Conversations¶
This workflow covers how to interact with AI Core through the floating chat panel, including starting conversations, asking questions, confirming actions, and reviewing conversation history.
Overview¶
AI Core is accessible from any screen in Progrid via a floating chat panel. You can ask questions about your data, request record creation or updates, and get contextual help based on what you are currently viewing.
Required permissions¶
AI User to start conversations and interact with the assistant
Start a conversation¶
Press Ctrl+Space from any screen to open the AI Core chat panel. A new conversation starts with a welcome message that includes context about the screen you are on.
Alternatively, navigate to to see all your past conversations and start new ones from the list view.
Ask questions about your data¶
Type natural language questions to query your Progrid data. The AI translates your question into database queries, respecting your permissions and the model whitelist.
Examples:
“How many open opportunities do we have?”
“Show me the top 5 customers by revenue”
“What is the average deal size this quarter?”
“List all contacts at Acme Corp”
“How many invoices are overdue?”
The AI uses the search_records, count_records, and aggregate_records tools to find
answers. Results are presented in a readable format within the chat.
Request actions¶
You can ask the AI to create or modify records. All write operations require your explicit confirmation before execution.
Examples:
“Create a new contact named Jane Doe at jane@example.com”
“Update the expected revenue on opportunity #42 to $50,000”
“Set all leads in the ‘New’ stage assigned to John to ‘Qualified’”
When the AI determines it needs to write data, it presents the proposed action with details of what will change. You see two buttons:
Confirm – Execute the action
Cancel – Reject the action and continue the conversation
Note
The AI can only modify models that have been whitelisted with write, create, or delete permissions in . If a model is read-only, the AI will explain that it cannot perform the requested action.
Contextual assistance¶
AI Core is aware of what you are currently viewing. When you open the chat panel from a specific record (e.g., a partner form or an opportunity), the AI includes that record’s context in the conversation. This allows you to ask context-specific questions like:
“What invoices does this customer have?” (from a partner form)
“What activities are scheduled for this opportunity?” (from a CRM lead form)
“Show me similar records” (from any record)
The context includes the current model, record ID, and key field values.
Review conversation history¶
Navigate to to see all your conversations. Each conversation shows:
Title – Auto-generated from your first message
Message Count – Total messages in the conversation
Last Activity – When the last message was sent
Status – Active, Archived, or Error
Click a conversation to view its full message history. Messages show role indicators (User, Assistant, or Tool) and timestamps.
Archive a conversation¶
To archive a conversation you no longer need:
Open the conversation from the list.
Click the Archive button in the header.
The conversation status changes to
archivedand it is hidden from the default list view.
Use the Archived filter in the search bar to view archived conversations.
Tips for effective use¶
Be specific. “Show me open opportunities over $10,000 assigned to the Sales team” works better than “Show me deals.”
Use field names. If you know the field name, include it: “What is the expected_revenue on lead 42?” is unambiguous.
Chain requests. The AI maintains context within a conversation. Ask a follow-up question about results from a previous query.
Check the model whitelist. If the AI says it cannot access a model, ask your administrator to add it to the whitelist.
Next steps¶
Model Access Control – Configure which models are available to the AI
Knowledge Management – Add documentation and procedures to the knowledge base
Security – Understand rate limits and audit logging