Antimetal uses artificial intelligence to monitor and analyze AWS usage patterns, enabling companies to save up to 75% on their cloud bills without needing to commit to long-term contracts or perform extensive manual optimizations.

When Antimetal reached out to The Kiln, they had 50,000 people in a Clay table, 100 inboxes warmed up, and were looking for strategy and implementation help to get a campaign off the ground.

"We wanted the ability to automate outreach in a personalized way, instead of send each email out individually", said Harrison Cuker, Antimetal's Head of Sales. "It would take forever do send this to a sales team, and we didn't have the time to do it without compromising pipeline."

The main issue that Antimetal wanted to solve was the personalizations themselves. With 50,000 people, there are very few commonalities that could be shared in copy while still reaching out to the entire list.

Here's what we did to create a campaign that fit the model:

1. Define what personalizations would be relevant to Antimetal's offer.

   We decided on running what we call an intent signal waterfall- this is used to provide relevant personalization to a list of contacts that otherwise have no connection to a company. It's easy to provide personalizations when there's an intent signal (ex: they interacted with a LinkedIn post), but the intent signal waterfall is meant to find something interesting to mention in the copy.
   
   To make this determination, we synced with Harrison to select 5 that fit the bill. These could be anything that show a buyer might be high intent- operates in an AWS intensive industry, is above a

   certain web traffic threshold, etc.
   
   

2. Rank the intent signals by strength, importance and cost

   Once we decided upon intent signals that signified a potential interest in Antimetal's services, we had to figure out which ones were most important, as well as those that were the most credit efficient. This involved deciding which Clay integrations were required for each intent signal, how many credits each integration costed, and how that compared of the strength of each intent signal according to Antimetal. Once we had those sufficiently ranked, we could move on to building out the table.
   
   

3. Writing copy

   There are two components to writing copy for an automated outbound sequence- the template and the personalizations. The template is static- no matter who we're sending the email to, the elements written in the core template remain the same. The personalizations are variable- these components of the copy can vary depending on the person, the segment the person is in, the company the person works for, or pretty much any differentiator that is worth mentioning or distinguishing from in an outbound campaign.

   
   Antimetal's offer was already incredible, so writing copy for the template itself was quite easy as it relied heavily on the offer itself (save up to 70% on AWS without spending anything). The main point of thought surrounded the personalizations themselves. Because they were generated by AI, we had to determine what kind of sentence we wanted the AI to write depending on the intent signal. We wrote 5 intro lines for the AI to model off of, one for each intent signal. Once we had copy finalized, it was time to create the automation
   
   

4. Creating the Clay table and optimizing for credit efficiency

   Once we had the spec created, we went into Clay and created the Clay table completely. From a technical perspective, the intent signal waterfall is by far the most cost-efficient way to create variance in personalization for a large list of cold leads. It works like any other waterfall in Clay does- first, it checks to see if a prospect matches an intent signal. If they do, then it stops all other integrations from running and writes the intro line using AI. If the prospect doesn't match the intent signal, then the next signal check initiates, and so on. This involves adding conditional runs to the Clay table to prevent unnecessary credit spend.
   
   Once the flow was built out and tested on 10 rows, we ran the entire 50,000 leads. We used an OpenAI API key to save money on credits, and this significantly reduced the cost of the project. From there, we sent the leads to a campaign in Instantly to send at high volume.
   
   "We're already starting to see positive responses a day in. It's incredible"
   
   This was a fun project to work on due to the scale of the lead list- with just this template, Antimetal can run outbound on this offer for as long as they'd like.