AI Chat
The AI Chat is powered by large language models (LLMs) and retrieves and synthesizes information from multiple sources, including Infor documentation and industry datasets, through AI Chat agents and tools.
Users can generate summaries and other text-based responses, and access direct links to CloudSuite applications and data tables for visual insights.
The AI Chat provides guided conversation starters for new users and contextual follow-up suggestions to support continued interaction.
AI Chat onboarding
After installation of Infor Mobile Hub and the first successful sign-in, the Assistant Overview page appears once per installation.
The Overview page introduces key AI Chat features and includes a link to the User Guide and an AI usage disclaimer.
This page appears only the first time the application is used after installation and does not appear again on subsequent logins.
Starting a conversation
To start a conversation, click the icon or select the message input field.
Default prompt suggestions appear above the message input field. You can select one of these suggestions or enter your own prompt.
The button becomes active when text is specified in the input field.
Your messages appear on the right side of the conversation view, and AI responses appear on the left side.
A Thinking indicator appears while the response is being generated.
The message input field expands automatically up to eight lines. If additional text is entered, the input field becomes scrollable. The input field remains visible above the keyboard while typing.
GenAI Assistant on a mobile displays real‑time execution metrics for active agent runs within the streaming status area. During response generation, the assistant shows both elapsed duration and total token usage for supervisor and sub‑agent runs. It updates continuously alongside the streaming response without causing layout shifts. These values are rendered using compact formatting to fit within the constrained mobile UI.
Token usage is not displayed when an MCP tool is invoked because such tools do not consume tokens. Once the response is complete, the final duration and token values are persisted in the conversation and remain unchanged. This provides users with consistent visibility into response performance without affecting ongoing interactions or controls such as the button.
Agent-level traceability
When one or more agents participate in generating a response, GenAI Assistant displays the participating agent names as selectable items within the conversation. Selecting an agent opens a detailed view that shows the agent's contribution, execution status, execution duration, and token usage.
Agent status indicates whether the agent completed successfully, partially completed, or encountered a failure. Execution metrics are displayed only for the selected agent, providing visibility into its individual activity and contribution to the response.
Agent details and execution metrics are retained as part of the conversation history, which help the same information to be viewed when previously saved conversations are reopened.
Stop Response
GenAI Assistant on a mobile supports response interruption through a stop control in the message input area.
When the assistant is actively generating a response, the button is replaced by the button of the same size. When you use the button, it cancels the active response generation, including response streaming, in‑flight tool activity, and any active sub‑agent processing. The cancelled run is marked as cancelled in the session history, and no session reset is required.
If is selected while a tool is executing, the current response generation is interrupted. If you reopen the conversation later, the tool execution completes and the full response is available in the conversation history.
After cancellation is completed, the input area returns to the button state. The conversation remains in the same session, and you can continue with sending the next message in an ordinary way.
While the cancellation is already in progress, repeated selection of the button is ignored.
AI response actions and feedback
Each AI response includes a option.
Clicking copies the response message to your device clipboard.
Follow-up suggestions
Follow-up suggestions appear after each generated response when the Show Follow Ups option is enabled in .
Suggestions are displayed in a horizontally scrollable list.
When you select a follow-up suggestion, it is inserted into the message input field. Some suggestions insert directly, while others include placeholders that you can modify before sending the prompt.
Managing conversation history
You can access previous conversations from the side menu.
A scrollable panel displays your chat history, and a search bar allows you to filter conversations by name or content.
Press and hold a conversation to rename or delete it.
Deleting a conversation requires confirmation before the conversation is permanently removed.
Paginated loading of conversation history
When you open a previous conversation from Conversation History, AI Chat initially loads the latest twenty conversation records. This makes the most recent context to appear quickly without loading the entire chat history.
As the user scrolls upward, older messages are loaded incrementally in sets of twenty. A loading indicator is displayed while additional records are retrieved. On refresh or re-entry into the conversation, AI Chat again loads the latest twenty records first, with older messages available through upward scrolling.
Formatted responses
Responses generated by AI Chat can include formatted content such as links, tables, or code snippets.
Links open in your device’s default browser.
Tables and code blocks can be scrolled to view the full content when the response contains more information than can be displayed in the visible area.
Citations are displayed at the end of the response, along with the generated content, and are not embedded within the main response body. The display of citations is dependent on the prompt. Citations are displayed only when the prompt includes additional relevant information, such as links or documents.
These types of citations are supported:
- In-app citations: Open within the application, such as document previews (for example, PDF files) or document detail views.
- External links: Redirect to a web browser for accessing content within the mobile application.
Usage permission for the Review Required tool
When an AI agent requires the use of a tool marked as Review Required, AI Chat pauses execution and displays a Tool Approval Card requesting user permission. The card lists the required tool or tools and provides approve and deny actions.
If approved, the agent executes the tool and continues processing the request. If denied, the tool is not executed and the current request is terminated, with a message indicating that the action could not be completed without approval. For multiple tools, approvals are requested sequentially, and no restricted tool is executed without explicit user consent.
GenAI Assistant can request additional information from users during an active conversation when more context is needed to complete a task. Instead of continuing with incomplete information, the assistant pauses execution and displays a clarification card directly within the chat conversation. It may contain one or more questions and include these response types:
- Single-select options
- Multi-select options
- Free-text responses
When multiple clarifications are required, they are grouped into a single card. Each clarification is presented as an expandable section, which helps you to review and complete the requested information before continuing.
When a response to a clarification request is needed, review the clarification questions displayed in the conversation and provide the requested information for each section. If the Other option is available, select it to specify a custom response and click to send all responses.
After submission, the assistant automatically resumes processing from the point where clarification was requested.
Completed clarification sections display a visual indicator showing that the required information has been provided. If required information is missing, the clarification cannot be submitted until all mandatory responses are completed.
Users may choose not to provide the requested clarification. In this case, the assistant may cancel the current action or respond based on the available information.
With this interaction, the GenAI Assistant gathers required details in a single step, reducing unnecessary follow-up questions and improving task completion accuracy.
In-line drill back hyperlink support
GenAI supports multiple in‑line drill back hyperlinks within assistant responses, which helps navigation links to be rendered directly inside generated content.
This capability enables APIs or agents to return data sets such as tables or lists with per‑item drill back links embedded inline, rather than as separate response attachments displayed below the response.
When deep agent support is enabled, agents can generate responses that include in‑line drill back hyperlinks. Selecting an in‑line drill back link triggers a navigation action through the standard navigation mechanism, even if the target does not resolve to an actual screen.
This enhancement improves usability by presenting actionable navigation links directly within generated tables or text, providing a more integrated and context‑aware experience.
Background responses and notifications
If you start a new conversation while another response is still being generated, the original conversation continues processing in the background.
When the response is complete, a toast notification appears indicating that the response has finished generating.
The notification message displays this message:Response completed. Tap to open.
Selecting the notification opens the completed conversation.
Toast notifications appear above the top navigation bar and display a brief message.
Shimmer placeholders are displayed while conversation history, content, or suggestions are loading.
Synchronizing conversations across devices
Conversations are automatically synchronized between the mobile and web versions of AI Chat.
This synchronization allows you to continue conversations seamlessly across devices.
Conversation history with Deep Agent enabled
Conversation History behavior in AI Chat is based on whether Deep Agent is enabled. The same conversation history endpoint is used regardless of Deep Agent state.
When Deep Agent is disabled, the system uses Schema Version 1 (V1) conversation messages. When Deep Agent is enabled, conversation history uses Schema Version 2 (V2) Deep Agent messages.
About AI Chat
Within the Conversation History side panel, an option labeled About AI Chat is available in the footer section. This option provides you with informational details about the AI Chat assistant.
When you select About AI Chat, a pop‑up dialog is displayed. The dialog presents relevant information about AI Chat, including details about the assistant and its purpose. When you close the dialog, you return to the current conversation without affecting chat history or interaction.
Endpoint selection based on Deep Agent flag
AI Chat automatically selects the appropriate GenAI endpoint based on the Deep Agent feature flag. When the feature flag is enabled, the app uses the /api/v2/chat endpoint. When disabled, the app falls back to the V1 endpoint, with no visible UI changes.
The V2 endpoint supports both streaming responses using the /messages/stream endpoint and synchronous responses. Endpoint switching is handled at runtime with no visible UI changes, and safeguards are in place to prevent cross‑session response echoing.
Focus modes in AI Chat
AI Chat supports focus modes that help you to interact with a predefined collection of agents. Focus modes are enabled using the FOCUSMODE feature flag and are applied by including a valid FocusMode value in the v2 or chat request.
When enabled and available, a plus (+) icon appears in AI Chat. You can select available focus modes using the button, which opens a panel for browsing and selecting focus modes. The panel supports scrolling and expansion to view the full list of focus modes.
Agent and focus mode selection
GenAI Assistant on mobile provides multiple ways to quickly select an Agent or Focus Mode directly from the chat input.
You can either select the + option in the AI Chat input area or specify @ to access a searchable list of available agents and focus modes. Only published and accessible items are displayed. Agents and focus modes are visually identified to help users distinguish between them.
Selecting the + option displays the agents and focus mode options. Selecting either option opens a searchable list of available agents or focus modes. When you specify @, a searchable suggestion list is displayed. The suggestion list filters matching results as you specify additional text.
Selecting an agent or focus mode from either the + menu or the @ suggestion list inserts it into the chat input as a badge. Messages sent with a selected agent are routed directly to that agent. Messages sent with a selected focus mode use the selected focus mode configuration for the conversation.
Selecting a new agent or focus mode replaces the current selection. The selection can be changed at any time without affecting the existing conversation session or chat history.
The @ suggestion list is available in both new and existing conversations and closes automatically when a selection is made or when the @ trigger is removed. (TECDOC-32659)
Upload and manage attachments in AI Chat
GenAI Assistant on mobile helps you to upload attachments from the chat input area using the + icon.
When you select the Attachment field, you can choose files from the supported sources such as Camera, Gallery, and Browser. One or more supported attachments can be added to the same message, including attachments from different sources, up to the supported limits.
These attachment limits apply:
- Up to five documents can be attached to a single message. Each document must not exceed 4.5 MB in size.
- When attaching one or more documents, the message must also include text.
- Attachments can only be included in user messages.
Selected attachments are displayed above the message input area before submission. You can remove individual attachments without affecting the remaining files. When the message is sent, all selected attachments are included with the request to the assistant.
After a successful submission, the selected attachments are cleared from the input area.
If message submission fails, all selected attachments remain available for retry. Files attached without sending a message are uploaded and retained for up to two days, after which they are automatically deleted.
You can return to a conversation from chat history and continue asking questions about attached files while the files remain within the retention period.
Unsupported file types, files that exceed the supported size limits, or attachments that exceed the maximum allowed number are rejected, and an appropriate validation message is displayed.
Refreshing conversation starter suggestions in AI Chat
GenAI surfaces conversation starter default suggestions to help users initiate interactions efficiently. When an administrator configures more than four default suggestions for a tenant, the GenAI interface displays only the first four suggestions by default.
When additional suggestions are available, a is displayed beneath the conversation starter suggestions. Selecting the loads the next set of up to four default suggestions based on the configured order.
Each subsequent selection of Refresh control advances through the remaining configured suggestions in groups of up to four. After all configured suggestions have been displayed, selecting Refresh again cycles the experience back to the first four default suggestions.
This interaction model ensures a bounded UI experience while you can access the full set of administrator-configured Conversation Starter suggestions in a predictable, rotational manner.
Limitations
Generated responses depend on available data and system configuration.
Responses can contain inaccuracies and should be verified before acting on them.
Offline use of AI Chat in Infor Mobile Hub is not supported.
Privacy and data handling
Use of AI Chat follows organizational privacy and security policies.
Diagnostic and usage information can be collected according to user's choice.
Refer to organizational policy documentation for details about data handling and retention.
Conversation data tracing and opt-out behavior
Conversation data tracing for AI Chat is controlled by the Opt-out of data tracing setting in the Infor OS Portal under GenAI.
If the option is disabled, conversation tracing remains enabled and AI Chat interactions continue to be visible in LangSmith. If the option is enabled, AI Chat conversation data is not traced and does not appear in LangSmith. This setting applies to new conversations and supports organizational privacy and compliance requirements.
AI Chat feedback
GenAI Assistant helps you to provide feedback on individual assistant responses through and actions displayed alongside the action.
Feedback helps improve response quality and provides a mechanism for reporting issues or suggesting enhancements related to AI-generated responses.
If you want to provide a positive feedback, select the icon on an assistant response. Optionally, you can leave a comment and click .
Positive feedback is associated with the selected response and helps identify useful and effective AI interactions.
Negative feedback
To provide negative feedback, click the icon on an assistant response and specify one of the available feedback methods:
- Quick Feedback: Helps you to report common response issues by selecting a feedback category and providing a description. Available categories include:
- Incorrect
- Incomplete Response
- Harmful Content
- Performance Issue
This option is intended for fast feedback submission directly from the conversation.
- Detailed feedback: Provides additional options for reporting issues or suggesting improvements. You can select one of these:
- Bug: Use to report problems with a response that require investigation. You can provide these details:
- Summary
- Priority
- Issue Category
- Description
- Steps to Reproduce
- Related Issue
- Supporting Attachments
When submitted, a ticket is created and associated with the selected assistant response. Relevant diagnostic information may be automatically included to assist with investigation.
- Feature Request: Use to suggest enhancements or new capabilities. You can provide:
- Summary
- Priority
- Description
- Related Issue
- Supporting Attachments
When submitted, a ticket is created to track the enhancement request.
- Bug: Use to report problems with a response that require investigation. You can provide these details:
Ticket lifecycle
Detailed feedback submissions are tracked through a ticket. The ticket id is created with tenant_id-sequence number. You can review the information included in the submitted ticket. If a ticket is resolved and later moved to a Fixed or Closed state, it cannot be reopened. If additional investigation or follow-up is required after closure, a new ticket must be submitted.
Information about ticket status and follow-up actions is available within GenAI admin settings.
In-line thoughts preference
When Deep Agent capabilities are enabled, with the GenAI Assistant you can control how much execution detail is displayed during AI interactions through the in-line thoughts setting in the User Profile settings.
The available options are determined by your administrator and may include:
- Minimal – Displays a thinking indicator with execution time and token usage while a response is being generated. The indicator is removed after the response is complete.
- Standard – Displays concise, plain-language summaries of agent activity along with execution time and token usage. Thoughts are displayed during execution time.
- Detailed – Displays the full execution trace, including agent and tool-level details. This option is available only to authorized administrative users.
To choose an in-line thoughts preference, select . The selected preference is applied immediately and used across all mobile assistant experiences.
The selected preference is stored for the current user and tenant. The preference is retained across sessions, application restarts, and future sign-ins.
Administrator and policy controls
Only the options enabled by the administrator are displayed. If an administrator removes a previously selected option, the setting automatically reverts to the administrator-defined default.
If Trace Opt-Out is enabled for your organization by the administrator, the effective setting is automatically set to Minimal and cannot be modified.
The In-line Thoughts preference changes only how execution information is presented in the user interface. It does not affect agent behavior, planning, tool execution, response generation, or token consumption.
Best practices for using AI Chat
Provide clear and specific prompts to improve response quality.
Verify generated information before taking action.
Confirm that the correct profile is active before submitting prompts.
Avoid entering sensitive information unless required by organizational policy.
Contact your administrator if AI Chat is unavailable or does not return expected results.
New Chat
When you select in AI Chat, the application starts a new conversation. Draft message is cleared and any previously selected Focus Mode is removed. A new chat session contains no draft text, composer state, or Focus Mode configuration. If you navigate to another conversation through Conversation History and then return, the unsent draft message is retained for that conversation. The icon remains enabled, so you can start a new conversation at any time.