Skip to main content

​
Suggestions

Suggestions
  1. Scheduled trigger: We send out suggestions at specific time intervals. On such an interval cycle, we trigger the TechWolf Skill Assistant. The interval can be changed depending on your needs.
  2. Fetch suggestions: To get those suggestions, we fetch them from our own API. More specifically, from the Suggestions Skill Profile endpoint, for all employees.
  3. Format suggestions: The Skill Assistant first determines which users should receive suggestions or reminders. The system determines this by checking when each user last received a suggestion and comparing it against the configured timing parameters for suggestions and reminders. The suggestions are then formatted based on the configurations for the message content. See how the message looks like in the How It Looks section.
  4. Send suggestions: We send out the suggestions to the messaging platform.
  5. Users receive the message: The messaging platform delivers the message to the correct user.

​
How the Skill Engine API max skills setting affects suggestions

The Skill Engine API uses the max_skills setting to return the top ranked Skills for an Employee based on their data. This setting does not directly mean that an Employee will only ever see that number of suggestions over time. The ranking is based on the Employee Skill Profile and includes Skills that were already accepted or rejected. The Skill Engine API then filters out already validated Skills before suggestions are shown in the Skill Assistant.

​
What happens near the limit

Rejected Skills can still influence the top max_skills ranking through an indirect effect. A rejection acts as feedback for the recommendation model and can change the confidence scores of related Skills. When an Employee is close to the limit, this recalculation can reorder the Employee Skill Profile and cause a different Skill to enter the top max_skills.
  1. The Skill Engine API returns the top max_skills Skills from the Employee Skill Profile.
  2. The Employee validates suggestions (for example, accepts 6 and rejects 4).
  3. The recommendation scores are recalculated based on the new feedback, including the rejection signal, which can change nearby Skill rankings.
  4. If this recalculation pulls another Skill into the top max_skills, that Skill can appear as a new suggestion.

​
What happens when data changes

When new Skill Events are added, probabilities can change and different Skills can enter the top max_skills ranking. This can create new suggestions for the Employee. Employee Skill Profile states after new Skill Events
  1. New Skill Events are added for the Employee.
  2. New inferred Skills are scored and compared against the current top max_skills ranking.
  3. Only the new Skills that enter the top max_skills become visible suggestions.
  4. Repeating this process over time can lead to more than max_skills validated Skills on an Employee profile.

​
Validations

Validations
  1. The user clicks submit on a card, which informs the messaging platform that it should send an update to the TechWolf Skill Assistant.
  2. The message is received by the Skill Assistant, which receives a list of confirmed and rejected skills.
  3. The TechWolf Skill Assistant stores this information, such as which skills, but also when and by whom.
  4. The feedback is sent to the TechWolf API, using the Employee Skill Profile feedback endpoint and can be tracked with the source field as “tw--bot”. To access feedback provided by an employee through the Skill Assistant, use the Employee List Skill Event endpoint with skill_profile_feedback as event_type. The resulting Skill Events with source field as “tw--bot” are the feedback events given using the Skill Assistant.

​
Employee Skill Management Tab

In addition to our Skill Assistant: We offer an interactive tab that enables employees to efficiently manage and track their skills. Within this tab, employees can:
  • View and Reevaluate Skills: Access their validated and rejected skills for assessment and potential updates.
  • Add Missing Skills: Identify and add any skills that may be lacking.
  • Self-rate Skill proficiency levels (optional): Update the proficiency level of a Skill.
Skill Management Tab Interface

​
Linking users

​
Through Custom Properties

An important step for the Skill Assistant to work is linking employee profiles in TechWolf to users in the messaging platform. This can be done by setting a Custom Property on the employee profile in TechWolf. The Skill Assistant will then use this Custom Property to link the employee profile to the user in the messaging platform.

​
Through Microsoft Graph

Another way to link employee profiles in TechWolf to users in the messaging platform is to use the Microsoft Graph API. This is the preferred method, as it is more secure and reliable. TechWolf will read the Employee ID from the Microsoft Graph API through the following endpoint: 
https://graph.microsoft.com/v1.0/users/<user id>?$select=employeeId
Notice the employeeId query parameter to filter out the Employee ID of the given Teams user.

​
Data flows

​
General overview

Data Flow Overview

​
Components

Microsoft Teams The Microsoft Teams environment used by the customer. Azure infrastructure TechWolf utilizes Azure infrastructure to facilitate communication between Microsoft Teams and the Skill Assistant. This infrastructure comprises:
  • Bot Service: Required by Microsoft Teams to create a bot. It acts as the intermediary between Teams and the Skill Assistant, handling message exchanges and event triggers from Teams users.
  • Entra Application: Handles authentication and authorization with Microsoft Teams and the Graph API.
  • Front Door: Routes incoming requests from the Bot Service or Skill Profile tab to the appropriate Skill Assistant deployment (EU/US).
Skill Assistant TechWolf’s backend system responsible for generating skill suggestions, processing feedback, and managing data for the Skill Profile tab. Skill Engine API The main TechWolf product. Handles skill-related data processing and integration.

​
Data processing

Types of data that is processed
  • Skill Suggestions: Automatically inferred skills for an employee, retrieved from the Skill Engine API.
  • Skill Feedback: Employee feedback on their suggested skills. Either validating that they have the skills, or rejecting the skills.
  • Skill Profile: A list of validated and rejected skills maintained in the Skill Engine API.
  • Microsoft User Data: Employee details retrieved from Microsoft systems, including name and email.
Data Processing Overview
Managed bySuggestionsFeedbackSkill ProfileMicrosoft User Data
Microsoft TeamsCustomerXXXX
Azure: Entra ApplicationTechWolf
Azure: Bot ServiceTechWolfXXX
Azure: Front DoorTechWolfXXXX
Skill AssistantTechWolfXXXX

​
Data storage

Overview of Data Storage Locations
Managed bySuggestionsFeedbackSkill ProfileMicrosoft User Data
Microsoft TeamsCustomerXXX
AzureTechWolf
Skill AssistantTechWolfXXX (No personal identifiable information stored)

​
Communication Protocols

All data exchanges occur over REST APIs secured by TLS 1.2 or higher.

​
Suggesting Skills

Suggesting Skills Process
  1. Fetch suggestions: On a regular basis, the Skill Assistant will fetch suggestions generated by the API.
  2. Routing the suggestion: The suggestion is sent to the Azure Tenant, where it is routed to the concerning employee.
  3. Sending the suggestion: The suggestion is sent through the ‘adaptive card’ format by the Bot Service to the customer Teams tenant.
Data Involved
  1. Teams ID, Internal ID, Suggestion
  2. Teams ID, Suggestion
  3. Teams ID, Suggestion

​
User Input

User input entails button clicks, messages and installation events. User Input
  1. User input request: When a user interacts with a suggestion, it is sent to the Bot service in the TechWolf Azure instance.
  2. Routing to Skill Assistant: The request is processed and routed by the Bot Service to the Skill Assistant. Only necessary information is retained, and names and email addresses are dropped.
  3. Execution of request: The Skill Assistant directs the request in proper format to the API, where the request is executed on the customer’s data.
  4. Response to user: A response, based on the request’s proper execution, is sent back to the Bot Service to route it back to the concerning user.
  5. Sending the response: The response is sent to the user by the Bot Service.
Data Involved
  1. Event Data, Feedback, MS User Data (Contains first name, last name and email address which cannot be dropped from the request. ref)
  2. Event Data, Feedback, MS User Data (Contains first name, last name and email address which cannot be dropped from the request. ref)
  3. Internal ID, Teams ID, Depending on the event: Feedback, Request suggestions or Skill Profile
  4. Teams ID, Custom response, Depending on the event: Suggestion or Skill Profile
  5. Teams ID, Custom response, Depending on the event: Suggestion or Skill Profile