Creating Tasks

Step-by-step guide to creating tasks in Better Analyst. Save a working chat, review its prompts and data sources, choose agents, and configure scheduling.

Learn how to turn a successful analysis into a reusable task that you can run again or schedule.

Before You Start

Have these ready:

  • A clear analysis goal
  • Data source (file or connector)
  • The steps you want to automate

Test your analysis in a regular chat first. Once it works, select Save as task.

Creating a Task

Step 1: Save a Chat as a Task

  1. Start or open a chat and complete the analysis you want to reuse
  2. Select Save as task in the chat header
  3. Better Analyst creates a task from the chat's prompts and attached data
  4. Open the Agents section's Tasks area whenever you want to review it again

Step 2: Review Data Sources

Choose the files or connections the task will analyze. Data attached to the original chat is carried into the saved task, and you can adjust the selection from the task page.

For Files:

  • Select from your uploaded files
  • Multiple files can be added

For Connectors:

  • Choose from your available connections
  • Configure any filters needed

Step 3: Choose Agents (Optional)

A task can run with one or more of your agents. The selected agent versions are frozen for scheduled and test runs, and the first agent in the ordered list is the Lead Agent.

Step 4: Review Analysis Steps

The user prompts from your saved chat become the task's ordered steps. Edit, add, remove, or reorder prompts so the sequence can run reliably each time.

Example Steps:

Step 1: "Calculate total sales by region for the current month"

Step 2: "Create a bar chart comparing this month to last month"

Step 3: "Identify the top 5 products by revenue growth"

Step 4: "Summarize key findings in 3 bullet points"

Step 5: Configure Enrichments (Optional)

Add data enrichment steps to enhance your analysis:

Sentiment Analysis Analyze text data for emotional tone:

"Analyze the sentiment of customer reviews and categorize as positive, negative, or neutral"

Language Translation Translate content between languages:

"Translate the product descriptions from Spanish to English"

Text Extraction Extract structured data from unstructured text:

"Extract company names, dates, and dollar amounts from these contracts"

Enrichments use AI Actions. Check your plan for your AI Action allocation.

Step 6: Set Schedule (Optional)

Choose when to run automatically:

  • Manual - Run only when you trigger it
  • Daily - Every day at a set time
  • Weekly - Specific day and time
  • Monthly - Specific date and time

Step 7: Run and Test

  1. Select Run for an on-demand run
  2. Review the resulting chat and the task's recent runs
  3. If you schedule the task, use Test in the scheduling dialog to send a test result
  4. Adjust the prompts, data, or agents if needed

Writing Effective Prompts

Be Specific

Good: "Calculate monthly revenue growth rate for Q4 2024" Bad: "Show me the trends"

Use Clear References

Good: "Using the sales_data column from the orders table..." Bad: "Using that data..."

Request Specific Outputs

Good: "Create a line chart with months on X-axis and revenue on Y-axis" Bad: "Make a chart"

Chain Results

Reference previous steps:

Step 1: "Calculate total sales by product category"
Step 2: "Using the results above, identify categories with >10% growth"

Multi-Source Tasks

Combine data from multiple sources:

Step 1: "From the Google Analytics connector, get sessions by channel for last month"

Step 2: "From the sales.xlsx file, get revenue by marketing channel"

Step 3: "Join these datasets by channel and calculate ROI per channel"

Using Enrichments

Sentiment Analysis

Perfect for:

  • Customer feedback analysis
  • Social media monitoring
  • Review aggregation
  • Support ticket categorization
"Analyze the sentiment of each customer review in the feedback column.
Add a new column with: positive, negative, or neutral.
Then summarize the overall sentiment distribution."

Language Translation

Perfect for:

  • International reports
  • Multilingual content
  • Localization workflows
"Translate all product descriptions from the source_text column
from German to English. Preserve the original meaning and formatting."

Text Extraction

Perfect for:

  • Document processing
  • Data entry automation
  • Unstructured data parsing
"From each invoice PDF, extract: invoice number, date,
vendor name, line items, and total amount.
Output as a structured table."

Troubleshooting

Task Fails to Run

  • Check data source is accessible
  • Verify connector credentials aren't expired
  • Review error message in run history

Results Not as Expected

  • Test steps individually in a chat
  • Make prompts more specific
  • Check data source has expected format

Enrichment Errors

  • Verify you have AI Actions remaining
  • Check the feature is enabled in settings
  • Ensure data format is compatible

Next Steps