Prompting best practices
Learn how to write clear, effective prompts to get the most out of Parabola’s AI chat and Custom transform step. In this session, the Parabola team shares best practices, real-world examples, and practical tips for turning plain language into powerful, production-ready workflows. Whether you're building reports, transforming data, or automating exports, this guide will help you use AI to move faster and build with confidence.
Topics covered
- How Parabola interprets your prompts
- Writing high-quality prompts for Custom Transform
- Formatting, logic, and edge case considerations
- Real prompt examples and breakdowns
- Tips from Parabola’s product and engineering teams
For more, explore Parabola's prompting best practices guide.
Overview & Purpose
The Parabola team gathered to share best practices for writing effective prompts when using the AI Chat Sidebar and the Custom Transform step. These tools help users—from operations and supply chain to finance—build and troubleshoot flows using natural language.
What is the AI Chat in Parabola?
- Described as an always-available assistant that lives alongside your flows.
- Helps with:
- Explaining how Parabola works
- Building flows based on described use cases
- Troubleshooting errors in real time
Do You Need to Be an AI Expert?
- No — Users just need to describe their processes naturally, like chatting with a colleague.
- The system is designed to be beginner-friendly and intuitive.
Top Prompting Tips
From Zach:
- Start by adding your data to the canvas. The AI uses visible data to make better recommendations.
- Tell the AI your data sources (e.g., Google Sheets, files, Shopify) to help it get context early.
From Adam:
- Include specific details in prompts:
- Data source (e.g., Shopify)
- Use case (e.g., sales report)
- Time range (e.g., last 30 days)
- Key fields (e.g., SKUs, revenue)
From Isaac:
- Be specific: Define inputs, transformations, and desired outputs.
- Think through your goals before starting.
- Don’t try to do everything at once—break big flows into bite-sized prompts.
From Brian:
- Precision matters more than verbosity.
- Avoid vague terms or excessive jargon.
- Call out edge cases and exceptions explicitly.
- Describe your process like you would in Excel or to a teammate.
Using the Custom Transform Step
- This step translates natural language into custom code.
- It's especially helpful for complex logic that would otherwise require many transformation steps.
- Best practices mirror the AI Chat Sidebar:
- Describe both the desired output and how to get there
- Be clear about formats (e.g., date, currency)
- Reference column names directly
New Feature: Improve Prompt Button
- Soon, users will be able to click a button to auto-improve their prompt using AI.
- It leverages all the tips discussed and sends the prompt to Co-Builder for enhancement.
Final Tips
- Isaac: “When in doubt, try it out.” You won’t break anything—iterate as you go.
- Zach: If you’re stuck, try clearing the conversation history in the chat to get a fresh perspective.
Closing Note
This was a collaborative discussion to help users get more value from Parabola's AI features. Whether you’re an AI novice or a seasoned builder, using natural, specific, and structured prompts will dramatically improve your experience.
For further questions, reach out to the Parabola team—or just ask Co-Builder.