Getting Started

Welcome to PromptLoop. This guide will offer a quick start to how the platform works and how to get started.

Quick Navigation:

Steps to Get Started:

• Step 1: Define Your Goal and Create a Task
• Step 2: Test and Edit Your Task
• Step 3: Run a Dataset on Your Task
• Step 4: Review, Filter, Search and Export

Introduction#

PromptLoop allows teams to leverage AI to research, build, and enrich datasets at lightning speed. This is accomplished first and foremost in a way that is non-technical, simple, and does not require an advanced knowledge of AI systems or prompting, despite our name :).

Principles

• Flexible - Access to hundreds of models and functions under a simple task based editor. Focus on the inputs (information you have) and the outputs (information you are looking for).
• Accurate - PromptLoop tasks are designed to stay within the bounds of what has been tested to work well. Whether extracting information from a website or categorizing industry terms, you will receive formatted and sourced answers from reliable models only using the relevant input data.
• Transparent - PromptLoop tasks and capabilities are transparent to the teams that rely on them. See exactly what your tasks are accomplishing and the steps they take to get there. This allows for peace of mind and helps you improve them with edits.

PromptLoop is built to help you accomplish repeatable AI research and analysis tasks, taking in sets of inputs (usually in the form of rows in a dataset) and returning formatted responses based on the task that you select.

PromptLoop is built around data, and usage is measured in terms of how much data the system is finding and generating. We have detailed information on this /credits. For teams that need support growing their business with reliable tools and data, we offer team packages that are customized with both support and model capabilities. These include detailed onboarding to set you up with the tools that you need.

To learn more and answer questions about your specific business you can book a demo here.

Core Concepts#

• Tasks - Actions for our system to accomplish for you. Tasks take inputs (like a website URL) and return outputs as new columns or rows in your dataset.

• Datasets - The repository for information you'll run tasks on. You can search, filter, save and share both inputs and outputs up to hundreds of thousands of rows.

• Integrations - Spreadsheets - All tasks can be used directly in Excel and Google Sheets for cell-specific operations.

Let's walk through the simple steps to get started:

Steps to Get Started#

1. Define your goal and create a task#

When you first log in, you'll want to create a task. A task is a function you can run repeatedly with different inputs to get consistent outputs.

For example, a task might:

• Extract company size and industry from corporate websites
• Determine if a company offers specific services
• Find contact information for key decision-makers

Creating your first task:

1. Navigate to the Tasks section in the top navigation
2. Choose from the task library of pre-built templates
3. Or select "Create New Task" to build a custom one

2. Test and edit your task#

Once you've created a task, you'll want to make sure it works as expected before running it on a full dataset.

Testing your task:

1. Every task has a built-in test page
2. Enter a sample input (like a website URL)
3. Click "Run Test" to see what output your task produces

Tips for testing:

• Try 3-5 different examples that represent the range of inputs you'll use
• Check that the outputs follow the format you need
• Look for any unexpected results or errors

If the task doesn't produce exactly what you need, use the editor to refine it:

• Be specific about formatting requirements
• Clarify what information to extract
• Specify how to handle exceptions or missing data

3. Run a dataset on your task#

Once your task is working correctly, it's time to run it on a full dataset.

Preparing your dataset:

1. Create a CSV or Excel file with the inputs your task needs
2. Ensure each column has a clear header name
3. Upload it on the Datasets page

Launching a job:

1. Navigate to your dataset and click the blue "Launch Job" button
2. Select the task you want to run
3. Map your dataset columns to the task inputs
4. Click "Launch" to begin processing

Your job will run in the background, processing each row through your task. The system will show you progress in real-time:

4. Review, filter, search and export#

When your job completes, your results will be saved as a new version of your dataset.

Reviewing results:

1. Open your dataset to see the original inputs plus new output columns
2. Use the search and filter options to examine specific results
3. Verify that the data meets your needs

Working with your results:

• Sort columns to identify patterns
• Filter to focus on specific criteria
• Export the entire dataset or selected portions
• Save versions with meaningful names for future reference

This entire process - from creating a task to exporting results - can take just minutes but save dozens of hours of manual work.

Next Steps#

Now that you understand the basics, explore these resources to get even more value:

• Custom Tasks Guide - Learn to create advanced tasks for specific needs
• Dataset Management - Tips for organizing and managing large datasets
• Team Collaboration - How to share and collaborate on tasks and datasets

Remember, each task you create is reusable. Build your library of tasks to automate more and more of your research and data work.