Fixing Inaccurate AI-Generated Copy in Personalization Campaigns

I am creating a personalization campaign on company growth with Clay. I have enriched all the company data with the relevant growth rates. I have outputted the growth rates in a column and have used AI to create copy based on this enrichment data. For some reason sometimes the growth rate data is not accurate in the created copy. For example, for a specific company the growth rate was 12%. The AI created copy based on a growth rate of 42% which is not correct. I have seen this happen a few times and it leads to awkward conversations if the data is not correct. Does anyone have an idea why this happens and how I can fix it? Is it a promt issue? See pictures for extra context.

9 comments

Channeled
APP
·
·
To fix the issue with incorrect growth rate data in AI-generated copy, you'll need to focus on improving your prompt engineering. Here's how to address this: 1. Set Clear Context • Start with a system prompt that explicitly defines the context and goal • Specify that the AI should only use the exact growth rate values from your data 2. Provide Specific Instructions • Include step-by-step instructions for the AI to follow • Explicitly tell the AI to reference and use the exact numbers from your growth rate column 3. Verify Output Format • Be explicit about how you want the numbers formatted and used in the copy • Use the preview feature to check if the AI is accurately incorporating the growth rate data before applying the formula to all rows If you continue to see inconsistencies, try iterating on your prompts and adjusting the wording until you get the desired results.
Sources: AI Formulas + Conditional Runs [Automated Outbound] - Clay University Prompt Engineering Crash Course [Clay 101] - Clay University
Channeled
APP
·
·
Did that answer your question?
Jerry v.
·
·
But why did it happen on only a handul of rows? Since I checked a couple and 95% of the outputs was correct. Is this still related to my prompt?
Channeled
APP
·
·
The inconsistency in AI-generated copy for a small number of rows can still be related to your prompt, but it's a normal part of the AI process that can be improved. When the AI doesn't give exactly what you're looking for on the first try, this is normal and part of the process. To fix these occasional inaccuracies: 1. Strengthen your system prompt by: • Explicitly defining the context and goal • Adding specific instructions to use exact numbers from your data • Including clear formatting requirements for numbers 2. Use the preview feature to check the AI's output accuracy before applying the formula to all rows. If you continue seeing inconsistencies in specific rows, try iterating on your prompt by adjusting the wording until you achieve consistent results across all rows.
Sources: Prompt Engineering Crash Course [Clay 101] - Clay University AI Formulas + Conditional Runs [CRM Enrichment] - Clay University
Channeled
APP
·
·
Did that answer your question?
Jerry v.
·
·
Go it. How do I use the preview function? Or do you mean test the promt on the first 10 rows?
Channeled
APP
·
·
When setting up a new AI enrichment, you can test your prompt on a small sample before running it on your entire dataset. Here's how: 1. During the action setup, click Save > Run First 10 Columns to test your prompt on a small sample. This approach is recommended because: ** It helps identify and fix any errors in advance ** Allows you to refine your prompts before running the full enrichment * Saves credits by avoiding potential issues with the entire dataset Once you're satisfied with the results from your test run, you can then apply the enrichment to the entire column.
Sources: Guide: Ways to save Clay credits | Documentation | Clay University Run Options
Channeled
APP
·
·
Was that helpful?
Channeled
APP
·
·
Hi Jerry v.! This thread was recently closed by our Support team. If you have a moment, please share your feedback: