> AI agents: this is one page from Mammoth Analytics documentation. The index of all pages as Markdown is https://docs.mammoth.io/llms.txt. Append `.md` to any docs URL, or send `Accept: text/markdown`, to get Markdown.

# AI dashboard creation

> Create a Mammoth dashboard by describing it in plain language, refine it with the Dashboard Assistant, keep its data fresh, and publish it to viewers.

This guide is for business analysts and managers who want a dashboard without writing SQL. It walks through Mammoth's AI dashboard creation: you describe the dashboard you want, and the Dashboard Assistant generates it from your data. For the full reference, see [Creating dashboards](https://docs.mammoth.io/learn/dashboard/creating-dashboards/).

**What you will build:** A shareable dashboard generated through natural language prompts. No code and no manual chart configuration.

---

## The challenge of traditional dashboard requests

Traditional dashboard creation workflows are slow:

1. You describe what you want to see to a data analyst
2. They write SQL, pull the data, and design charts
3. You review the layout and request changes
4. They iterate, often taking days or weeks
5. The dashboard becomes stale, and the analyst is bottlenecked on updates

AI dashboard creation shortens steps 2 to 4: you describe the dashboard, and Mammoth generates a first version you can refine by conversation.

---

## Requirements

- A dataset in Mammoth with at least one View (a View from a transformation pipeline works best)
- Natural language descriptions of what you want to see
- A Mammoth workspace with at least one project

---

## Step 1: Prepare your dataset

Before creating the dashboard, make sure your data is clean and properly labeled. The AI reads your View's column names, data types, and sample values.

1. Open or create a dataset in Mammoth
2. Check the column **Explore Cards** to see data types and value distributions
3. Rename columns to be descriptive (e.g., `created_at` → `OrderDate`, `usd_amount` → `Revenue`) using **Rename Column** in the column menu
4. Make sure the View you want to use has the transformations applied

**Good dataset characteristics:**
- Clear, human-readable column names
- Correct data types (dates as dates, numbers as numbers, categories as text)
- No excessive null values in critical columns

---

## Step 2: Start a new dashboard

1. Click **Dashboards** in the left sidebar
2. Click **Create Dashboard**
3. In the dialog, browse to your project, select the dataset, and choose the View to use as the data source
4. Click **Select**

A full-page prompt screen opens.

---

## Step 3: Describe your dashboard with natural language

The prompt is where you tell Mammoth what you want to see. The more specific you are, the better the result.

**Example prompts:**

*For sales analytics:*
> "Show pipeline by stage using a funnel chart. Add a table showing top 10 sales reps by closed-won amount this quarter. Include a line chart showing deal velocity by week."

*For financial reporting:*
> "Create a P&L dashboard with a table showing revenue, COGS, and operating expenses by month. Highlight months where expenses exceeded budget in red. Add a stacked bar chart showing revenue by product line."

*For operational KPIs:*
> "Show 4 KPI cards at the top: total units produced today, defect rate (%), on-time delivery %, and open work orders. Below, add a line chart showing production throughput by hour and a table of the top 10 open issues."

**Prompt best practices:**

**Recommended approach:**
- Name the chart types you want (bar, line, pie, scatter, table, KPI card)
- Specify metrics clearly (e.g., "sum of revenue", "count of orders", "average days to close")
- Describe filters or groupings ("by stage", "by month", "by region")
- Use your actual column names from the dataset

**Approaches to avoid:**
- Vague descriptions ("show the data" is too generic)
- Omitting column names ("show sales" ← clarify: Amount, Revenue, or Price?)
- Requesting charts the data can't support (Mammoth generates from your data and cannot create data that does not exist)
- Describing complex multi-step calculations (do those in the View's pipeline first, and keep the prompt to aggregations and basic math)

---

## Step 4: Generate and review

1. Type your description in the prompt box and click the **send button**
2. Mammoth generates the dashboard and opens the editor, with the Dashboard Assistant chat on the left, the Filters panel in the middle, and a live preview on the right
3. Review the result:
   - Are the charts the types you requested?
   - Do the metrics look correct?
   - Is the layout readable?

If something is off, tell the Dashboard Assistant in the chat. Common refinements:
- "Move the table below the chart"
- "Use a bar chart instead of a line chart"
- "Add a date range filter"
- "Show the top 5 instead of top 10"

---

## Step 5: Refine and add interactivity

Keep chatting with the Dashboard Assistant to adjust any aspect of the dashboard: charts, layout, pages, filters, and theme.

**Filters:** the AI creates filters for relevant categorical, date, and numeric columns, and they appear in the Filters panel. Ask the assistant to add, remove, or change filters (for example, "Add a filter for region"). See [Dashboard filters](https://docs.mammoth.io/learn/dashboard/dashboard-filters/).

**Themes:** ask for a theme by name or pick one from the **Theme** menu. See [Dashboard themes](https://docs.mammoth.io/learn/dashboard/dashboard-themes/).

**Pages:** for larger dashboards, ask the assistant to split the content into pages with tab navigation. See [Dashboard navigation](https://docs.mammoth.io/learn/dashboard/dashboard-navigations/).

---

## Step 6: Keep the data fresh and share

**Keep the data current:**

1. Schedule a **Dataset Refresh** automation (Automations → New Automation → Dataset Refresh) if the source is a connector
2. In the dashboard's sync settings, turn on **Sync automatically** so new rows come into your draft when the pipeline runs, and **Publish new data automatically** so they reach viewers

**Share the dashboard:**

1. Click **Publish & Share** (top right) and click **Publish**
2. Set **Access Type**:
   - **Authenticated**: viewers must sign in to Mammoth
   - **Public**: anyone with the link can view (no sign-in required)
3. Copy the share link and send it by Slack, email, or any other channel

---

## Results to expect

- You get a first version of the dashboard from a description, then refine it by conversation instead of reconfiguring charts by hand
- Business users can create and adjust dashboards without writing SQL
- AI-generated dashboards are a starting point: check the metrics against your data before sharing

**Example scenario:**
> A marketing director needs a dashboard showing campaign performance by channel. They describe what they want ("Show impressions, clicks, CTR, and conversion rate by channel as a table. Add a line chart of daily impressions over the last 30 days."), review the generated dashboard, and publish it. If they want to tweak the layout the next day, they ask the Dashboard Assistant to change it.

---

## Common patterns and prompts

**E-commerce dashboard:**
> "Show total revenue, order count, and average order value as KPI cards. Add a line chart of daily revenue for the last 90 days. Show top 10 products by revenue as a table. Add filters for date range and product category."

**HR and people analytics:**
> "Create 4 KPI cards: total headcount, open requisitions, average tenure, and voluntary turnover %. Add a bar chart showing headcount by department. Add a line chart showing new hires per month for the last 12 months."

**Customer support metrics:**
> "Show average response time, resolution time, customer satisfaction score, and ticket volume as KPI cards. Add a line chart of tickets created versus resolved by day. Show top 10 ticket types as a bar chart."

These prompts assume your View already has the columns they refer to (for example, a status or resolution time column). Prompts can only chart what the data contains.

---

## Limitations

**AI dashboard creation works best for:**
- Aggregated metrics (sums, counts, averages)
- Standard chart types (bar, line, pie, scatter, table, KPI card)
- Straightforward filters (date range, category select)

**Do these in the View's pipeline first, then build the dashboard on top of it:**
- Complex calculations (multi-step math across joined datasets)
- Derived columns that need conditional logic

---

## Variations: dashboard suites and email delivery

**Build a dashboard suite:**
Create multiple complementary dashboards (for example, Sales, Finance, Operations), each on its own View.

**Email the data:**
Use a **Messaging** automation to email the underlying View as a CSV attachment on a schedule.

---

## Next steps after your first dashboard

1. **Start with your simplest use case**: pick one dashboard your team needs and describe it in natural language
2. **Share the dashboard**: show it to stakeholders and iterate based on feedback
3. **Add more dashboards**: once comfortable, create dashboards for other teams

---
Source: https://docs.mammoth.io/guides/use-cases/ai-dashboard-creation.md · Updated: 2026-10-13