# Functional Capabilities The chatbot offers a wide range of functional capabilities designed to support internal Customer Service agents in their day-to-day operations. Leveraging both information retrieval and generative AI, the chatbot is able to streamline agent workflows, improve the accuracy of responses, and handle complex queries across multiple languages and conversation threads. At the core of the chatbot is its generative AI functionality, which allows it to synthesize information into coherent, concise, structured and easy-to-read responses. These responses are synthesized from the knowledge the chatbot retrieves, structured to make them suitable for customer service communication.\ The AI model generates responses in a conversational style, ensuring the information is tailored to the specific context of the query. For example, if an agent asks for troubleshooting steps, the chatbot will produce a step-by-step guide that is both clear and concise. ### 5.1 Application switch Users who have access to multiple applications can seamlessly switch between them using the application switcher located at the top header, on the right side of the screen. * For **Super Admins**: The switcher displays a list of all available applications. * For **other users** (Admins, Knowledge Managers, or Agents): The list includes only the applications to which they have access. If a user has access to only one application, the switcher is disabled, and no switching options are available. **Note**: A user can have different roles in different applications but only one role per application. When the user switches to a different application, their role is automatically updated to reflect their specific permission level within the selected application. ### 5.2 Knowledge Bots A *Knowledge Bot* is a specialized chatbot that leverages a specific database to generate responses to user queries. It is designed to provide targeted, relevant answers based on the data it has access to. #### 5.2.1 Knowledge bots listing The listing page displays all the Knowledge Bots available within the current application. The most recently created bots appear at the top of the list. Users can view detailed information, edit, delete, or add new Knowledge Bots. The list includes the following details for each bot: * ID * Name * Logo (if available) * Creation date * Creator’s user role ***Access Roles:*** * **Super Admins**: Can view and manage Knowledge Bots within the current application (he has access to all applications). * **Admins**: Can view and manage all Knowledge Bots within the current application. * **Knowledge Managers**: Can view only the Knowledge Bots assigned to them; they do not have permissions to add, edit, or delete. * **Agents**: Do not have access to this page. #### 5.2.2 Add / Edit Knowledge Bot Only *Super Admins* and *Admins* have permission to create, update, or delete Knowledge Bots. When creating a new bot, a modal window will open containing the input fields: * **Name**: The name of the Knowledge Bot. * **Answer Language**: The default language used by the knowledge bot to respond. * **Client-specific Instructions**: Additional rules or guidelines for the chatbot. * **Output Length**: Specifies the approximate length of responses in characters, influencing the level of detail. * Small: 50 characters * Medium: 200 characters (default) * Large: 1,000 characters * Extra Large: 10,000 characters * **Formality:** Sets the tone of responses. * Informal (default) * Formal * **Expertise:** Sets the level of details in response. * Low * Medium (default) * High * **Design:** * **Logo**: Upload a logo (supported formats: PNG, JPG; max size: 7MB) * **Background Color**: Customizable via a color picker. * **Left menu Background & Text Colors**: Customizable via a color picker. * **Data Masking Fields:** If data masking is activated in the application, a list of available masking flags is provided for selection to secure sensitive information. The design fields allow customization of the chat interface’s appearance, including colors. When editing an existing Knowledge Bot, the same modal appears with pre-filled information. To delete a Knowledge Bot, a confirmation modal prompts the user to verify the action by entering the bot’s name. #### 5.2.3 Knowledge Bot details The *Knowledge Bot Details* page displays all information about the selected bot: general details and data sources from the current knowledge bot, which will be detailed in *5.3 Data sources.* The general details of the knowledge bot are: * Logo and Name * Creator’s Name * Answer Language * Response Output Length * Creation Date * Expertise Level * Application Name * Formality Level * Total Data Sources: Number of Files / Web Pages * Data Masking Status (Enabled / Disabled) If *Data Masking* is enabled, a tooltip shows the specific flags applied. If *Client-specific Instructions* are provided, a dedicated section displays these instructions. #### 5.2.4 PII data masking Personally Identifiable Information (PII) Data Masking is a feature that can be enabled or disabled at the application level. When adding or editing a Knowledge Bot, if Data Masking is enabled for the application, users can configure specific masking rules for that bot. These flags are used to anonymize or protect sensitive questions in Chat before they are sent to the LLM. On the Knowledge Bot details page, if masking flags are in use, the Data Masking field will be *Enabled*. Clicking the info icon reveals which flags are applied, ensuring users understand what data is being protected. ### 5.3 Data sources For each Knowledge Bot, users can add data sources from which the chatbot will retrieve responses. These sources enable the chatbot to integrate with various sources to provide accurate and real-time information to agents. The supported data sources include: * **Internal Knowledge Repository**: The chatbot can retrieve information from the system’s internal repository, where knowledge materials can be uploaded in formats such as Microsoft Word (.docx), Excel (.xlsx), PowerPoint (.pptx), PDF (.pdf) and Text (.txt) files. * **Web Scraping**: The chatbot has the ability to scrape data directly from websites, ensuring it can access real-time external information when necessary, expanding the range of data beyond internal sources. * **Major Knowledge Management Systems**: Integration with internal and external Confluence and SharePoint instances. These integrations support SSO authentication, allow users to select specific documents or pages for scraping, and provide options for sensitive data masking and scheduled updates to keep information up to date. On the Knowledge Bot details page, users can view the list of data sources associated with the current knowledge bot and have the option to upload new ones. #### 5.3.1 Data Sources listing The data sources listing section displays all materials linked to the current Knowledge Bot. This interface allows users to manage and monitor the content that the chatbot references when generating responses. Items are ordered by the most recent update date, descending. For each data source, the following details are shown: * **Data Source**: The name of the material, helping identify its origin. * **File/URL**: Displays the name of the uploaded document or the URL of the scraped webpage / Confluence / SharePoint. This allows user to quickly recognize the specific files or pages that have been incorporated into the chatbot's knowledge base. * **Status**: Indicates the current processing status of each item: * **Success**: The material has been processed successfully and is ready for use. * **Pending**: Processing is ongoing; the user must wait until it is complete. * **Not Refreshed**: The material was not replaced or updated successfully through recurrency or manual update. * **Fail**: The material could not be processed. * **Actions available**: * **File details**: Open the details of the data source. * **Replace**: Allows modification or replacement of the data source. * **Delete**: Removes the data source if it is no longer relevant. Users can also search for specific materials. For hierarchical sources such as Confluence or SharePoint, pages are displayed in a tree view that reflects their structure from Confluence / SharePoint. Folders in Confluence or SharePoint do not have individual detail pages; they serve only to replicate the structure visually. Processing statuses influence available actions: * If **Success**, all options are available. * If **Pending**, no actions can be performed until processing is complete. * If **Not Refreshed**, all options are available * If **Fail**, users can Replace or Delete, and a tooltip will explain the failure reason. In this case, the details page is unavailable as no data has been ingested. #### 5.3.2 Data Ingestion To add new data sources, users should select the **Upload** tab, which offers four options corresponding to supported data types: * **Document:** for PDF, TXT, XLSX, PPTX, DOCX files * **URL:** for websites * **Confluence:** for Confluence pages * **SharePoint:** for SharePoint documents and pages After selecting the data source type, additional fields specific to that type will appear. Once the data source is saved, it remains in a **Pending** state until processing begins. After processing starts, the system checks every 5 minutes for new chunks. If none are created within this interval, the data source will be marked as **Failed**. This process ensures that large files are processed efficiently without timeout issues; if a file remains stuck, it will be marked as failed after retries. ***5.3.2.1. Documents ingestion*** The knowledge bot can retrieve information from the internal repository where knowledge materials are uploaded in formats such as PDF (.pdf), TXT (.txt), XLSX (.xlsx), PPTX (.pptx), and DOCX (.docx). * Users can upload up to 10 documents at once, all the same type. * The maximum file size per document is 30MB. * Ingested data is extracted as plain text (excluding images or comments). **Chunking and ingestion configuration:** \ Depending on the file type, users can set additional settings: * **Divide text by** – provides options to divide the document into manageable chunks, for optimal processing: * **Split by words:** Divides content into chunks of 200 to 4,500 words (PDF, DOCX, TXT). * **Split by page:** Separates content based on document pages (PDF). * **Split by paragraph:** Separates by paragraphs (PDF, PPTX, DOCX). * **Split by slide:** For PowerPoint documents. * **Chunk settings:** * **Data Source Description:** A brief description of the content to improve accuracy and relevance of information retrieval. * **Overlap:** Number of words repeated between chunks to enhance retrieval context. Higher overlap enhances accuracy but may increase processing load. The option is not available for Excel files. * **Add description to each chunk:** If enabled, the description is prepended to each chunk, to improve retrieval and accuracy. * **Excel-specific options:** * **Header type:** * **No header** * **Columns** – the header is applied to columns, and the header section is displayed horizontally * **Rows** – the header is applied to rows, and the header section is displayed vertically * The header is added to each chunk * **Header size:** Number of rows or columns comprising the header. * **Priority:** A value between 1 (low) and 100 (high), indicating the importance of the content for retrieval, to ensure critical information is emphasized in responses. When adding Excels as data source, if the user does not select a specific chunking strategy then the default strategy is used: to check the maximum number of rows / columns that can fit in a chunk as long as the size is smaller than the embedding model max token input (ADA Models have a max token input of 8191 tokens). For PDFs data sources, the files are processed with Azure Document Intelligence with the Markdown content format. ***5.3.2.2 Web Scraping*** The chatbot has the ability to scrape data directly from websites, ensuring it can access real-time external information when necessary, expanding the range of data beyond internal sources. The user can specify one or multiple URLs, up to 50, from where the data will be scrapped. Additional settings can be made to configure how the data is ingested. * **Recurrence:** Set how often the content is refreshed automatically (e.g., no recurrence, 1 day, 2 days, 4 days, 1 week, 2 weeks, 1 month, 2 months, 3 months). * **Divide text by:** Set how to split the page in chunks, for optimal processing * **Split by words:** Divides content into chunks of 200 to 4,500 words * **Depth:** Determine whether to include only the current page (`0`) or the current page and all subpages (`Infinite`). * **Chunk settings** * **Data source description** – to provide a clear and descriptive name for the content; this helps improve accuracy and relevance of information retrieval and enhances response quality * **Overlap** – sets the number of words repeated between consecutive chunks to improve retrieval context; higher overlap enhances accuracy but may increase processing load. * **Priority** * **Data source priority** - A value between 1 (low) and 100 (high), indicating the importance of the content for retrieval, to ensure critical information is emphasized in responses. * **Use alternative method** – to enable an alternative method only if the previous standard method provided insufficient content The ingested data can be updated recurrently, so that the information is always up to date. Based on the recurrency selected, the webpages are scraped again, with the same settings as they were scraped before. ***5.3.2.3 Confluence integration*** Before using the integration, an initial configuration by the Admin team is required. Failing to complete this step will prevent users from successfully ingesting content from this source. For more details, refer to [**Confluence Configuration**](https://concentrixgpo.atlassian.net/wiki/spaces/Kbot/pages/1121878165/Functional+Documentation#5.7-Configure-data-ingestion) section. **Confluence** can be integrated as a knowledge source, and the Knowledge Manager or Admin or Super Admin provides the main URL that serves as the starting point for scraping the pages. * Authentication is via SSO, allowing access to pages the user has access to in Confluence. * Up to 500 pages can be loaded from the starting Confluence URL. If this limit is exceeded, the user will be notified. * Pages are shown hierarchically, similar to Confluence’s structure, to facilitate navigation. * Users can select individual pages or the entire structure from the displayed hierarchy. After selecting the Confluence pages to be scraped, users can adjust additional settings to configure how the data will be ingested. ***Advanced settings*** * **Recurrency**: Frequency of automatic updates (e.g., no recurrence, 1 day, 2 days, 4 days, 1 week, 2 weeks, 1 month, 2 months, 3 months). * **Divide text by:** Set how to split the page in chunks, for optimal processing * **Split by words:** Divides content into chunks of 200 to 4,500 words * **Chunk settings** * **Overlap:** sets the number of words repeated between chunks to improve retrieval context; higher overlap enhances accuracy but may increase processing load. * **Priority** * **Data source priority:** The importance level of the content in retrieval, between 1 (low) and 100 (high), to ensure critical information is emphasized in responses. * **Use alternative method:** to enable an alternative method only if the previous standard method provided insufficient content Scraped Confluence pages are stored as individual data sources, excluding attachments and images. In the data sources list, these pages are displayed in a hierarchical structure - like Confluence, to make it easier for users to locate specific pages. If a parent item is not selected but its child items are, those child items will be linked to the nearest selected parent item. The ingested data can be updated on a recurring basis to ensure the information remains current. Based on the selected recurrence, only the pages that have been updated will be scraped again, using the same configuration as before. If an authentication failure occurs during a scheduled update, the data source will be marked as **Not refreshed**. In this case, the *Replace* option becomes available, allowing the user to re-enter valid credentials ***5.3.2.4. SharePoint integration*** Before using the integration, an initial configuration by the Admin is required. Failing to complete this step will prevent users from successfully ingesting content from this source. For more details, refer to [**SharePoint Configuration**](https://concentrixgpo.atlassian.net/wiki/spaces/Kbot/pages/1121878165/Functional+Documentation#5.7-Configure-data-ingestion) section. **SharePoint** can be integrated as a knowledge source, and the Knowledge Manager or Admin or Super Admin provides the main URL that serves as the starting point for scraping the pages. * Authentication is via SSO, allowing access to files and pages the user has access to in SharePoint. * Up to 500 items can be loaded from the starting SharePoint URL. If this limit is exceeded, the user will be notified. * Files / pages are shown hierarchically, similar to SharePoint’s structure, to facilitate navigation. * Users can select specific files, pages or the entire structure from the displayed hierarchy. In the *data sources list*, the scraped files and pages are displayed in a hierarchical view that mirrors the SharePoint structure, making it easier for users to locate specific files / pages. If only child items are selected without their parent, those items are linked to the first selected parent or, if no parent is selected, promoted to top-level items. After selecting the SharePoint items to be scraped, users can adjust additional settings to configure how the data will be ingested. ***Advanced settings*** * **Enforce access control:** when enabled, responses generated by the chatbot will be restricted to the data the Agent has access to in SharePoint. This ensures sensitive or restricted information is not disclosed to unauthorized users. * **Recurrency:** Users can set how frequently the data is automatically refreshed and re-scraped. (e.g., no recurrence, 1 day, 2 days, 4 days, 1 week, 2 weeks, 1 month, 2 months, 3 months). * **Divide text by:** Set how to split the page in chunks, for optimal processing * **Split by words:** Divides content into chunks of 200 to 4,500 words * **Depth**: Determines whether only the current page (`0`) or the page and all subpages (`Infinite`) are included in the embedding process. * **Chunk settings** * **Data source description:** description for the content; this helps improve accuracy and relevance of information retrieval and enhances response quality * **Overlap:** sets the number of words repeated between chunks to improve retrieval context; higher overlap enhances accuracy but may increase processing load; * **Priority** * **Data source priority:** The importance level of the content in retrieval, between 1 (low) and 100 (high), to ensure critical information is emphasized in responses. * **Use alternative method**: Enables an alternative scraping method if the standard approach provides insufficient content. Scraped SharePoint files / pages are stored as individual data sources. The document types that can be scraped from SharePoint are the ones supported by platform (pdf, txt, xlsx, pptx, docx). The pages supported from SharePoint are webpages. WebParts are not supported. In the data sources list, the scraped files / pages are displayed in a hierarchical structure - like SharePoint, to make it easier for users to locate specific files. If a parent item is not selected but its child items are, those child items will be linked to the nearest selected parent item. When adding new data sources from SharePoint, *Knowledge Managers / Admins / Super Admins* can enable access restrictions for *Agents*, ensuring that chatbot responses only include data accessible to them in SharePoint. This maintains data security and compliance. The ingested data can be updated on a recurring basis to ensure the information remains current. Based on the selected recurrence, only the files / pages that have been updated will be scraped again, using the same configuration as before. If an authentication failure occurs during a scheduled update, the data source will be marked as **Not refreshed**. In this case, the *Replace* option becomes available, allowing the user to re-enter valid credentials. ***5.3.2.5. Salesforce integration*** Before using the integration, an initial configuration by the Admin is required. Failing to complete this step will prevent users from successfully ingesting content from this source. For more details, refer to [**Salesforce Configuration**](https://concentrixgpo.atlassian.net/wiki/spaces/Kbot/pages/1121878165/Functional+Documentation#5.7-Configure-data-ingestion) section. **Salesforce** can be integrated as a knowledge source, and the Knowledge Manager or Admin or Super Admin provides the main URL that serves as the starting point for scraping the articles. * Up to 500 items can be loaded from the starting URL. If this limit exceeds, the user will be notified. * Users can select specific articles or the entire structure from the displayed hierarchy. After selecting the Salesforce articles to be scraped, users can adjust additional settings to configure how the data will be ingested. ***Advanced settings*** * **Recurrency:** Users can set how frequently the data is automatically refreshed and re-scraped. (e.g., no recurrence, 1 day, 2 days, 4 days, 1 week, 2 weeks, 1 month, 2 months, 3 months). * **Divide text by:** Set how to split the page in chunks, for optimal processing * **Split by words:** Divides content into chunks of 200 to 4,500 words * **Chunk settings** * **Data source description:** description for the content; this helps improve accuracy and relevance of information retrieval and enhances response quality * **Overlap:** sets the number of words repeated between chunks to improve retrieval context; higher overlap enhances accuracy but may increase processing load; * **Priority** * **Data source priority:** The importance level of the content in retrieval, between 1 (low) and 100 (high), to ensure critical information is emphasized in responses. Scraped articles are stored as individual data sources. These contain some fields from Salesforce articles, as described [here.](https://concentrixgpo.atlassian.net/wiki/spaces/Kbot/pages/981893378) The ingested data can be updated on a recurring basis to ensure the information remains current. Based on the selected recurrence, the articles will be scraped again, using the same configuration as before. If an authentication failure occurs during a scheduled update, the data source will be marked as **Not refreshed**. In this case, the *Replace* option becomes available, allowing the user to re-enter valid credentials. #### 5.3.3 Recurrency The SharePoint / Confluence / Web Site / Salesforce ingested data can be updated on a recurring basis to ensure the information remains current. On the Advanced Settings section a new field is available, and the user can set how frequently the data is automatically refreshed and re-scraped. (e.g., no recurrence, 1 day, 2 days, 4 days, 1 week, 2 weeks, 1 month, 2 months, 3 months). Based on the selected recurrence, the files / pages are scraped again using the same configuration as before. For Confluence / SharePoint only the files/pages that have been updated will be scraped again. If an authentication failure occurs during a scheduled update, the data source will be marked as **Not refreshed**. In this case, the *Replace* option becomes available, allowing the user to re-enter valid credentials. #### 5.3.4 Data Source details For each data source, users can view detailed information by selecting either the entire row or the details icon from the data sources list. The detailed view includes: * **Details Block:** Displays the file name or page URL, along with the associated Knowledge Bot name. * **Hierarchy Block:** Shows the structure for Confluence and SharePoint data sources. * **Updates Block:** Indicates the date and time of the last update. * **Settings Block:** Contains details about the chunking strategy applied to the data source, including: * **Divide Text Method:** The method used to split the content. * **Chunk Overlap:** The number of words repeated between chunks to preserve context. * **Average Chunk Size:** The number of words per chunk. * **Recurrence:** The schedule for automatic updates, if configured. * **Data Masking:** Indicates whether masking is enabled or disabled. * **Priority Block:** Displays the current priority level and allows adjustments. * **Search Field:** Enables users to search for specific text within the chunks of the current data source. * **Chunks List:** Shows all chunks created for current data source based on the selected chunking strategy, with details including: * Content view * Number of tokens * Number of words * Screenshots (if available, applicable for PDFs and PPTX files split by words or paragraphs) Users can edit individual chunk content or delete specific chunks as needed. **Data Masking Information:** \ For Data masking information, if enabled, a tooltip displays the masking flags applied to the data source, indicating which sensitive entities are masked. **Hierarchy Block** reflects the same structure as seen in the data sources list, highlighting the current data source within the hierarchy. Users can navigate to other items in the hierarchy to view their details. Note that folders cannot be selected since they do not contain detailed content. #### 5.3.5 Replace data source Users can manually update ingested data when needed. From the data sources list or a data source details page, click **Replace** to open a modal that allows updating or replacing the data. Replace Modal — fields and settings: * Data source name * File name / URL * Knowledge Bot name * Upload section (visible if the current data source is a document) * Advanced settings (depend on data source type; previously selected values are prefilled): * Divide text by — same options as on upload * Overlapping * Chunk overlap * Data source description * Add description to each chunk (documents only) * Recurrence — for Web, Confluence, or SharePoint * Authentication — for Confluence and SharePoint * Data masking — if enabled for the application For documents, the upload section is displayed but uploading a new file is optional. The user can add another file to be ingested to replace the existing one, or if they want to use the same file but with different settings, no document needs to be provided. If a new document is uploaded, this needs to have the same type as the original one. For Confluence and SharePoint, the URL of the current data source is displayed and cannot be changed. After authentication, the hierarchy form Confluence / SharePoint is displayed starting with the current data source URL. The previous selected sources are selected by default, and the user can make changes and select / unselect other items. If new items were added in Confluence / SharePoint under the current item, then these will appear in the hierarchy, unselected by default, and the user can select them if he wants. If some previously selected items were removed from Confluence / SharePoint, then these will not appear in the hierarchy and on replace these data sources will be removed. The items that were scraped before and remain selected, will be scraped again if there were changes made to them. If there are no changes on these items, then they will not be scraped again. For SharePoint, an additional option is present, to *Enforce access control* to give restricted access to Agents on data sources. For Confluence and SharePoint, the recurrency can be set/updated only on the main item from hierarchy and the value will be inherited in all the children of that main item. On confirmation, the platform re-scrapes the selected source(s) and updates the data source. #### 5.3.6 PII data masking If data masking is enabled for the current application, Admins and Knowledge Managers can apply masking to data sources. Masking is applied at ingestion (or on replace) so sensitive values are masked inside chunks and remain masked in chatbot responses. When adding a data source, the Data Masking section lists all available masking flags. By default, the flags that are set on current knowledge bot are preselected. Users may change the selection for the new data source. For documents, a data masking section appears for each uploaded document and can be different for each of them. For the rest of the data sources (Web, Confluence, SharePoint), a single Data Masking configuration applies to all selected pages/documents. In the Replace modal, Data Masking appears under Advanced Settings with previous selections prefilled. On the Data Source details page, if masking flags are applied the *Settings* block shows **Data Masking** as *Enabled*; an info icon reveals which flags are active. If a data source has recurrence enabled, the same masking flags are re-applied automatically on each refresh. #### 5.3.7 Delete data source Super Admins, Admins and Knowledge Managers can delete data sources from either the data sources list or the data source details page. Deletion requires confirmation via a modal to prevent accidental removal. When a data source is deleted, its chunks are removed, and the Knowledge Bot can no longer use that source for responses. If the data source is a Confluence page or a SharePoint item that has child items, the confirmation modal offers two delete options: * **Delete Embedding** — Removes only the selected data source. Child items are re-linked to the next available parent in the hierarchy; if no parent exists, they become top-level items. * **Delete Hierarchy** — Removes the selected data source and all its child items. ### 5.4 Users The Users Management page lists all users who have access to the current application and their Knowledge Bot assignments. Accessible by Super Admins, Admins, and Knowledge Managers, available actions depend on the viewer’s role. ***Visibility and permissions*** * Super Admin * Can view all users (Admins, Knowledge Managers, Agents). * Can add, edit, and delete any user. * Admin * Can view Admins, Knowledge Managers, and Agents within the current application. * Can add, edit, and delete Admins, Knowledge Managers and Agents. * Cannot edit or delete his own Admin user. * Knowledge Manager * Can view Knowledge Managers and Agents assigned to the Knowledge Bots they manage. * Can add, edit, and delete Agents assigned to their Knowledge Bots. * Cannot add, edit, or delete other Knowledge Managers. Each row from users list represents one User–Knowledge Bot assignment and shows: Name, Email, Knowledge Bot, Role. The list can be filtered by Knowledge Bot or Role and searched by first name, last name, or email. Knowledge Bot list from filter contains only the ones that have specific users assigned. Super Admins roles are not displayed in Users list. ***Adding users*** When adding a user, the following information can be provided: * Name * Surname * Email * Knowledge Bot – list with all Knowledge Bots from current application to which the current user has access * Role – the roles lower than current user When adding an *Admin*, the Knowledge Bot is disabled because the Admin has access to all knowledge bots from current application. When adding a *Knowledge Manager or Agent*, assigning a Knowledge Bot is required and only one can be added at once. To grant a user access to multiple Knowledge Bots, add one entry per Knowledge Bot using the same email. A user may create only users with roles lower than their own. Each user can have only one role per application. An Admin user may create other Admins and users with roles lower than their own. A Knowledge Manager user may create only users with roles lower than their own. Each user can have only one role per application. ***Validation Rules:*** * Name, Surname, Email and Role are mandatory. * Name and Surname should have between 2 and 255 characters, and only letters * Email should have email format * Knowledge Bot is mandatory if role is Knowledge Manager or Agent * If the email belongs to a Super Admin, an error is shown. * If the email is already added on the current application as Admin, an error message is displayed. * If the email is already added on the current application as Knowledge Manager / Agent and now is added again with a different role, an error message is displayed. * If the email exists with a different role, an error is shown. * Knowledge Managers or Agents can be promoted to Admin, gaining access to all knowledge bots in the current application. * If the email exists with the same role (Knowledge Manager / Agent) but for a different Knowledge Bot, the user details are updated with the new knowledge bot access. ***Editing users*** Current user can edit only the users that have a lower role, except admins that can update other admins. When editing an user, the name, email and role can be updated. When editing a Knowledge Manager or Agent, then knowledge bots list reflects all the Knowledge Bots the current edited user has access to. These assignments cannot be changed in the edit flow. The role can be changed to Agent, Knowledge Manager or Admin and the user will have the same access to all knowledge bots. ### 5.5 Chat Chat functionality uses generative AI models to produce human-like, easy-to-read responses. These responses are synthesized from the data sources the knowledge bot retrieves, structured to make them suitable for customer service communication. \ The AI model generates responses in a conversational style, ensuring the information is tailored to the specific context of the query. Before starting the conversation, the user can configure some elements: * **Knowledge Bot:** Selects which bot (and its data sources) to use. The list shows only bots from the current application that the user can access and have data sources added. * **Language:** Response language. Defaults to the bot’s configured language but this can be changed to another one. * **Advanced Configuration:** Optional, knowledge bot specific configuration that can be made so that the chatbot responds according to business needs. * **New Chat Window:** Opens a fresh conversation (advanced instructions are not available in this mode). The **Advanced Configuration** options are by default filled with the options set on knowledge bot and any change affects only the behavior of the current chat window of the user. The **Advanced Configuration** section allows Super Admins, Admins and Knowledge Managers (but not Agents) to optimize content and fine-tune the chatbot’s output to ensure it aligns with specific project needs. These features provide greater control over how the chatbot interacts and delivers responses. * **Client Specification Instructions:** This field allows user to insert specific instructions that guide the behavior of the chatbot. These instructions help customize the responses and refine the chatbot’s accuracy by aligning it with client-specific guidelines and requirements. Use this feature to ensure the bot’s behavior and output meet your project’s unique needs. * **Output Length:** This setting allows user to specify the desired length of the chatbot’s responses, depending on the context of the customer interaction and the level of detail required. * **Formality:** The **Formality** setting controls the tone of the chatbot’s responses. Users can adjust the level of formality in the output to match the desired communication style, ranging from casual to highly formal, ensuring that the bot’s responses are appropriate for the target audience or use case. * **Expertise Level:** This setting controls the depth and complexity of the chatbot’s answers. Adjust the **Expertise Level** to determine how detailed and technical the responses should be. Use this feature to ensure the bot provides information at the appropriate level of expertise, whether for general queries or more in-depth, specialized topics. Advanced Configuration option is not available for “New chat window”. Chat page UI can be themed per Knowledge Bot (configured on bot settings): * Top-row logo * Background color * Left-menu background and text color When Chat page is opened, no history will be shown, even precious conversations took place. Persistence of previous questions / answers will be kept only at tab level. So, the users start with a clean page and will be able to see the history of questions / answers while the user's session is active. Any close and reopen of the tab will clean the questions history. If the application has no data sources, then the chat page remains accessible, but all controls are disabled, and an explanatory message is shown. #### 5.5.1 Questions and Responses There are two types of questions: **Main questions**, which starts a conversation, and **Follow-up questions**, that continue a conversation and receive the answers considering the previous questions and answers from the same conversation. The chatbot is designed to handle **multiturn conversations**, where a query involves multiple exchanges or follow-up questions. The chatbot retains the context of each ongoing conversation, allowing it to respond appropriately even as the dialogue evolves. * **Conversation History Retention**: The chatbot keeps track of previous queries and responses in a session, which is particularly helpful when an agent needs to handle complex customer issues over several interactions. * **Context Awareness**: By remembering earlier parts of the conversation, the chatbot can adjust its responses dynamically. For example, if an agent follows up on a previous query, the chatbot can provide information based on the entire conversation context, not just the most recent input. If rephrasing is enabled for the application, the chatbot will attempt to rewrite the user’s question before generating an answer. The rephrased question (when produced) is shown beneath the original. Answers are generated from the Knowledge Bot’s configured data sources that are uploaded with success. When the data source has recurrency or manual replace, while the status is Pending / Processing / Not refreshed, then the current version of data source is considered for generating the response. If the chatbot cannot find the response in the available data sources, then a message is displayed. If rephrasing fails and the model still returns an answer, that response is flagged with a light-red highlight to indicate a potential hallucination or out-of-context content. For SharePoint sources with enforced access control, *Agents* receive answers only from documents they can access in SharePoint. Knowledge Managers, Admins and Super Admins receive answers from all data sources available to the Knowledge Bot. The data sources that are *In processing* or *Failed* are not considered for generating an answer. If the data source is with recurrency and on the next update the authentication fails, then the status is *Not refreshed* and the current version of the data source is considered for getting the response. If the response contains *Links*, then these are visible with link specific style and are clickable. If current knowledge bot has data masking enabled, and flags selected, then the question is masked using those flags. If the data source is not masked, then knowledge bot data masking is applied on question and context together. When an answer is displayed, the system provides additional features to enhance the user experience and ensure transparency in how the response was generated: * **Source Documents**: The system shows the source materials that were used to compose the answer. This allows users to trace the information back to its original documents for validation or further exploration. The user can open the page / document in a new tab, if this is possible, or download the document, if not. If the chatbot cannot generate a response, the sources section is not displayed. * **Feedback Submission**: Users can provide feedback on the quality of the answer: * **Thumb Up/Down**: Users can quickly rate the answer with a thumbs-up or thumbs-down. * **Detailed Feedback**: A text field is available for users to submit more detailed feedback about the answer, which helps improve the chatbot’s accuracy and relevance over time. * **Copy Answer**: This feature is available to streamline the interaction between the chatbot and other tools. Users can easily copy the generated response and paste it into other platforms or communication channels, improving workflow efficiency. * **View entire source** opens a modal that lists the exact chunks from data sources from where the response was generated. Each chunk has a link to open/download the document from where they are part of. If screenshots are available for these chunks, then these will be displayed below the chunk. * **View thumbnails** is an option available only if there are screenshots available (data source is a **PDF** or **PowerPoint** document). This provides a visual reference to the exact sections of the document used, making it easier for users to verify the content. A maximum of 5 thumbnails will be shown by default, with an option of expanding the view and viewing more. By clicking on a thumbnail, the image will be opened large in a popup view. If the same screenshot appears multiple times, it should be displayed only once. #### 5.5.2 Multiple conversations Users may run multiple chat sessions in parallel by clicking *New Chat Window*. Each chat opens as an independent window and can use a different Knowledge Bot and response language. Multiple windows can be opened at once, each acting independently of each other and having by default the settings of the main window in the moment of launch. Even if on the main page a new application was selected, the already opened windows will keep the original values. Closing the window will lead to losing previously asked questions. #### 5.5.3 Smart prompts Smart Prompts helps Users retrieve relevant answers faster by surfacing common, previously asked questions. * **Top questions:** Users see the five most asked questions of the past 30 days. The list shows rephrased questions if rephrasing is enabled; otherwise, it shows original question. Similar questions are grouped. Questions with masked data, that were disliked, or that produced no valid response are excluded. * **Auto‑fill suggestions**: While typing, Users receive up to five suggested questions that match the entered characters, drawn from the most frequent queries. Suggestions exclude disliked questions and those containing masked data. * **Cached answers**: If an User asks a previously asked question, the system returns the same cached answer to reduce latency and cost. A new response is generated if the prior answer was disliked, relevant data sources were updated, chatbot settings changed, or data masking was enabled. Answers sourced from SharePoint items with enforced access control are not cached. #### 5.5.4 Feedback Mechanism Users have the ability to rate the chatbot’s responses directly through the interface. Feedback options allow them to rate the accuracy and usefulness of the response, and they can provide additional comments if they encounter issues or see room for improvement. * **Rating System**: Users can give a rating (thumbs up/down) after receiving a response. This feedback is used by the system to fine-tune future responses. * **Additional Comments**: Users can also submit comments explaining why they found a response helpful or unhelpful, providing deeper insights for Knowledge Managers and the development team to enhance the chatbot’s performance. #### 5.5.5 Multilingual Capabilities One of the chatbot’s key strengths is its ability to engage in multiple conversations in different languages simultaneously. This is especially valuable for global support teams working in various regions and time zones. Agents can query the chatbot in one language while their colleagues use it in another. This enables large, diverse teams to operate efficiently without language barriers. ### 5.6 Reporting The Reporting Panel provides Key Metrics to Knowledge Managers, Admins, and Super Admins, offering valuable insights into the chatbot’s performance and effectiveness. * **Query Volume and Usage:** Tracks the number of queries submitted over time, providing data on chatbot usage, including total queries handled and breakdowns by specific use cases (e.g., retrieving knowledge base information or live data). * **User Satisfaction:** Measures agent satisfaction with chatbot responses, collected through direct feedback and ratings. User satisfaction is typically scored on a scale, with optional comments for detailed insights. #### 5.6.1 Reporting Dashboard The Reporting page provides various data insights related to Knowledge Bot interactions. Accessible by Super Admins, Admins, and Knowledge Managers, with data visibility tailored to each role. The Dashboard page presents high level information regarding Costs, Feedback, Requests, Models, Providers, Regions, Users. Charts data can be filtered by Start date, End date, specific Use Case, or Assistant (Knowledge Bot). The default range is the current month. All use cases are selected by default, but users can choose a main use case or a specific one. All knowledge bots are selected by default; but the user can select one or more specific knowledge bots. Once the filters are selected and applied, the charts are updated to show information for the filtered data. If charts have no data, then they are not displayed. The use cases considered for reporting are also filtered by role. That means that an Admin should see the use cases initiated by other Admins, Knowledge Managers and Users for current application. A Knowledge Manager should be able to filter only by the assistants they are assigned, and the use cases considered are the ones from assistants (knowledge bots) they have access and are initiated by other Knowledge Managers or Users that are assigned to those assistants. ***Charts & Metrics*** * **Cost Pie Chart**: Shows total costs for the main use case based on current filters. The hovering displays detailed info such as total requests and costs per use case. Available only to Super Admins. * **Feedback Pie Chart**: Visualizes feedback as Negative, Positive, or No Feedback for current filters. Hovering shows exact counts. * **AI Model Breakdown**: Displays request counts per AI model used for current filters; viewable only by Super Admins. * **AI Provider Breakdown**: Shows request counts per AI provider for current filters; viewable only by Super Admins. * **AI Region Breakdown**: Displays request counts by region for current filters; viewable only by Super Admins. * **Top Users**: Lists the top 10 users that made requests, ordered by the number of requests in the selected period and filters (only counting main use case requests). * **Top Embedding Sources**: Shows the number of sources used for embeddings for current filters. ***Overview Details*** * **Total Cost**: Total expenditure in dollars for all requests to LLM during the selected period and filters; viewable only by Super Admins. * **Total Requests**: Counts all user requests, including question queries, rephrasing, follow-up enhancements, file uploads, etc for current filters. * **Average Cost / Request**: Calculated as total cost divided by total requests for current filters; viewable only by Super Admins. * **Average Input Tokens**: Average size of input prompts sent to the LLM, measured in tokens (\~4 characters per token) for current filters. Includes questions, prompts, and context. * **Average Output Tokens**: Average size of the responses generated by the LLM for current filters. * **Average AI Request Response Time**: The time (in seconds) for a request response to happen, for a single use case for current filters. This overview helps monitor costs, request activity, and performance, facilitating insights into chatbot operation efficiency and user engagement. #### 5.6.2 Reporting List The Reporting Dashboard can be switched to a list view, which displays detailed information about use cases in a tabular format based on the selected filters. To access this view, click the **"View Data in List"** button. The page includes a filtering section, the list table, and an export button. The filters are the same as on chart view and the selected filters remain the same when switching between chart view and list view. When **Apply** button is clicked, then the selected filters are applied on the list. When **Reset** button is clicked, the filters go to default state and refreshes the list accordingly. Use cases table contains information about the use cases that were triggered, and the following information is displayed: * **Date:** When the use case was triggered. * **Use Case Description:** Brief description of the use case. * **Assistant:** The Knowledge Bot involved. * **User:** The user who initiated the use case. * **Model:** The AI model used; available only for Super Admins. * **Feedback Grade:** Thumbs-up or thumbs-down feedback, if available. * **Total Cost:** Cost of executing the use case (zero for cached responses); available only for Super Admins. The list is ordered descending by date and supports pagination. Clicking a row takes the user to a detailed view of that use case. Clicking the Export button downloads a CSV file containing the same data as shown in the list. Cost information (Total Cost, Input Cost, Output Cost) is available only to Super Admins. Exported data includes: Date, Application, Provider, Main Use Case, Use Case Description, Model, Total Cost, Input Cost, Output Cost, User, Input Size, Output Size, Duration, Assistant, Agent Name, Original Question, Rephrased Question, Response, Full Input Prompt, Full Output, AI Feedback (Up/Down/Empty), Feedback Comment. #### 5.6.3 Reporting details The Reporting Details page provides comprehensive information about a specific use case. The page includes two main sections: the Metrics section and the Prompt/Response section. **Metrics section** displays use case information such as: * **Date:** When the use case was executed. * **Application:** The application in which the use case occurred. * **Knowledge Bot:** The bot involved. * **AI Model:** The AI model used (accessible only to Super Admins) * **Region:** Geographic region of the request. * **Costs:** Cost details * **User Name:** The user who triggered the use case. * **Input/Output Size:** Number of tokens for input and output. * **Duration:** Time taken for processing. Cost Information (accessible only to Super Admins) * **Total Cost:** Overall cost of the use case. * **Input Cost:** Cost for processing the input. * **Output Cost:** Cost for generating the response. *Note:* For cached responses, the cost is shown as zero. **Prompt/Response Section** * **Prompt**: Displays the user’s question and the full prompt that was sent to AI for current use case. * **Response:** Shows the AI’s generated answer. When a Knowledge Manager or Admin views details, only the user's question and context are shown in the Prompt section; the full prompt is hidden. When a Knowledge Manager or Admin have the Level 2 set (this can be done by Super Admins from Admin Dashboard), then they will see the full prompt, as Super Admin. #### 5.6.4 Chat logs The **Chat Logs** page displays a comprehensive list of chat conversations conducted within the application. Users can access this page by clicking the "View Chat Log" button located on the Reporting Dashboard. This feature is available to users with Knowledge Manager, Admin, and Super Admin roles. A chat log session commences when a user logs in and begins querying a specific assistant. All primary and follow-up questions within the same assistant are recorded in a single chat log session. When the user switches to a different assistant, a new chat log session is initiated. Additionally, if a chat log session begins on a different day, a new session will be created. The chat logs table includes the following details: * **Chat log id** * **Date** – the date and time when the conversation started * **Assistant** – the knowledge bot on which the conversation took place * **User** – The user that asked the questions from the conversation * **Actions** * **View details** The chat logs list can be filtered so that the user sees only the ones he is interested in: * **Date** – filter the conversations from that specific date * **Assistant** – A multiselect filter containing all accessible Assistants for the current user, with all options selected by default. * **User** – A multiselect filter featuring all users with access to the same assistants as the current user, also with all options pre-selected. **Note**: Knowledge Managers are limited to accessing conversations from assistants they have permission to use. Each table row includes a checkbox for selecting the corresponding chat log, as well as a header checkbox that allows users to select or deselect all logs on the current page. This selection facilitates the export of chosen chat logs. The **View details** action takes the user to **Conversation Details** page to view the detailed conversation, main and follow-up questions. The page contains a general details block with the following information: * Chat log id * Start date * Assistant * User Below details section, the full conversation of the chat log is displayed, main and follow-up questions, in the same order as they were asked, and for each main question, the follow-up ones are displayed: * Time Stamp * Language * Original Question * Follow-up Question * Rephrasing * Response - first 20 characters from response and **View response** button which opens the full response. * Feedback - Thumbs down /Thumbs up * Feedback Comment **Export functionality** On both the Chat Logs list page and the Conversation Details page, an "Export" button is available. This feature allows users to export details from selected chat logs (determined via checkboxes) or from the current chat log when accessed from the details page. Upon clicking the **Export** button, a notification informs the user that the export process has commenced. Once the file is ready, a follow-up message will confirm its availability, and the file is downloaded. The export file contains: * Chat log id * Assistant * Time Stamp * Language * Original Question * Follow-up Question * Rephrasing * Response * Feedback - Thumbs down (-1)/Thumbs up (1) * Feedback Comment The conversations and questions are exported in the same order as they took place: Main questions and their follow-up questions. ### 5.7 Configure data ingestion For Confluence, SharePoint or Salesforce integration, an initial configuration needs to be done. On Configure Data Sources page, Admin or Super Admin can configure them. The page contains the list with currently configured integrations for current application: * Data source type * Base Url * Actions * Edit * Delete The admin can create new ones, update the existing ones and delete them. When a new configuration is added, the Admin needs to provide de following details: * Data source type * Confluence * SharePoint * Salesforce * Base URL - needs to be unique for current application * Client Id * Tenant Id – only for SharePoint * Client Secret **Note**: Do not use "/" at the end of Base URL When adding a new configuration, the values are saved. When editing a configuration, the Client ID, Tenant Id and the Secret are not prepopulated with current values, and the user needs to provide the values again. ### 5.8 Data retention The system includes an automatic data retention mechanism that deletes user data, such as questions, answers, and feedback after a specified period. This period is set at the application level and, by default, is 180 days. ***Data to be Deleted*** * Conversations: Questions, answers, context, and rephrased questions. * Completion Data: Prompts, responses, feedback text, and grades. The deletion process is fully logged and auditable. Once the retention period expires, the data is marked as "soft deleted" with the date and time of deletion. Daily, the system records details of deleted items, and this information is accessible only to Super Admins for auditing purposes. ### 5.9 Data Anonymization and Security Given the sensitive nature of customer data handled by the chatbot, several measures are in place to ensure data privacy and security. * **Data Anonymization**: The chatbot applies data anonymization techniques to protect any personal or sensitive customer data it interacts with. For example, it may mask customer names, addresses, or account numbers in its responses, depending on the query and security policies in place. * **Access Control**: Role-based access (RBAC) ensures only authorized users (Agents, Knowledge Managers, Admins, Super Admins) can access specific features. Secure single sign-on (SSO) and Multi-Factor Authentication (MFA) may also be used. * **Compliance with Security Standards**: The chatbot operates within Concentrix Azure tenant, ensuring that it meets all necessary security and compliance standards, such as GDPR. Regular penetration tests and security assessments are performed by the company’s Information Security Team to ensure the chatbot remains secure and compliant with industry regulations. Future compliance targets include ISO27001 and SOC2 certifications. --- # Agent Instructions: Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://docs.ixhello.com/knowledge-bot/functional-documentation/functional-capabilities.md?ask= ``` The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.