The Insights Agent is your conversational partner in data analysis. Like any large language model, its quality scales with the quality of your prompts — think of it as a capable junior analyst: the more context and direction you give, the sharper the output.Documentation Index
Fetch the complete documentation index at: https://docs.caplena.com/llms.txt
Use this file to discover all available pages before exploring further.
The Golden Rule: Context Is Everything
The Insights Agent doesn’t intuitively know your intent — it reacts to the data and the prompts you provide. To get precise, actionable answers, include as much context as possible: the timeframe, the specific category, or the survey column you want to analyze.Iteration tip: Treat the first answer as a starting point. If it’s too broad, follow up: “That’s interesting — can you show me specific verbatim examples for that negative trend?”
How the Agent Shapes Its Responses
The Insights Agent doesn’t use a fixed template. It dynamically selects the visualizations, charts, or summaries that best answer your specific prompt. A simple question might return a clean text summary. A broader request about “most common issues” will automatically generate a multi-layered analysis with statistical trends, sentiment breakdowns, and representative customer comments.What to Do: Master the Conversation
These strategies consistently produce better, more targeted answers.| Strategy | Why it works | Example prompt |
|---|---|---|
| Onboard yourself first | Quickly understand the scope of a new project | ”What data can I see in this project?” or “What are the most common themes here?” |
| Define your persona | Tailors output to your specific role and priorities | ”My role is Product Manager for the web app. What feedback is relevant for me?” |
| Be ultra-specific | Prevents generic answers by targeting exact data points | ”Compare sentiment on ‘Ease of Use’ between Q1 and Q2 for Enterprise users.” |
| Explain by example | Sets a clear template for format or tone | ”Summarize in three bullet points, then one Key Action Item for the dev team.” |
| Check active filters | Ensures you’re analyzing the right data subset | Check the system note in chat to see which dashboard filters are currently active. |
What to Avoid: Common Pitfalls
Watch out for these habits:- Blind trust: The Agent is highly accurate, but it augments your expertise rather than replacing it. If an insight contradicts your intuition, investigate it — ask for an explanation of its reasoning or cross-reference the raw data.
- Ignoring filters: If results feel “off,” a dashboard filter is often still applied. Always verify active filters in the Meta Text at the start of your response.
- Expecting magic knowledge: The Agent can only analyze data uploaded to Caplena. It cannot access internal documents, Slack messages, or external market trends unless they are part of the current project.
- Vague terminology: Avoid pronouns like “it” or “that” when the conversation has covered multiple topics. Use explicit category or segment names to keep the Agent on track.
- Using it as a manual calculator: Don’t ask the Agent to do arithmetic on numbers already shown in chat. The Agent provides pre-calculated Key Quantitative Findings from the raw dataset — manual math in a chat thread introduces errors. For value mappings or mathematical operations, use the Smart Column feature instead.