Connect Clay, the ultimate outbound tool, with AI Agents that can answer anything from websites.
I've gone down the rabbit hole so you don't have to.
Here's a how-to-guide on powering up your outbound on #Clay with custom #AIAgents. π€
A bit of context first.
What's the value of an agent on outbound?
ChatGPT assumes if it lacks info, but agents deep dive for answers.
They'll scrape and research thoroughly until they have a good answer.
Example.
You're selling AI services and want to check if a company is already using AI.
If you ask chatgpt: "check this website and tell me their AI use-cases" - it'll flop.
An AI agent is better equiped to answer this.
It's problem with it? Reliability.
It takes a few tests to get it to make accurate results.
Don't take it from me - take it from Clay's top expert @xaviercaffrey13
If you don't mind taking that risk - that's it for all folks π
But if you do care about this... I've found an alternative.
π‘ What if we connect Clay to a different agent? Is it possible?
- Yes. It is π
I dived into the rabbit whole and will skip you the learn curve.
I'll show you how I connected Clay to an AI Agent on Relevance.
Why did I do it?
1. I'm relatively fresh to Clay - so this was a learning exercise.
2. Wanted to test out Clay's possibilities to integrate to external software.
3. Wanted to experiment with Agents once again π
The use-case was to have an agent research and scrape though a website to find use-cases of AI for that company, then present the findings as:
- "I really like how you're using AI for [use-case]."
1. Pick an AI agent provider
I've done my dose of experiments with @RelevanceAI_ and was confident this could be possible with them.
Plus they have pre-build templates making it easy to get started.
3. Test the agent & refine
Make sure it's getting the output you want.
4. Now - let's get the API key information from Relevance.
This will get a little bit tecnical - but don't worry (nothing GPT can't guide us though).
5. Now go into Clay
You should have a column with company websites for this example.
Next steps will feel tecnical - but beleive me, nothing a bit of ChatGPT can't solve quickly.
We'll be setting up 2 API calls from here.
1. A column with an HTTPS POST api call to activate the agent and send it the custom information we want it to enrich. In this case the company domain.
2. A column with GET call to get the response from the agent back into Clay.
I'm not the most saavy - so ChatGPT helped out on this stepπ
What I did was:
1. Provide ChatGPT the API documentation from relevance
relevanceai.com/docs/agent/agent-api
2. Give the API keys to Relevance
3. Ask it to guide me step by step on what to input into Clay's API connector.
And it worked like magic. It told me exactly what to input into Clay and worked on the first try.
Now, once you've ran both API's succesfully (getting "200" as result), you'll be able to extract the response from the GET Api.
And that's it! πͺ
Last things to consider:
β If Clay Agent gets fixed - that's a 10x easier way to do itπ€
β Even though Relevance is more reliable - it's also a more expensive.
Test credits will run out fast. Yet Overall, not too expensive for this type of enrichment.
This thread serves as a tutorial for
β How to connect an API to clay
β How Clay agent works and it's limitation
β How to connect an alternative agent to Clay
β And the possibilities it can open up for outbound