Learn and master Clay with Clay university

Before diving into the Find step of our FETC framework, let's address a topic every Clay creator asks us:

"How can I stretch my Clay credits and save money?"

We'd like to offer a reframe here. The question isn't how to save credits in Clay, but rather how to maximize the impact of your work.

Consider this example: you could save credits by choosing the cheaper 1-credit Claygent Helium AI models over the 3-credit Claygent Argon model. But if your data quality drops even 10% across 100,000 accounts, you've saved $700 at the cost of 10,000 potential leads.

That seems counterproductive.

There are many ways to test and iterate with AI in Clay without spending credits—which we'll cover below—but remember that your ultimate goal is not to save credits, it's to generate revenue.

Let's walk through four simple strategies to help you optimize your credit investments for the best results.

🎨 Strategy 1: Use the Metaprompter to Get Better Prompts Faster

One of the most common ways people waste credits is through prompt iteration—running the same enrichment several times just to refine what they really meant.

That's where the Metaprompter comes in.

It lets you write in plain English what you want the prompt to do, and then it translates that into a structured, optimized prompt for you.

Before the Metaprompter, you'd run a test, tweak the prompt, re-run it—burning credits every time. Now, you can get it right on the first try.

For example, let's say you want to create a subject line that starts with "Congrats on" and pulls from someone's LinkedIn profile. You can just tell the Metaprompter:

"Write a subject line that starts with 'Congrats on' and references the person's most recent award, job change, or any other reason to celebrate. Keep it short, and make sure it sounds like something a human would actually say."

Just like that, you get a clean prompt without running through multiple versions.

🛠️ Strategy 2: Pick the Right Clay Tool for the Task

In Clay, not all enrichments are created equally. The one you choose can make a big difference in how many credits you use.

Let's break down the most important three:

First, use AI Formulas when you can

• These are free. No credits required.
• Use them when you want to manipulate or clean up data you already have—like summarizing job titles, counting education experiences, or combining data from multiple columns.
• AI Formulas are deterministic. That means they follow rules, not guesses.
• They're perfect for simple logic, formatting, or text extraction.

Second, use "Use AI" when you need reasoning and don't need the Internet

• The "Use AI" enrichment is great when you need more intelligence or creativity—like summarizing company descriptions or generating a sentence about someone's background.
• The key point: "Use AI" doesn't touch the internet. That's why it's less credit-intensive.
• Want to save even more? You can connect your own API key from OpenAI or Anthropic. This cuts your cost by up to 90%, and 90% of Clay users input their own AI API keys into Clay.

Third, use Claygent when you need the internet

• Claygent is Clay's AI-powered web researcher. It goes out to the internet, pulls info from websites, and returns real-time results.
• Use it when you're doing last-mile discovery—like pulling content from a company's careers page, or finding quotes from an exec in a press release.
• It's powerful. Just use it wisely.

🔄 Strategy 3: Conditional Runs

Conditional runs are the bedrock of data optimization in Clay. They allow you to add if/then logic to columns, so that specific rows won't run when certain conditions aren't met.

For example:

• Only run Claygent if no other data was found
• Only generate an email intro if there's a valid job title to reference

This is the same logic that powers waterfalls—and we cover it more in our Waterfall lessons.

🧪 Strategy 4: Always Test Before You Scale

This one's simple but critical.

Never run enrichments on your full table without testing first.

Start with 5–10 rows. Make sure the output looks good. Then scale to 50. Then 500.

Two quick tips here:

1. Turn off your table runs when you're building. This prevents accidental full-table processing while you're testing.
2. If you do trigger a full run by mistake, don't worry. You can always pause the table, stop the runs, and adjust.

Think of this as your insurance policy. It keeps you from blowing through credits because of one small mistake.

📌 Your Playbook for Optimizing Credits in Clay

To wrap up, here's your playbook for saving credits in Clay:

1. Use the Metaprompter to write smarter prompts before you run them
2. Choose the right enrichment for the job—start with AI Formulas, use "Use AI" with your own key when you need logic, and save Claygent for when you truly need web research
3. Be smart with your conditional runs—only run rows that get the data you need and ignore everything else
4. Test before you scale—start small, review your results, then run the rest

By making a few smarter decisions up front, you can save hundreds—or even thousands—of credits over time.

And that means more budget and more bandwidth to explore what really matters: building powerful, AI-powered GTM workflows.

In our next lesson, we'll show you how to do exactly that—starting with Find, the first step in the FETC framework.

‍