How to Build a Data Analysis Assistant

A Data Analysis Assistant is the best option when you want to create a reusable chart, table, or summary from a specific set of data.

This article walks through how to create a new chart or insight step by step.

When to use a Data Analysis Assistant

- you already know the question you want to answer
- you want a specific chart, table, or summary
- you plan to rerun the analysis later
- you may want to use the result in a Slideform project

Example use case

In this example, we will create a chart showing monthly pipeline created for the last 12 months from a selected data source.

1. Create a new Assistant

To get started, go to AI Assistants > Assistants and click New.

The next screen presents the available Assistant types. Select Analysis.

2. Select the output type 

Choose an output type: Chart, Table, or Dynamic Text.

3. Select the data source 

Use the dropdown to select one or more of your Slideform connected data sources. 

Depending on the type of data, you may need to select the worksheet, table, and data columns. If your data source is large, use filters to limit the data volume that will be ingested by the AI. You can also manually limit the number of results if needed. 

Alternatively, you may enter an SQL query to select the data. Click the Manual Query checkbox and enter the query.

Use the preview button to examine the connected data.

4. Enter a prompt

In the prompt field, describe exactly what you want.

Example: "Create a chart of monthly pipeline created for the last 12 months."

Strong prompts usually include:

- the metric
- the date range
- the breakdown or grouping
- the output you want
- any formatting preferences

5. Give the insight a name

6. Select a branding profile and an AI model

If you want the chart or table to follow a specific visual style, select a branding profile before generating the result.

This is especially useful for:

- client-facing charts
- executive reporting
- outputs that will be embedded into a presentation

Slideform supports mutliple AI models.

7. Generate the chart or insight

Click Test to run the Assistant and review the result in the preview area.

Depending on the prompt, the output may include:

- the final chart
- a summary
- a table behind the chart
- artifacts such as SQL or Python used to create the result

8. Refine if needed

If the result is close but not right, update the prompt and rerun.

Helpful refinements include:

- specifying a shorter or longer date range
- changing the grouping
- asking for sorting
- requesting a different output type
- adding instructions like `make this client-ready` or `include a short summary`

9. Save the insight 

If you are happy with the results, click Save. Saving the assistant allows you to run this analysis again as an automated input into Slideform reports or presentations.

The Code window allows you to view any code created by the Assistant. This is provided for validation purposes, but you can copy the code to run elsewhere if you want.

If you want to modify the analysis, enter the desired changes in the text box and hit enter. This will automatically modify your prompt. Then click Test to re-run the process.

The Start Over button deletes the previous plan, script, and results and starts the request again in a fresh context window.

Data Analysis Assistant best practices

- Build one clear analysis per Assistant.
- Keep the data scope narrow when possible.
- Name the Assistant based on the output it creates.
- Save successful results as insights.
- Use chart branding profiles for presentation-ready outputs.