
You can access the AI assistant configuration page in the following ways:
- open the Launch pad page, locate the Harmony features section and click Configure next to one of the AI assistant features
- open the Configuration menu, locate the Platform setup and tools section, and select Artificial intelligence. Locate the AI assistant card and click Configure.
Tip: You can also use the search bar in the header to quickly navigate to the AI assistant configuration page by searching for Launch pad or Artificial intelligence.
Prerequisites
Doceboai.cloud domains
For the AI assistant page to load correctly, your firewall or network infrastructure must not be blocking the *.doceboai.cloud domains. Check with your IT department to ensure that *.doceboai.cloud domains are placed on the allow list.
Configuring the AI assistant
The AI assistant configuration page is organized in two areas. The Configuration area at the top contains one section per feature, where you set which features are active and how each one behaves. The Customization area below controls how the assistant presents itself and how it communicates, and applies to all AI assistant features.
To enable or disable all AI assistant features at once, click Enable all or Disable all at the top right of the page. One option or the other will be shown, depending on the features' enablement status.
Enabling and managing AI assistant features
The Configuration area contains one section per feature: Search and content discovery, Tutoring, Platform guidance, and Conversational insights. Each feature can be enabled or disabled independently using the dropdown at the top right of its section. The dropdown has three options:
- Disabled: the feature is not available to any user.
- Preview mode: only for you: the feature is enabled only for you, to test it before making it available to everyone.
- Enabled for all users: the feature is active for its intended audience (all users for Search and content discovery and Tutoring; Superadmins and Power Users with the AI assistant permission for Platform guidance; Superadmins for Conversational insights).
Some AI assistant features offer additional settings that further refine their behavior. The sections below describe each feature in detail.
Search and content discovery
Users can ask questions and get AI answers from platform content, discover related learning resources, and ask about their learning status. Set the availability of this feature using the dropdown at the top right of the card, and configure the following options:
- Allow general AI knowledge in responses: when selected, if platform content doesn't provide an answer, the AI assistant may use its general knowledge. Deselect to restrict responses to the indexed platform content only.
- Allow catalog content in responses: when selected, answers will also draw from free catalog content visible to the user. Deselect to restrict responses to the courses and learning programs in which the user is actively enrolled.
Learn more on the content used for AI assistant responses.
Tutoring
Users can ask questions, get clear explanations, and receive helpful suggestions about the course or learning program they are viewing. Set the availability of this feature using the dropdown at the top right of the card, and configure the following option:
- Enable Tutoring for all courses and learning plans: when selected, the AI tutor is automatically active on all courses and learning programs. When deselected, the AI tutor must be enabled explicitly from the Agents tab of each course or learning program.
The setting defined here acts as the default for every course and learning program on the platform. Administrators can override the default on individual e-learning courses, ILT courses, and learning programs.
Platform guidance
Superadmins and Power Users can get contextual platform guidance, step-by-step instructions, and administrative support from the AI assistant. Set the availability of this feature using the dropdown at the top right of the section.
Granting Platform guidance access to Power Users
As a Superadmin, you can grant Power Users access to Platform guidance by adding the AI assistant permission to their Power User profile. Once this permission is assigned, all Power Users associated with that profile can use Platform guidance.
For more information on how to assign permissions to Power User profiles, see Creating and managing Power Users.
Conversational insights
Superadmins can ask natural-language questions about learning data and receive immediate answers powered by platform analytics. Conversational insights covers enrollments, completions, learning programs, users, and time spent on learning. Set the availability of this feature using the dropdown at the top right of the section.
Conversational insights is available to Superadmins only. Power Users cannot access Conversational insights.
Customization
The three Customization settings are independent of each other, and each one applies to every conversation with every user on the platform.
| Setting | What it controls |
|---|---|
| Assistant name | The name the assistant uses to introduce itself and to refer to itself throughout the platform. Use it to align the assistant with your organization's branding. |
| Custom instructions | The rules the assistant always follows. This is where you set behavioral boundaries, terminology, escalation paths or content-sourcing rules that shape what the assistant does. |
| Communication style | The tone and delivery the assistant uses when responding. Use it to match the assistant's voice to your organizational culture or to the expectations of your audience. |
Please note: The communication style controls how the AI assistant sounds, while custom instructions control what it does. Keep tone requests out of the instructions field, because the communication style already covers them.
Setting a custom name
In the Assistant name area, type the name you want your users to see in the Name field. Leave the field empty to keep the default name, AI assistant.
Defining custom instructions
In the Custom instructions area, type your rules in the Instructions field. The field accepts up to 2,000 characters. The assistant applies these instructions across all conversations with all users, so review them carefully before saving.
Best practices for custom instructions
The following practices help you write instructions the assistant can apply consistently.
- Be specific, not aspirational. Write concrete rules the assistant can check itself against, not vague goals. "Direct HR questions to the HR portal" works, while "Be helpful with HR topics" does not give the assistant anything actionable.
- One rule, one sentence. Keep each instruction short and self-contained. If you need multiple rules, list them as separate sentences rather than combining them into one long instruction, as this reduces the chance the assistant drops or blends requirements.
- State the action, not just the topic. Do not just name a subject area, say what the assistant should do about it. Instead of "Be careful with legal topics", write "Never provide legal advice, direct users to the legal team instead."
- Avoid overlap with communication style. Do not duplicate tone instructions here, such as "Be friendly", as that is already covered by the style dropdown. Use this field for rules, not tone.
- Use concrete terminology. Name the exact terms, tools, or resources the assistant should reference, such as a platform name, a department name, an email address, or a URL, rather than general descriptions like "the right team".
- Set boundaries where it matters most. Prioritize instructions for topics with real consequences if handled incorrectly, such as compliance, HR, legal, safety, or brand-sensitive areas, rather than trying to cover every possible scenario.
- Test the instruction like a rule, not a suggestion. A well-written instruction should be something you can evaluate objectively: did the assistant follow it or not? If it is open to interpretation, tighten the wording.
- Keep it current. Review the custom instructions periodically, especially if the org structure, the tools, or the policies referenced in the rules change, such as a new support email address or an updated escalation process.
Examples of custom instructions
- If asked about HR policies such as vacation or leave, do not answer directly, direct the user to check with their manager or the HR portal.
- If a user asks about a technical issue you cannot resolve, tell them to contact the IT helpdesk at support@example.com.
- When "onboarding path" is mentioned, this always means the mandatory training program that new hires complete in their first 30 days.
- Acme Manufacturing produces industrial packaging equipment for the food and beverage sector. Most learners are plant technicians and line supervisors.
Choosing a communication style
In the Communication style area, use the dropdown to select the tone the assistant uses when responding to users. The available options are described below.
| Style | How the assistant responds |
|---|---|
| Default | The assistant uses its standard tone, with no additional styling applied. |
| Warm and encouraging | The assistant briefly acknowledges the user's situation before answering and uses supportive, reassuring phrasing, particularly when helping with errors or troubleshooting. |
| Friendly and conversational | The assistant uses a casual register and everyday phrasing, sounding like an approachable colleague rather than a formal support agent. |
| Thorough and explanatory | The assistant explains the reasoning behind its answers and adds a brief rationale to each step, favoring understanding over brevity. |
| Formal and polished | The assistant uses complete, formal sentences without contractions or colloquialisms, keeping a precise and professional register. |
The AI assistant on extended enterprise
On extended enterprise clients, the Artificial intelligence settings have not yet been updated to the new unified AI assistant configuration page. Each extended enterprise client still uses the previous interface, with separate Harmony search and Harmony copilot settings.
Tutoring does not have a dedicated setting for extended enterprise clients and automatically inherits the configuration set on the main platform. Administrators can still enable or disable tutoring granularly on individual courses and learning programs.
Harmony search on extended enterprise
On extended enterprise platforms, you have the option of enabling or disabling Harmony search selectively, only on specified domains.
To access the configuration:
- Select Navigation menu > Configuration (spanner icon) > Platform setup and tools > Extended enterprise management
- In the list, locate the extended enterprise client of interest and click the gear icon on its row. This will open the settings page for that client.
- In the vertical navigation, select Artificial intelligence and locate the Custom settings option.
Client without custom settings:
If a client has the Custom settings check box de-selected, then the settings of the main platform will apply. Accordingly, these clients will inherit the same Harmony search enablement (Enabled for all users, Disabled, or Preview mode) and options (Allow general AI knowledge, Allow catalog content in responses) that you have configured on the main platform.
Client with custom settings:
On the other hand, for an extended enterprise client that has Custom settings enabled, you can enable or disable Harmony search independently of the root platform. Specifically, in the Harmony search options you can choose:
-
Enable Harmony search for all users: Select this check box to enable the feature for all users on this client, even if it is disabled or in Preview mode on the main platform.
When enabled you can also set whether to Allow general AI knowledge in responses and Allow catalog content in responses on this client. - Deselect the check box Enable Harmony search for all users to disable the feature for all users on this client, even if it is enabled on the main platform.
→Note, however, that if you enabled Preview mode on the main platform, Harmony search will continue to be visible, only to you, also on this client. For more information see the chapter Preview mode and extended enterprise.
Harmony Copilot on extended enterprise
On extended enterprise platforms, you have the option of enabling or disabling Harmony copilot selectively, only on specified domains.
To do this:
- Select Navigation menu > Configuration (spanner icon) > Platform setup and tools > Extended enterprise management
- In the list, locate the extended enterprise client of interest and click the gear icon on its row. This will open the settings page for that client.
- In the vertical navigation, select Artificial intelligence and locate the Custom settings option.
- If a client has the Custom settings check box deselected, then the settings of the main platform will apply. Accordingly, these clients will inherit the same Harmony copilot settings (Enabled for all administrators, Disabled, or Preview mode) that you have configured on the main platform.
- On the other hand, for an extended enterprise client that has Custom settings enabled, you can enable or disable Harmony copilot independently of the root platform by checking or unchecking the Enable Harmony copilot for all administrators option.
Please note. If you enabled Preview mode on the main platform, Harmony copilot will continue to be visible, only to you, also on this client. For more information, see the chapter Preview mode and extended enterprise.
Preview mode and extended enterprise
Preview mode of any AI assistant feature can only be enabled on the main platform. When a feature is in Preview mode on the main platform, it is visible only to you in:
- The main platform.
- Extended enterprise clients without Custom settings for Artificial intelligence.
- Extended enterprise clients with Custom settings for Artificial intelligence and the feature disabled.
On extended enterprise clients with Custom settings for Artificial intelligence and the feature Enabled for all users, the feature will be visible to everyone on that client, even if it is in Preview mode on the main platform.
Note on AI assistant feature modes
Preview mode, where an AI assistant feature is enabled only for you, is a client-side setting stored in your browser. If you clear cache/cookies, open an incognito window, switch browsers, or sign in from a different device, the preview flag is removed. In those cases, the platform reverts the feature to its disabled state.
When a feature is enabled for all users, however, the setting is persistent and does not depend on your browser.