Introduction

Your Docebo learning platform is powered by Artificial Intelligence (AI), with several AI features included in your platform to enhance the way enterprises deploy and manage their learning programs, internally and across the extended enterprise, and is designed to deliver automated and personalized learning at scale.

All users in your platform - learners, managers, and administrators - can benefit from AI-powered features, and as with any features and functionalities in your platform, we want to ensure that you know exactly what’s going on and how to best use these innovative and time-saving features. This article outlines some questions and answers that you may have about Docebo’s artificial intelligence, as well as where you’ll find the features in your platform.

How does artificial intelligence work in my Docebo platform?

The artificial intelligence in your platform lets your platform automatically perform some of the actions that you would do manually yourself, either as a learner or an administrator, such as tagging training materials included in courses and the social learning assets that you contribute to the Discover, Coach & Share module,  or suggesting which courses are most relevant for learners.

The concept is simple: We use specialized algorithms to safely and securely analyze nearly all of the different types of content that you’re adding to your platform and how you’re using the platform, and find ways to enhance and simplify the learner or administrator experience.

What do I need to know about how Docebo uses and/or stores the data used in algorithms?

Refer to the Legal notes chapter of the article Managing artificial intelligence settings to learn more about the usage and storage of data in Docebo’s AI algorithms.

Which technology does Docebo use for the AI features? Where will these data be hosted?

We use many proprietary or third-party state of the art technologies in the field of text-to-speech, speech-to-text, text generation models, and natural language processing. We also use third-party technologies to improve or enhance our AI models, including services from AWS, Google Cloud Platform, and Microsoft Azure OpenAI. We assess and update our technologies and models periodically, including the ones required for Docebo AI features. 

Will other Docebo customers have access to my data processed by AI features?

The content and data that are analyzed and used in Docebo’s AI algorithms are used only for the purpose of enhancing the learner or administrator experience in your specific platform. Docebo does not share or sell any data used for these algorithms with Docebo partners or customers.

Where can I find AI features in my platform?

The list below is an updated outline of where you’ll find AI-powered features in your Docebo platform:

AI-powered search

Docebo’s AI analyzes learning content of any nature, understanding its content and using that content to exponentially improve learning content discoverability with the platform’s search feature. Upon sharing a new learning asset or creating new training material, AI analyzes the asset to produce search results that are more relevant. For example, a video is searched and discovered by its spoken content to produce much more effective learning experiences by putting content that the learners want in front of them more easily and more effectively. Learn more about the search engine in your platform.

Auto tagging

AI analyzes learning assets (formal and informal), identifying key phrases and creating tags for the assets automatically. This helps make content easier to find using the AI-powered search mentioned above. For example, upon sharing a  video as a new learning asset, AI “listens” to the entire video, understands its key phrases, and it generates up to 10 suggested tags for the learning asset uploader in order to help with content categorization and search. Tags can also be manually edited if necessary, and, given the stochastic nature of the language understanding algorithms used in auto tagging, human supervision of the suggested tags is recommended. It is also recommended to enable the AI retraining option to improve the AI algorithm. Learn more about auto tagging settings and about creating and sharing formal content and informal content in your platform.

Docebo’s AI needs a minimum number of words (16 words) to process to be able to compute tags, and that, for security reasons, auto tagging is not performed on private content (that is, content requiring authentication to be accessed). When uploading private content, we suggest you add tags manually. Please note that only Superadmins and experts can add tags manually when uploading an informal asset, while users can add or edit manual tags only once the asset has been published.

The auto tagging feature may encounter problems when processing SCORM and XAPI training material created through the use of certain authoring tools embedding textual contents directly into programming files, and compressing images so that their quality does not allow Optical Character Recognition (OCR). We are currently working to confront this limitation by collaborating with some of the major authoring tools providers.

Auto skill assignment

You can automate the task of assigning skills to content with the auto skill assignment feature. You can manage this feature in the artificial intelligence configuration panel. Artificial intelligence will analyze learning content as it is uploaded to your platform, identifying automatically what skills from the skills catalog are most relevant to each piece of content, and assigning them to the content. This makes adding skills to your content effortless and helps your learners find the right content to develop their skills. Please note that the skills listed in the Docebo catalog are only available in English. 

Content recommendation

There are several places within your learning platform in which the AI will provide personalized suggestions on training materials, courtesy of our content recommendation feature. Suggestions also come in a variety of flavors: below there are details on the major applications of content recommendation. More details on the principles and the criteria used by content recommendation to produce suggestions. You can disable content recommendation in the artificial intelligence configuration panel.

• Invite-to-watch: Invite-to-watch elevates the social learning experience by making it simple to share content with peers. When a learner shares an asset in Discover, Coach & Share and wants to invite a specific audience to watch that asset, AI suggests the employees who would benefit most from it. Learn more about this feature. Also note that some other considerations apply, such as the rules of visibility of users for the content being watched.
• Enrollment suggestions for administrators: Administrators will see an AI-suggested list of learners when enrolling them into a course, so they spend less time searching and brainstorming which users will benefit most from the course.
• Formal content suggestions for managers: Managers can see suggestions for courses in which they should enroll their team members from the My team area of the platform, allowing them to continue their learning journey without the intervention of the administrator.
• Formal content suggestions for learners: As learners enroll in a course, they are provided with suggestions of other courses that they may be interested in. If the AI’s suggestions are not enough, Learn will automatically suggest additional courses based on course categories. Note that it is not possible to manually update the list of suggested courses, as each list is dynamically generated to cater to the individual learner's experience. Learn more about learners’ self-enrollment in courses and learning plans.
• Next content suggestions for learners: Learners are provided with accurate suggestions of informal content that their colleagues are contributing in the platform when they’re already watching an asset so that they can focus on relevant content that creates an individualized learning experience Learn more about Discover, Coach & Share.

Which languages are supported by Docebo AI algorithms?

Most of Docebo’s AI features (including the platform's AI-powered search) work with our available product languages. However, the auto tagging feature works only for English, Spanish, French, German, Italian and Portuguese content at the moment, but there is full intention to add support for a number of other platform languages for these features over time. We will roll out these capabilities for other languages, together with incremental enhancements and maintenance of all AI functionality. Keep an eye on the Product updates page (opens in a new tab) for more information in the future.

The auto skill assignment feature has been tested in English, but the model supports all the platform languages. Ongoing efforts are in place to continuously improve the quality of assigned skills and expand the capabilities of the feature.

Best practices on audio and video for auto tagging

The auto tagging algorithm can extract tags also from any spoken audio of multimedia learning contents (both formal and informal) that you upload on your platform. However, there are some best practices for the correct interpretation of spoken audio by the state-of-the-art speech-recognition technology the company relies on:

• there are no significant background noises that could cover the spoken audio
• there is no or very little overlap between the speakers (in the event that more speakers speak)
• the spoken audio language is the same throughout the course of the recording
• a very strong speaker accent can lead to less precision

An audio track with any of the above characteristics extends the ability of speech recognition and comprehension technology to clearly distinguish spoken words and turn them into text, on which the automatic encoding algorithm operates. In these cases, the suggested tags produced through auto-tagging may be poor, irrelevant, or non-existent.