The GTM Agent Harness: From Signals to Pipeline in Production

Julia Nimchinski:
And next, we welcome Ellen, co-founder of Warmly. What a treat! Super excited for this one, and and can’t wait to ask you what’s in your agenda Coas. I know it’s warmly and Claude, must be.

Alan Zhao:
In my agenda for today, or just in general?

Julia Nimchinski:
in your AGENTIC OS, what’s…

Alan Zhao:
Oh, gentlemen.

Julia Nimchinski:
Because we… yeah.

Alan Zhao:
Yeah, it’s, just for me personally, because I also run marketing, it’s a lot of Claude hooked up into all the APIs that we have access to. And then a lot of codecs running in the background, too.
create these self-iterating loops, because what you do is you give access to all your systems, and then you provide as much context, and you try to build these very custom-made, one-to-one apps on top of all of your specific organizational data. So hooking up to Google Ads, service account, LinkedIn through Zapier MCP, Meta through their API, these are all the ad platforms that we have.
And then also creating, because we have all the intent data, creating, like, the system of feeding signals and pairing it against CRM data to create the automated outreach across LinkedIn and email.
So then that’s… those are the five surface areas that you cover, and then all the while, saving down the MD files into CloudMD, To then use that as context for further iteration once you get more data pulling in from production.

Julia Nimchinski:
And can’t wait to get into it, Ellen.

Alan Zhao:
Awesome. Alright. Well, today, I think this, this, this, we’ll be talking about Warmly’s demo, about how this all works, and then if there’s time at the end, I’m happy to go through my own process, which is,

Julia Nimchinski:
I would love to have, yeah. If you can do it, that would be amazing.

Alan Zhao:
Alright, I’ll try to be really efficient with the time today. And for those of you who don’t really know Warmly, we started off as intent data providers, so we were really focused on first-party signal of website de-anonymization, and then we built on top of that an inbound agent, we’re calling it, which is basically a chatbot.
And on top of that, we built personal landing pages, smart pop-ups, retargeting. And the whole idea is that the website is the place that, without question, it is, it is a signal, whether or not you need to act on every single website visitor, is determined based off of your company, so we have to filter traffic.
But the idea is, like, you are trying to optimize anybody who comes to the site and knows about you. So that’s the first part of our platform. The second part of our platform that fewer people actually know about, but, we’re very excited about, is called the TAM agent.
It’s in order to bring more traffic to your website, we built this TAM agent that builds your total addressable market Because we also have our own contact database, our own ZoomInfo provider, and we have the intent signals, so together, you can map out your TAM To deliver, from a demand gen perspective, bulk campaigns. For example, for us, myself, I had to run a Drift with Sunsetting campaign.
Drift is a chatbot, they sunset about 2-3 weeks ago, and just needed to pull the list of anyone who’s ever used Drift, and put them through campaigns. I’ll show you guys how to do that in a second. So these are the two. It’s one thing to bring traffic to the site that you care about, the other one to convert the traffic and to retarget them, once they already know about you.
And underneath it all, I think everyone’s already heard about this today, but we do try to build our own version of the context graph, which is the interconnected nodes between people, companies, employment, signals, deals.
And then, ultimately, what you want to get is the buying committee of the companies that you care most about, that you’re trying to target, and we do save down the activity feed, like, we call it the ledger, of everything that’s happened to each person and company, and we’ll show you that in a bit.
Underneath this all is 200 million contacts in the database, 10 plus vendor waterfalls just for website de-anonymization, and then we have first, second, and third-party intent, like Bombora, for example, but also new hires, job openings, news. 10Ks, 10Qs, and then we hook it to your CRM so it’s seamless. And then underneath it all, we also connect to a lot of your favorite tools.

Alan Zhao:
Something that I think people have been talking about is once you get to the level of, I have all the data and execution is basically infinite, then you need to create these decision engines, which we call, like, these, like, trust gates.
And the reason for this is because you only have so many emails that you can send out a day, so many… even fewer LinkedIn messages you can send out a day, so you have to use that really sparingly. If you’ve already sent them an email or a LinkedIn message, you don’t want to send it to them again.
And these are all, like, the little micro-decisions to make sure that, everyone’s getting exactly what they need and nothing more. And then evaluation is like, okay, now that you’ve seen the results of it, you want to backtest it into, oh, sorry, you want to update the policy layer, which is basically the decision layer of your agents, so that they make better decisions in the future.
And then these things inform our two agents right here. So, coming into the product, like I was saying before, we started off with website visitor deanonymization. So, you already have the full visibility into anyone who’s ever visited your website.
So, this is where they are on the sequencing, this is, pages viewed, and then if you want to look inside, you can see all the activity, the chat messages, the pages that they visited. This is just a random one that I’m pulling up. We also have… oops, sorry about that. Refresh this. We also have social intent, so you can follow key people.
Let’s say we’re looking to follow any time chatbot, conversation marketing, or Drift Alternative is mentioned. We can see the posts of people mentioning these topics, and then we can straight up directly send them into an orchestration. Orchestrate. And what that does is, we, for these posts.
this is mostly LinkedIn, we scrape the engagers of the posts, and the likers, and the commenters, and then you can put them into sequences, or push them into, LinkedIn ads, or webhooks, or wherever it needs to go, or your CRM. So this is our signals agent, and we have a large list of signals here.
We have a bunch of other signals, identified contact visitors, Twitter, social engagers, chat engagement, Bombora Research Intent. We have templates, so it’s very easy to use. And as an example of what happens is, for any single one of these, like.