> For the complete documentation index, see [llms.txt](https://docs.smaq.io/documentation/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.smaq.io/documentation/get-started/tutorials/chat-with-smaq-ai.md).

# Chat with SMAQ AI

SMAQ AI isn't a generic chatbot — it knows your connected data, your datablocks, your dashboards, and your playbooks. It can fetch numbers, build visualizations, run multi-step analyses, and push results back onto your dashboards.

### Prerequisites

* A project with at least one connected data source.

***

### Where to find chat

* **In any project**: click **Chat** in the side menu.
* **Inside any dashboard**: click the chat icon in the dashboard header. The AI has access to the current dashboard's context.
* **On a public dashboard**: viewers can chat about the data without signing in. See [Public dashboards & sharing](/documentation/features/dashboard/public-share-and-chat.md).

<figure><img src="/files/sCyzk3RH69iCaaeedt4J" alt=""><figcaption></figcaption></figure>

{% file src="/files/G18iLngVhsKdrY9utTpH" %}

### What the AI can do

The chat experience is powered by a team of specialized agents. You don't pick agents directly — the supervisor routes your question to the right one. See [What SMAQ AI can do](/documentation/features/chat/capabilities.md) for the full list.

A non-exhaustive sample:

* **"What was my ROAS last week vs the week before?"** — fetches the data and replies with a number + a comparison.
* **"Show me a line chart of spend by day, last 30 days, by channel."** — builds the chart inline. You can click **Add to dashboard** to push it onto a dashboard.
* **"Why did conversions drop on Tuesday?"** — runs a multi-step analysis: pulls the data, compares against typical days, checks campaign-level changes, and writes up findings.
* **"Run my weekly review playbook."** — executes a saved [playbook](/documentation/features/chat/playbooks.md).

### Step 1 — Start with a clear question

Be specific about:

* **What metric** (Spend, Conversions, ROAS, etc.)
* **Which source** (if you have multiple — "Google Ads spend" vs "Meta spend")
* **The time window** ("last 30 days", "Q1", "yesterday vs the day before")

Examples that work well:

* "What's my Google Ads CPA by campaign for the last 14 days?"
* "Top 5 fatigued ads on Meta in the last 7 days."
* "Compare GA4 ecommerce conversion rate by channel, this month vs last."

Examples that need clarification (the AI will ask):

* "How am I doing?" — too vague.
* "What's the best channel?" — best at what? Over what period?

### Step 2 — Use suggested questions

When you open the chat, you'll see a strip of suggested questions tailored to your connected data. Click any to send it as your first message — they're a good way to discover what's possible.

### Step 3 — Watch the thinking

For multi-step questions, the AI shows you what it's doing — fetching data from Google Ads, comparing periods, running an aggregation, etc. You can collapse this if it's noisy.

### Step 4 — Add results to a dashboard

When the AI returns a chart or table, hover over it and click **Add to dashboard**. Pick the target dashboard and page. The widget lands there, fully editable like any other widget.

<figure><img src="/files/SOSsbrv02CMiYQSFjIEy" alt=""><figcaption></figcaption></figure>

### Step 5 — Iterate

Treat chat as a conversation:

* "Now break that down by device."
* "Filter to just the search campaigns."
* "Compare against the same period last quarter."

The AI keeps the context of the conversation. Start a new chat when switching to an unrelated topic — see the [Chat History](/documentation/organization/review-historical-user-messages.md) view to find old conversations.

### Best practices

* **One topic per chat** — keeps context tight and the AI's answers focused.
* **Be specific** — "ROAS by campaign last week" beats "campaign performance."
* **Use playbooks** for repeated analyses — see [Build your first playbook](/documentation/get-started/tutorials/build-your-first-playbook.md).
* **Click "Add to dashboard"** to build a dashboard incrementally through conversation, instead of configuring each widget manually.

### What's next

* [What SMAQ AI can do](/documentation/features/chat/capabilities.md) — full agent reference.
* [Playbooks](/documentation/features/chat/playbooks.md) — codify your analyses.
* [Build your first playbook](/documentation/get-started/tutorials/build-your-first-playbook.md) — step-by-step tutorial.