> 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/features/chat/capabilities.md).

# What SMAQ AI can do

SMAQ's chat is powered by a roster of specialized agents coordinated by a supervisor. You never pick agents directly — the supervisor routes your question. But knowing what each agent does helps you ask better questions.

### The agents

#### Concierge / Supervisor

Routes incoming questions. Decides whether to answer directly, fetch data, generate a chart, run a playbook, or chain multiple agents. Not directly visible; it's the orchestrator.

#### Fetch Data Agent

Pulls raw data from your datablocks. Used when you ask for numbers — "What's my ROAS for the last 7 days?", "Top 5 campaigns by spend last month."

Best questions:

* Single-fact: "What was my GA4 conversion rate yesterday?"
* Comparisons: "Spend last week vs the week before."
* Filtered: "Conversions from mobile only, last 30 days."

#### Visualization Agent

Turns fetched data into charts, tables, metric cards, pivot tables. Used when you ask for a visual or when a number alone wouldn't be enough.

Best questions:

* "Show me spend by day, last 30 days, as a line chart."
* "Bar chart of conversions by channel for this month."
* "Pivot table of campaigns by week."

After the chart renders, **Add to dashboard** to keep it.

#### Search Datablock Agent

Finds the right datablock to answer a question. The richer your project datablocks (with good names and descriptions), the better this agent performs.

When you ask a question and the AI says "I'll use the 'Google Ads weekly campaigns' datablock to answer that" — that's Search Datablock at work.

#### Business Agent

High-level analysis and recommendations. Multi-step reasoning.

Best questions:

* "Why did ROAS drop last week?"
* "What channel should I spend more on?"
* "Should I pause this campaign?"

The Business Agent often chains Fetch Data, Visualization, and Search Datablock to build its answer.

#### Dashboard Agent

Dashboard-specific queries and modifications.

Best questions:

* "What does this chart show?" (inside a dashboard)
* "Add a chart showing weekly conversions to page 2."
* "Filter this whole page to just the search campaigns."

#### Context Retrieval Agent

Pulls relevant info from your [Knowledge Base](/documentation/features/knowledge-base.md) when answering.

Not directly invokable — it runs whenever another agent needs your business context.

#### Auto Analysis Agent

Surfaces noteworthy findings without being asked. Used in automations and on dashboard load.

#### Insight Discovery Agent

Proactive opportunity surfacing. Behind features like "Did you know your Tuesday ROAS is 2× your Monday ROAS?"

#### Report Agent

Generates polished reports from dashboards. Drives [Generate Report AI](/documentation/features/reports/generate-from-dashboard.md).

#### Competitive Analysis Researcher

Pulls in competitor signals — the Meta Ad Library, Google Transparency Center, public web sources. Used by the [creative intelligence](/documentation/features/creative-intelligence.md) automations and on demand from chat.

Best questions:

* "What is competitor X running on Meta?"
* "Compare our messaging against competitor Y."

#### Performance Marketing Education Agent

Explains marketing concepts. Useful when you want context, not just data.

Best questions:

* "What is impression share lost to budget?"
* "How is iOS 14+ attribution different from before?"
* "When should I use Smart Bidding vs Manual?"

#### Postprocess Data Agent

Reformats and transforms data — currency conversion, deduplication, aggregation rollups. Runs invisibly during multi-step queries.

#### Execute Playbook Agent

Runs [playbooks](/documentation/features/chat/playbooks.md) on demand. When you ask a question that matches a saved playbook, this agent executes it.

#### Format Response Agent

Polishes the final response — formatting, headlines, tone. Last step before display.

***

### Capabilities at a glance

| You want to...                      | The AI uses...                                     |
| ----------------------------------- | -------------------------------------------------- |
| Get a single number                 | Fetch Data + Format Response                       |
| See a chart                         | Fetch Data + Visualization                         |
| Get a recommendation                | Business Agent (often + Fetch Data, Visualization) |
| Understand something on a dashboard | Dashboard Agent + Search Datablock                 |
| Run a saved analysis                | Execute Playbook                                   |
| Compare against competitors         | Competitive Analysis Researcher                    |
| Learn a marketing concept           | Performance Marketing Education                    |

### What the AI can't do

* **Modify your ad accounts.** SMAQ never writes back. The AI can recommend a negative keyword; you still copy-paste it into Google Ads.
* **Predict the future with high confidence.** Statistical projections are available; promises aren't.
* **Read between the lines of your strategy.** Tell the AI directly via the [Knowledge Base](/documentation/features/knowledge-base.md).
* **Access data outside your connections and KB.** No external web search beyond competitor-research connectors.

### Tips for better chat

* **Front-load specifics.** "Show me Google Ads ROAS by campaign for Q2 as a bar chart" → one shot. "Show me ROAS" → three back-and-forth clarifications.
* **Use playbooks for repeated investigations.** See [Playbooks](/documentation/features/chat/playbooks.md).
* **Rich KB = better answers.** See [Knowledge Base](/documentation/features/knowledge-base.md).
* **Iterate.** "Now break that down by device" is faster than starting over.

### Related

* [Playbooks](/documentation/features/chat/playbooks.md).
* [Knowledge Base](/documentation/features/knowledge-base.md).
* [Chat with SMAQ AI (tutorial)](/documentation/get-started/tutorials/chat-with-smaq-ai.md).


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.smaq.io/documentation/features/chat/capabilities.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.