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

# How to Use Ask AI

> Use Formo's AI assistant to explore analytics data with natural language questions and generate SQL queries automatically.

Ask natural language questions about your app and get instant answers. The AI generates SQL queries, runs them, and returns a chart and summary.

## Part 1: Open Ask AI

Click **"Ask AI"** in the left sidebar (chat bubble icon) to open the chat interface. You'll see your conversation history on the left and the chat window on the right.

<Frame>
  <img src="https://mintcdn.com/formo/5S3MPZWP8GYgbzVs/images/ai.png?fit=max&auto=format&n=5S3MPZWP8GYgbzVs&q=85&s=c50d8476296fc92e85607f7651c5858d" alt="Ask AI interface with chat history" width="963" height="441" data-path="images/ai.png" />
</Frame>

## Part 2: Ask Your First Question

Start with something straightforward. Click in the message box and type a question:

**Example:** "How many unique wallets connected this week?"

Hit enter. The AI will:

1. Parse your question
2. Generate a SQL query from it
3. Run the query against your data
4. Return an answer, with a chart when relevant

<Frame>
  <img src="https://mintcdn.com/formo/5S3MPZWP8GYgbzVs/images/text-to-sql.png?fit=max&auto=format&n=5S3MPZWP8GYgbzVs&q=85&s=dd135ef7dbfbe3bb19cdb1604c082448" alt="AI generates SQL and shows chart result" width="418" height="191" data-path="images/text-to-sql.png" />
</Frame>

## Part 3: Understand the Response

Each Ask AI response includes:

* **A plain English answer** summarizing what the data shows
* **A chart or dashboard proposal**, when the question calls for one
* **"How I calculated this"**: a collapsible section showing the SQL and tool calls the AI used to get the answer

## Part 4: Ask Follow-Up Questions

Ask follow-ups without repeating context:

**You:** "How many unique wallets connected this week?"
**AI:** \[Shows 12,453 wallets]

**You:** "How many of those were new?"
**AI:** \[Filters to new wallets. Understands "those" = the wallets from last question]

**You:** "What were their top referral sources?"
**AI:** \[Breaks down new wallets by referrer. Still understands the context]

The AI maintains conversation memory, so each question builds on the previous one. No need to re-ask context.

## Part 5: Use the Insights Page for Automated Analysis

Beyond Ask AI, visit **Insights** in the sidebar for AI-generated analysis of your app. It has two parts: a summary of wins, issues, and opportunities, and a cohort analysis across acquisition, revenue, and churn. See [Insights](/features/product-analytics/insights) for the full breakdown.

<Frame>
  <img src="https://mintcdn.com/formo/d1KXaJbWihiCN18P/images/insights.png?fit=max&auto=format&n=d1KXaJbWihiCN18P&q=85&s=edf3493fc689c40a60e9bc1fc9a67b09" alt="AI-generated insights page" width="1586" height="795" data-path="images/insights.png" />
</Frame>

| Section                 | Answers                                         |
| ----------------------- | ----------------------------------------------- |
| **Wins**                | What improved recently, with deltas             |
| **Issues**              | What's underperforming and needs attention      |
| **Opportunities**       | Untapped segments or behaviors                  |
| **Acquisition Quality** | Which channels bring in the most valuable users |
| **Revenue Insights**    | Weekly revenue/volume trend vs. the prior week  |
| **Churn Prediction**    | Which behaviors signal a user is about to churn |

## Part 6: Save Your Questions as Dashboards

To check a question weekly instead of re-asking it:

1. Ask your question in Ask AI
2. Click **Add to Dashboard**
3. Choose an existing board (or it saves directly if you only have one)

The chart stays on your dashboard and updates with new data whenever anyone on your team opens it.

## Example Questions

<Note>
  Be specific with dates and metrics. Instead of "How are we doing?" try "What's my DAU compared to last week?" or "Show me conversion by traffic source for the last 30 days."
</Note>

### User acquisition

**Prompt:** "Where are my users coming from this month?"

**What you'll learn:** Top referrers, UTM sources, and traffic channels.

**Follow-up:** "Which referrer has the best conversion rate?"

### Active users

**Prompt:** "How many active users did we have last week?"

**What you'll learn:** DAU trend with day-over-day changes.

**Follow-up:** "Show me daily active wallets for the last 30 days and compare to the previous 30 days"

### Conversion funnel

**Prompt:** "What percentage of visitors connect their wallet?"

**What you'll learn:** Top-of-funnel conversion rate.

**Follow-up:** "Of those who connect, how many complete a transaction?"

### User retention

**Prompt:** "What's my week-1 retention rate?"

**What you'll learn:** How many users return after their first visit.

**Follow-up:** "Show me retention by acquisition source"

### Revenue analysis

**Prompt:** "What's our daily revenue trend for the past month?"

**What you'll learn:** Revenue/volume metrics and trends.

**Follow-up:** "Break this down by day of week"

### User segments

**Prompt:** "Which customers have the highest usage?"

**What you'll learn:** High-value user identification and behavior patterns.

**Follow-up:** "How many whales (net worth > \$100k) do I have, and what's their average transaction volume?"

### Geographic distribution

**Prompt:** "Which countries have the most active users?"

**What you'll learn:** Geographic breakdown of your user base.

**Follow-up:** "Which country has the highest average transaction value?"

### Feature adoption

**Prompt:** "How many users visited the /stake page this week?"

**What you'll learn:** Feature page traffic.

**Follow-up:** "Of those, how many completed a staking transaction?"

### Drop-off analysis

**Prompt:** "Where do users drop off in the transaction flow?"

**What you'll learn:** Conversion bottlenecks.

**Follow-up:** "Is drop-off higher on mobile or desktop?"

### Anomaly detection

**Prompt:** "Are there any unusual patterns in my data this week?"

**What you'll learn:** Spikes, drops, or outliers worth investigating.

**Follow-up:** "What might have caused the spike on Tuesday?"

### Acquisition quality

**Prompt:** "Which channels produce users with the highest retention?"

**What you'll learn:** Which acquisition sources bring the best long-term users.

**Follow-up:** "Correlate signup source with 30-day retention: which acquisition channels have the best retention rates?"

### Activation

**Prompt:** "Which early action predicts long-term retention?"

**What you'll learn:** The "aha moment" event that separates retained users from churned ones.

**Follow-up:** "What percentage of users complete that action in their first session?"

### Revenue cohorts

**Prompt:** "Are recent customers more valuable?"

**What you'll learn:** Whether newer cohorts are spending more or less than older ones.

**Follow-up:** "Break down average revenue per user by weekly cohort"

### Behavioral insights

**Prompt:** "What do power users do that casual users don't?"

**What you'll learn:** Behaviors that differentiate high-value users from the rest.

**Follow-up:** "Which features have the biggest usage gap between power and casual users?"

### Churn prediction

**Prompt:** "What behaviors signal a user is about to churn?"

**What you'll learn:** Early warning signals for at-risk users.

**Follow-up:** "Find the top 10 users by session count, show their activity over time, and flag any that look like potential churns"

### Other questions

**Revenue attribution:**

> "Which UTM campaign drove the most transaction volume?"

**Competitive insights:**

> "How many of my users also use Uniswap?"

**Time patterns:**

> "What time of day do most transactions happen?"

**Platform comparison:**

> "How does mobile vs desktop conversion compare?"

## Best Practices

<Tip>
  **Be specific with time ranges.** Instead of "How many transactions?" say "How many transactions in the last 7 days?" The AI will ask if you're unclear.
</Tip>

<Tip>
  **Use follow-ups to drill deeper.** Don't try to ask everything in one question. Start broad, then ask follow-ups to explore specific angles.
</Tip>

<Tip>
  **Ask about wallet labels.** If you've set up [wallet intelligence](/guides/wallet-segmentation), ask the AI to segment by label. "Show me DAU by wallet label" reveals which user types drive value.
</Tip>

<Tip>
  **Copy SQL for custom charts.** If a generated query is useful beyond this one answer, copy it into the Explorer or a SQL chart on your dashboard.
</Tip>

<Tip>
  **Check Insights daily.** It refreshes the first time you open it each day, so reviewing it each morning surfaces issues early.
</Tip>

## FAQ

<AccordionGroup>
  <Accordion title="Does the AI have access to all my data?">
    Yes, Ask AI can query all events, properties, and custom SQL tables in your project schema. It respects your project's permissions, so team members only see data they have access to.
  </Accordion>

  <Accordion title="How far back can I ask about historical data?">
    You can query as far back as your project has data (usually from your first SDK install, months to years ago). Ask away: "Show me DAU for the last 6 months."
  </Accordion>

  <Accordion title="Can I ask the AI to create a new event or property?">
    No, Ask AI only reads and analyzes data. To create new events or properties, configure them in your SDK or contract events setup.
  </Accordion>
</AccordionGroup>