> ## Documentation Index
> Fetch the complete documentation index at: https://docs.arya.so/llms.txt
> Use this file to discover all available pages before exploring further.

# Arya Report Chat

> Ask about your data in natural language — answers with charts and insights.

**Arya Report Chat** is AI embedded in reports that answers natural-language questions with charts, tables and insights.

<Frame>
  <img src="https://mintcdn.com/agevole/cM9eEPJE8FVhTNOb/images/report/en/report-chat.svg?fit=max&auto=format&n=cM9eEPJE8FVhTNOb&q=85&s=1911c31b7a14c0cb076027afd1eb76e5" alt="Chat window with user question and response with inline chart" width="1200" height="720" data-path="images/report/en/report-chat.svg" />
</Frame>

## How it works

<Steps>
  <Step title="Open a dashboard">Any dashboard (or the whole workspace) → click **Ask Arya** top right.</Step>

  <Step title="Ask the question">
    Write a question in English or Italian. Examples:

    * *"Why is Q2 down 15% vs Q1?"*
    * *"What's the average conversion rate of my reps?"*
    * *"Show the 5 customers with highest margin this year"*
    * *"Is there a negative trend in a specific geography?"*
  </Step>

  <Step title="Arya analyses">
    Queries the underlying data (respecting user permissions), computes, identifies patterns, generates response.
  </Step>

  <Step title="Rich response">
    Output includes:

    * **Explanatory text** (why the result is what it is)
    * **Inline charts** (if the question needs them)
    * **Tables** with detail
    * **Suggestions** (follow-up questions)
  </Step>

  <Step title="Iterate">
    Follow up with more questions — Arya keeps conversation context.
  </Step>
</Steps>

## Types of questions that work well

<AccordionGroup>
  <Accordion title="Quantitative questions">
    *"How many deals did I close won this quarter?"*
    *"What's the average deal value in Enterprise?"*
  </Accordion>

  <Accordion title="Trends and comparisons">
    *"Has revenue grown or declined in the last 6 months?"*
    *"Compare Mario vs Luigi's performance in the last 3 months"*
  </Accordion>

  <Accordion title="Causal analysis">
    *"Why did I lose so many deals at the Negotiation stage?"*
    *"What's the main reason for customer churn?"*
  </Accordion>

  <Accordion title="Segmentation">
    *"Which industries am I selling best in?"*
    *"Show customers with ARR > 50k but no activity in the last 30 days"*
  </Accordion>

  <Accordion title="Forecasts">
    *"End-of-quarter closing projection based on current pipeline"*
    *"How many new customers can I expect next month at current trend?"*
  </Accordion>
</AccordionGroup>

## Costs (AI credits)

Every question consumes credits proportional to complexity:

| Question type             | Estimated credits |
| ------------------------- | ----------------- |
| Simple query (count, sum) | 2-5               |
| Analysis with charts      | 10-20             |
| Deep causal analysis      | 20-50             |
| Forecasts                 | 30-80             |

See [AI credits](/en/arya-101/lesson-10-ai-and-aryachat).

## Accuracy and verification

<Warning>
  Arya does its best but can make mistakes. Always verify critical data by cross-checking with the original report. AI:

  * **Knows**: statistics from your workspace, numeric comparisons, patterns
  * **Can be wrong on**: complex causal inferences, forecasts with little data
  * **Doesn't know**: external context not in the data (market, competitors, events)
</Warning>

Every response includes a **"Show me the data"** option to see how Arya reached the conclusion.

## Privacy

Questions and data aren't used for AI model training. Every query is isolated to your workspace and respects the asking user's permissions.

## Best practices

* **Be specific**: "this quarter" beats "recently"
* **One thing at a time**: short questions → precise answers
* **Iterate**: first question for lay of the land, then dive deeper
* **Verify**: for important decisions, export data and sanity check

## Frequently asked questions

<AccordionGroup>
  <Accordion title="Can I share an Arya response with my team?">Yes. Every response has a unique shareable link (requires workspace login).</Accordion>
  <Accordion title="Can Arya generate new dashboards?">Not directly, but when it generates charts you can click **Save as dashboard** to persist them.</Accordion>
  <Accordion title="Does it work on sensitive data (salaries, invoices)?">Yes, respecting permissions. If a user has no access to salary data, Arya doesn't include it in responses.</Accordion>
</AccordionGroup>
