Skip to main content
Omnifact is designed to work like a natural conversation, allowing you to communicate with the just as you would with a colleague. Understanding how conversation flow works in Omnifact will help you get the most out of your interactions.

​
How Conversation Works in Omnifact

Conversation Flow Example
In Omnifact, conversations follow a natural back-and-forth pattern just like chatting with a knowledgeable colleague. What makes this powerful is that the agent keeps track of your entire conversation history within a chat session, allowing for follow-up questions without needing to repeat information. This natural conversation pattern helps you efficiently navigate from broad questions to specific, actionable information, making even complex topics more accessible and your interactions more productive.
Pro Tip: You don’t need to repeat yourself! Omnifact remembers what you just said.Instead of asking: “Summarize the Q3 Financial Report PDF” and then “What were the risks in the Q3 Financial Report PDF?”You can just say: “What were the risks?”

​
The Power of Follow-Up Questions

Follow-up questions are key to getting the most out of Omnifact. They allow you to:

Dive Deeper

Start with broad questions, then use follow-ups to explore specific aspects of the topic.

Clarify Information

If the AI’s response isn’t clear or complete, ask follow-up questions to get the information you need.

Build on Responses

Use the AI’s answers as a foundation for your next questions, creating a productive chain of inquiry.

Refine Results

Gradually narrow down broad answers to get precisely what you’re looking for.

​
Follow-Up Question Examples

Here are examples of effective follow-up questions in different scenarios:

Example: Marketing Campaign Planning

Initial Question: “Can you help me brainstorm ideas for our summer product campaign?”Effective Follow-ups:
  • “Which of these ideas would work best for our target demographic of young professionals?”
  • “How should we adapt the messaging for social media channels specifically?”
  • “Can you suggest a timeline for implementing the third idea you mentioned?”
Initial Question: “Can you summarize the key points from the quarterly report I just attached?”Effective Follow-ups:
  • “What are the most significant changes compared to last quarter?”
  • “Can you explain more about the customer retention statistics you mentioned?”
  • “What recommendations would you make based on these findings?”
Initial Question: “Can you explain what ‘data normalization’ means in database management?”Effective Follow-ups:
  • “What are the benefits of normalizing data?”
  • “Can you give me a simple example of how to normalize a database?”
  • “What are the potential downsides of over-normalizing?”

​
Tips for Effective Conversations

Be Specific

The more specific your questions, the more relevant Omnifact’s answers will be.

One Topic Per Chat

For optimal results, start a new chat when switching to a different topic.

Try Different Approaches

If a response isn’t helpful, ask Omnifact to “explain it differently” or “try another approach.”
Remember that follow-up questions work best when they’re related to the current conversation. For completely new topics, starting a fresh chat will give you better results.

​
Context Window Warnings

Every AI model has a maximum memory limit, known as its “context window”. This dictates how much of the previous conversation it can remember at any one time. As your conversation grows very long, you may see a warning banner appear indicating that you are approaching the model’s context limit.
  • When this happens, the AI may begin to “forget” the earliest messages in your chat.
  • While the conversation can continue, you might notice a drop in response quality, speed, or context awareness.
  • For optimal performance, we recommend starting a new chat when you see this warning, summarizing any necessary context from the previous chat into your first prompt.

​
Agent Actions Timeline

When you send a request to Omnifact, you can observe the step-by-step process in real time:
  • Action Indicators: During response generation, an ordered list of actions appears above the streaming text.
  • Types of Skills: You might see the AI Thinking, Searching the web, Fetching URL, Retrieving from Uploaded Files, Generating image, Running code, Delivering files, or calling connected MCP tools.
  • Transparency: Each action shows relevant context, like the exact search query being run, the code executed, or the number of documents retrieved. This transparency helps you verify that the AI is gathering the correct information before it formulates its final answer.

​
In-Chat Approvals for State-Changing Actions

When an autonomous agent attempts to run an external tool configured with an Approval policy (such as updating a CRM record or deleting a database entry via MCP Servers):
  • Interactive Banner: The agent pauses execution and displays a Tool Approval Banner directly above the chat input.
  • Inspect Request: You can expand the banner to review the tool name, target integration server, and parameter payload.
  • Allow or Deny: Click Allow to approve the action or Deny to reject it. If denied, the agent is informed and adapts its response without modifying any external data.
Tool approval banner above the chat input asking to allow a Jira MCP tool, with Deny and Allow buttons

​
Connecting Workplace Tools for Chat

When a tool requires you to connect your own account (such as HubSpot, GitLab, or Jira):
  • Connect to use: Tools become available in chat once you connect your account in Account Settings > Connections.
  • Reconnecting tools: If a tool connection expires, the agent will let you know in chat and direct you to your Connections settings to reconnect.

​
Next Steps

Now that you understand how conversation flow works in Omnifact, you can: