Insights
Insights is an AI agent that lets you ask natural-language questions about the conversations your agents have already had with customers. Instead of digging through the inbox manually, just ask — Insights analyzes the data and answers with text, tables, and charts.
What you'll learn
- What Insights is and how it differs from your agents
- How to ask a question
- How to read a streamed answer with charts
- How to answer follow-up questions from the agent
- How to manage your past insight tasks
What is Insights?
Insights queries the history of conversations between your customers and your agents. It's built to answer analytical questions — trends, summaries, and rankings — based on conversations that already happened.
Typical questions include:
- What questions do customers ask most often?
- Which product gets discussed the most?
- What topics are trending right now?
Insights is different from your agents. An agent answers questions using your database or knowledge base in real time (e.g. "Which items are out of stock?" or "What's the stock for Product A?"). Insights instead analyzes conversations that already took place, to help you spot patterns and trends across all of them.
The Insights app UI is currently in Bahasa Indonesia — placeholders, empty states, and progress messages appear in Indonesian even when this guide is in English.
Insights is available to directors — workspace Owners and Admins — from the Insights item in the sidebar.
Ask a question
Open Insights from the sidebar. You'll see a text box where you can type any question in plain language, plus a few suggested questions to get you started.
- Type your question in the text box
- Press Enter or click the send button
- The agent starts working in the background — you'll see it appear under "Recent" with a Working status badge
Be specific about time ranges and scope for the best results — e.g. "How many customer messages did we get this week?" or "Who are our top 10 most active customers this month?"
Read the answer
Insights streams its answer in real time, word by word, similar to a chat assistant. For questions that require scanning a large number of messages (like finding the most-asked topic), it processes every matching message rather than a small sample, so you'll see a progress bar while it works.
Depending on the question, the answer may include a chart or table alongside the written explanation — for example a bar chart of channel breakdown, or a ranked list of top askers.
Each task shows a status badge: Working (in progress), Needs Input (the agent is asking you something), Done (answered), or Failed.
Answer follow-up questions
Sometimes Insights needs more context before it can give an accurate answer — for example, if your question doesn't specify a time range. When that happens, the task status changes to Needs Input and the agent's question appears in the thread.
- Read the agent's clarifying question
- Type your answer in the reply box at the bottom
- Send it — the agent resumes and continues its analysis
You can also keep the conversation going after the agent has answered — just type a follow-up question or ask it to dig deeper into a specific part of the answer.
Manage past tasks
Every question you ask is saved as a task under "Recent" on the Insights home screen. Click any task to reopen its full thread — including the original question, the streamed answer, and any charts.
Open a task and click the trash icon in the header to permanently delete it if you no longer need it.
Expected outcome
You now know how to ask Insights natural-language questions about your customer conversations, read streamed answers with charts, respond to clarifying questions, and manage your past tasks.
That's it — you've completed the Orxa product guide! You now know how to create agents, add knowledge, connect integrations, manage conversations, work with your team, and use Insights to analyze your conversation data.