Clay Icon

Issue with Claygent’s Output Not Consistently Following Instructions

·
·

Not sure if this is a bug, but it bugs me most days. When I ask Claygent to perform a task like find the name of a decision maker, and explicitly instruct it to put something like 'decision maker not found' as the output if it can't find anyone it still goes off piste. Just done this on a list of 1000 and it's used these variations: N/A Not found decision maker not found Unknown Not identified yet Which means I then have to filter by 'does not contain any of'. Every time I do this. Just thought I'd flag in case it is a bug!

  • Avatar of Channeled
    Channeled
    APP
    ·
    ·

    To ensure consistent output formatting from Claygent, you should:1. Set clear instructions in your system prompt that specify exactly how you want the output formatted when no decision maker is found. 2. Be explicit about the output format you want, including the exact phrase to use when no results are found. For example, specify "Use exactly 'decision maker not found' (no variations) when no decision maker can be identified." 3. Leverage Neon model features, which excel at structured data extraction and consistent output formatting. The Neon model allows you to format results into specific columns, making it easier to work with the results without additional manipulation. Sources: Prompt Engineering Crash Course [Clay 101] - Clay University Prompt Engineering Crash Course [Limitless Research] - Clay University

  • Avatar of kushagra
    kushagra
    ·
    ·

    have you tried using the json output? that should solve for this

  • Avatar of Channeled
    Channeled
    APP
    ·
    ·

    To ensure consistent output formatting in Claygent, you can:1. Enable JSON output formatting when setting up your task. This will help structure the data in a consistent format. 2. In your system prompt, be explicit about: • The exact output format you want • The specific phrase to use (e.g., "decision maker not found") • Step-by-step instructions for the AI to follow 3. Define specific column outputs and data types (Text, True/False, Number, URL, Select, or Email) to ensure consistent formatting across all results.This structured approach will help maintain consistency and eliminate variations in the output format. Sources: Prompt Engineering Crash Course [Limitless Research] - Clay University Use AI integration overview | Documentation | Clay University

  • Avatar of Tanvi R.
    Tanvi R.
    ·
    ·

    Hi Naeem, thanks for reaching out! Kushagra is right, JSON output should solve for this! Specifying output structure in your prompt alone isn't enough. Select JSON Schema as the output structure of your prompt and then select Generate from prompt. This will use AI to automatically prefill your output fields with the correct JSON formatting. When your prompt has clear parameters on what output fields it should contain, it will be able to execute more effectively and consistently! Here's a quick tutorial explaining JSON Schema for AI Output Enhancement: https://www.loom.com/share/4f7b6e6ecba44811af419a1ae0932d1b Feel free to give this a try and let me know if this resolves your issue!

  • Avatar of Channeled
    Channeled
    APP
    ·
    ·

    Hey there - just wanted to check in here to see if you needed anything else! Feel free to reply back here if you do.

  • Avatar of Channeled
    Channeled
    APP
    ·
    ·

    We haven't heard back from you in a bit, so we're going to go ahead and close things out here - feel free to let us know if you still need something!

  • Avatar of Channeled
    Channeled
    APP
    ·
    ·

    Hi Naeem A.! This thread was recently closed by our Support team. If you have a moment, please share your feedback: