Guys, I was working on the raw data for the first time... and after using openAI API on 5K records, it cost me $22.8 for just data enrichment with AI (it does not include contact details enrichment). And I used the cheapest model, i.e., 4.1 mini. Is that normal?
That cost seems high! You can cut your AI costs by up to 90% by connecting your own OpenAI API key instead of using our managed one. When you use your own API key, you won't consume any Clay credits for AI actions.
AI credit costs in Clay are based on prompt and token usage, and 90% of our users connect their own API keys to save significantly.
Is that what you were looking for?
Yes, I'm using my own API key and that's what costing me $22.8 for 5K records data enrichment
That cost is actually normal when using your own OpenAI API key. The price difference between Clay's managed key and your own is significant - it's around 0.003 cents vs 1 cent per call.
When you use your own API key, you're paying OpenAI's standard rates directly, which can add up quickly with large datasets. The cost depends on your prompt complexity and token usage.
While using your own key saves Clay credits, it may not always be more economical for high-volume operations since Clay buys API access in bulk.
Did that help clarify the cost for you?
How can I reduce the cost? I'm already using the cheapest model.
Here are several ways to reduce your AI costs:
Use AI Formulas for data manipulation like summarizing job titles or combining columns - they cost zero credits.
Use the Metaprompter to craft effective prompts on the first try instead of burning credits on multiple iterations.
Implement conditional runs - only process rows that meet specific criteria so you're not wasting money on enrichments that won't yield useful results.
Switch to even cheaper models like "4.1 Nano" when available - check your Configuration tab to see if you can go lower than 4.1 mini.
Be aware of model selection - credit costs fluctuate based on which AI model is selected, so double-check you're using the most cost-effective option before execution.
Did that help you find ways to cut costs?
If you still need help reducing your AI costs or have more questions about optimizing your usage, I’m here to assist. Would you like to share more details about what you’re trying to achieve or any specific challenges you’re facing?
