Clay: How to Get Started

Top 6 Skills to Learn

July 25, 2024 • Estimated Reading Time: 8 minutes

Hey, Claymakers!

Everyone asks - how and were can I get started with Clay? What should I focus on?

If you missed it, Clay summer school start soon - we announced cohort training.

The next cohort starts July 29. It’s a 2-week intensive crash course covering Clay's most important features (Prompt engineering using AI, no-code formula building, etc.).

We also have Clay University (short video how-tos), which is a good primer.

I’m biased, but truly believe that Clay is becoming a must know for GTM. Whether you’re using Clay at your company or personally, I predict that Clay will be a differentiator for job applicants, too We’re already starting to see an uptick in Clay being mentioned in job descriptions.

So what should you learn?

If I were to start learning from scratch today, I’d focus on the following capabilities:

  1. How to do a company or lead search.

  2. Order of operation for adding enrichments. (They all have their own prerequisite data points.)

  3. Formulas

  4. Claygent

  5. Prompting best practices

  6. HTTP API

Below are some tips for each:

Top 6 Skills to Learn

1. Company & People Search

The standard search starts off with a company or people search that looks like the LinkedIn search you’re used to.

However, most people don’t realize there’s several other ways to start searches:

  • Find local businesses with a Google search - very helpful if you’re focusing on physical location of companies.

  • Storeleads - a DB for eCommerce data.

  • CRM Imports - great if you already have named account lists.

  • CSV imports

  • Ocean.io - for finding lookalikes

Whatever it is, you ultimately want to analyze the firmographics & technographics of your best customers so you can search based on that criteria.

The best accounts are usually those that

  • Convert at the fastest rate

  • Have the highest ACV

  • Match your bullseye use cases

2. Enrichments & Order of Operations

This was a lot of trial and error for me. But, It doesn’t have to be for you. Here’s an example how you need to get certain data to unlock new enrichments (due to their inputs)…

Company Name —> needed for —> Find Domain —> needed for —> Enriched Company Profile —> needed for —> finding the employee count and other information that you might find on a Company LinkedIn profile.

3. Formulas

This is the easiest one. You just need to know how to write clearly in English.

Going to [Add Enrichment] and then [Formula] brings you to the next screen below. This is where you write in natural language and then Clay will use AI to generate a formula you can apply to your data. Some common examples are:

  • Extract the domain URL from this email [Email Column]

  • Output a green check box if [Employee Count Column] value is between 50 - 500

  • Scoring accounts or contact by doing something like:

    • Add 5 points if [Column A] meets a specific criteria.

    • Add 5 points if [Column B] meets a specific criteria.

    • And so on…

Then you’ll see what the output will look like so you can confirm if it meets your expected outcome.

4. Claygent

Claygent is our AI Agent that you provide prompts so it can automate all your team’s manual research at scale. Pretty much any publicly available information can be scraped and brought in programmatically.

There is a lot of unpack here. But, I have 2 tips that will help you extract more value faster.

  1. Our prompt library is growing and we now have over 150. A great combo is Find a 10k URL and Analyze a 10K for key initiatives.

  2. Try out the Neon model if you want more structure outputs so you can break out the information into different columns.

5. Prompting Best Practices

Prompt engineering can be frustrating, and it does take some iterations, but you’ll be amazed at what you can do. That’s why the former Director of AI at Tesla thinks “the hottest new programming language is English.”

Here are 11 quick tips you can check out my LinkedIn. It provides more details and examples of what a bad and good prompt looks like.

1. Be as specific as possible2. Define the scope of your ask3. Specify the format that you want for the output4. Assign a persona to ChatGPT or your AI Agent5. Describe the style you're looking for6. Describe the tone you're looking for7. Give a timeline, or date range8. Provide examples of what you want the output to look like9. Define location of where the AI Agent should look10. Suggest effort ... this is one of my favorites... and surprising how it makes an impact: "Be as thorough as you possibly can. My job depends on this."11. Seek explanations

6. HTTP API

This was the hardest to learn, at least for me.

HTTP API is how you connect Clay to basically any service that has an API. For example, you could connect to ImportYeti to pull shipping information (suppliers, weight of shipments, last shipment and more.)

There’s actually a ton of these very specific databases. If you are using one, and doing manual lookups, then you can probably automate that data coming into Clay and save a ton of time.

This is also a huge unlock for creativity.

Someone who was interviewing automated the analysis of a LinkedIn profile, used AI to write song lyrics specific to that person/company, and then used an AI music service to create a rap video.

That’s all for this week!

Around The Kiln

Fellow GTM Engineer, Matthew Quan, made a video and announced our new Salesforce package. This is how you can add an “Launch Enrichment” button in SFDC to kickoff Clay tables that finds the right leads for an account, researches them, enriches with work email and phone number, and then creates and uploads those leads so they are ready for prospecting within a minute.

Reply

or to participate.