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

# Understand

> How Adapt investigates your questions and delivers trusted answers

The second step in Adapt's Ask-Understand-Act framework. After you ask a question, Adapt investigates across your connected data sources to provide trusted, cited answers.

## How Adapt Investigates

When you ask a question, Adapt's AI agent:

<Steps>
  <Step title="Plans the approach">
    Determines which data sources to query and what information is needed
  </Step>

  <Step title="Gathers data">
    Queries your connected tools—Snowflake, HubSpot, Stripe, etc.
  </Step>

  <Step title="Analyzes results">
    Processes the data, performs calculations, and identifies patterns
  </Step>

  <Step title="Synthesizes answer">
    Combines findings into a clear, comprehensive response with citations
  </Step>
</Steps>

## Trusted Answers

### Source Citations

Every answer includes citations to the data sources used:

```
Your MRR grew 12% last month, from $98,400 to $110,200.

Sources:
• Stripe subscription data (as of today)
• 47 active subscriptions analyzed
```

You can always trace an insight back to its source.

### Data Transparency

When Adapt queries your data, you can see:

* **Which tools** were accessed
* **What queries** were run
* **Raw data** that informed the answer

Click "Show details" on any response to see the full investigation.

## Multi-Source Analysis

Adapt can combine data from multiple sources in a single answer:

**Question:**

```
Which customers are at risk of churning?
```

**Adapt investigates:**

| Source         | Data Retrieved                                  |
| -------------- | ----------------------------------------------- |
| **Stripe**     | Payment history, failed charges, plan changes   |
| **HubSpot**    | Last contact, support tickets, engagement score |
| **Product DB** | Login frequency, feature usage, activity trends |
| **Intercom**   | Recent conversations, sentiment                 |

**Combined insight:**

```
5 accounts show churn risk signals:

1. Acme Corp - No login in 45 days, 3 open support tickets
2. Beta Inc - Downgraded plan, payment failed twice
3. Gamma LLC - Usage dropped 80% this month
...
```

## Analysis Capabilities

### Data Processing

Adapt can process and transform data:

* Aggregations (sum, average, count)
* Grouping and segmentation
* Time-series analysis
* Percentage calculations
* Trend identification

### Statistical Analysis

For deeper questions, Adapt performs statistical analysis:

* Correlation analysis
* Cohort comparisons
* Significance testing
* Forecasting and projections

### Code Execution

When complex analysis is needed, Adapt runs code in a secure sandbox:

```python theme={null}
# Adapt might run analysis like this:
import pandas as pd

# Load and process data
df = pd.read_csv('sales_data.csv')
monthly = df.groupby('month')['revenue'].sum()
growth = monthly.pct_change()

# Calculate insights
avg_growth = growth.mean()
best_month = growth.idxmax()
```

You can see the code that was executed and verify the logic.

## Visualization

Adapt automatically generates visualizations when helpful:

### Charts

Request specific chart types:

```
Show me a line chart of revenue by month
```

```
Create a bar chart comparing regions
```

```
Visualize the distribution of deal sizes
```

### Tables

Complex data is formatted into sortable, searchable tables:

| Customer  | Revenue  | Growth | Risk Score |
| --------- | -------- | ------ | ---------- |
| Acme Corp | \$45,000 | +12%   | Low        |
| Beta Inc  | \$32,000 | -5%    | Medium     |
| Gamma LLC | \$28,000 | -15%   | High       |

### Dashboards

For comprehensive views, Adapt can create multi-visualization dashboards combining charts, tables, and key metrics.

## Web Research

For questions requiring external information, Adapt searches the web:

```
What are the latest trends in B2B SaaS pricing?
```

Adapt will:

1. Search for recent articles and reports
2. Extract relevant information
3. Summarize findings with source links
4. Combine with your internal data if relevant

## Deep Research

For complex research questions, Adapt can spawn sub-agents to investigate in parallel:

**Question:**

```
Do a competitive analysis of our top 3 competitors
```

**Adapt's approach:**

* Sub-agent 1: Research Competitor A
* Sub-agent 2: Research Competitor B
* Sub-agent 3: Research Competitor C
* Main agent: Synthesize findings into comparison

This parallel approach delivers comprehensive research faster.

## Knowledge Base

Adapt maintains a knowledge base of your organization's context:

* **Definitions**: How you define metrics like "active user" or "MRR"
* **Business rules**: Your fiscal year, quota structures, etc.
* **Preferences**: How you like reports formatted
* **History**: Learnings from past conversations

This context helps Adapt give more relevant, accurate answers.

Learn how to manage this context in the [Knowledge](/features/knowledge) guide.

## Verification

For critical insights, Adapt can verify claims:

1. Cross-reference multiple data sources
2. Check for data quality issues
3. Flag any inconsistencies
4. Note confidence levels

```
Revenue grew 15% last quarter.

Verification:
✓ Confirmed in Stripe data
✓ Matches Snowflake records
✓ Consistent with HubSpot deal data
Confidence: High
```

## Next Steps

Once Adapt understands your question and provides an answer, you can:

<CardGroup cols={2}>
  <Card title="Ask Follow-Ups" icon="comments">
    Drill deeper with follow-up questions—context is maintained
  </Card>

  <Card title="Take Action" icon="bolt" href="/features/act">
    Have Adapt act on the insights—draft emails, update records, etc.
  </Card>
</CardGroup>