Building an AI Workforce for GTM Execution

Julia Nimchinski:
And our next session, we are talking AI workers for real GTM execution, and Neha, welcome, such a pleasure. How have you been?

Ameya:
Hey, Julio, really good to be on. I’m doing really well. Excited to come here and talk to everyone. Just curious about this format. Is it… is it interactive, or can audience, reply in chat?

Julia Nimchinski:
Yeah, so we are livestreaming the whole thing for the audience, we are just with the speakers, we’re obviously on Zoom, and they’re watching us on the side. Yes, it’s interactive, they’re sharing their questions on Slack, DMs. And communication channels. But yeah, before we get started, Amaya, I’m just curious your thoughts on what’s the biggest misconception when it comes to AI co-workers?

Ameya:
Yeah, I think the… well, I’m talking to a lot of marketing leaders every single week, rolling out our AI solutions to that, and the… Three things that I’ll say they keep getting confused about is It’s hard to understand what an AI agent actually is, since everyone’s doing agent washing on SaaS platforms. And then there’s some business folks that are more mature in their AI usage that are using things like custom GPTs. I know folks that are using NHN workflows and things like that as well. There, they really get stuck thinking that. advancing beyond that in their usage of AI requires more technical know-how. So the biggest one is that the right way to use AI is to treat it like a teammate. So think about creating and curating context, onboarding procedures, documents, knowledge sources, writing down and describing your business process like you would to a teammate you were onboarding. And if you approach it like that, you can actually create really, really sophisticated AI workers faster and easier than you ever thought possible. So… I think we’re also in that adoption cycle where Everyone has entered the trough of disillusionment after the initial hype, so there’s a lot of folks out there right now, and a lot of companies saying they can do really incredible things with AI, but when you actually get into their platform and product. it’s not the case, so… that’s what I’m seeing.

Julia Nimchinski:
Love it. Where would you put 2026 in terms of the Gartner hype cycle?

Ameya:
Gosh, I mean, I think 2026, we’re definitely in a trough of disillusionment right now in the hype cycle, but to be… to be completely honest, I’m not a big fan of Gartner. I think it’s a pay-to-play organization, and I don’t think anyone really even takes the reports seriously. I don’t ever read them myself. No one’s asked us yet what we’re ranked on it.

Julia Nimchinski:
Well… That’s a spicy take. Live events unfiltered. Let’s move into the demo.

Ameya:
Yeah, let’s pop in. I’m gonna keep the slideware really, really tight, because what I actually want to do is just show you guys, AI workers running. So, let’s just present quickly and do a quick intro to Everworker. So, good to meet you all. My name is Amandaishmuck. I’m head of marketing at EverWorker. No, I am not an AI agent talking to you, even though something very strange appears to be happening with my camera lighting in the backdrop here. But yeah, we’re a worker, we help businesses do more with more with AI, and what we help you do is grow revenue, reduce costs, and get a competitive edge. with a always-on agentic AI workforce. So, a little bit about us, like, well, we’re well-backed, VC-backed business, and the entire exec team, including myself, has been in the AI space for a number of years. I’ve been at a couple AI companies in the HR and recruiting space and marketing prior to this as well. So, really, what we’re seeing in the market is This is where a lot of people are at. AI chat assistants using things like GPT projects, starting to figure out how to prompt and manage context. Go ahead, Julia. Question? No? Okay. And from there, we started to see the market evolve last year with AI agents. So, think about, like, how broad N8N knowledge has become, and the emergence of really, like, dev tool-focused platforms to create AI agents with code. And these are pretty cool. You can do different things, like skills with them, and execute tasks a lot faster. Claude’s co-work feature helps you do this as well. But what we’re focused on delivering is AI workers, so think of these as Full-on digital employees that execute an entire business process from end to end, and at their highest level of maturity. they’re operating as always-on employees that you basically set an objective to, and then set and forget, and they run at scale autonomously. That… future vision there. We’re nearly there with one of the AI workers we use, which is our SEO marketing AI worker, but a lot more are starting to get there, and they’re all available in our templates. So, really, what an AI worker is, is think of it as an extension of you and your own expertise as a marketer. So, it’s able to analyze, underside, descend, decide, execute, and then learn continuously over time, and this is all enabled through the platform. And really, one of our core contentions when we looked at the market was that business people need an AI platform that’s built for them and not developers, so that’s what we’ve built. And then furthermore, when we looked at the other paths to AI from the different sales conversations I have, like. Late last year, most things were failing, and they were failing in POC and demo, because IT was building them. And then, when I’m talking to folks that are getting quotes from consultancies, they’re getting quotes like $60,000 for a 6-month AI strategy roadmap with no delivery, or if you’re talking to BCG, they quoted a large supermarket chain in the Nordics that we’re working with. $6 million for AI agents and told them it would take it 2 years to deliver them, which is just an absurd cost, absurd time to market, and, like, the value just isn’t there, so… We’re actually priced at about 10% of all of those. 60K will get you up and running with 15 AI workers, and we’ve built an entire catalog of AI worker templates for sales, marketing. finance, customer support, HR, recruiting, and we can deliver and customize those to you really, really quickly. So, that’s enough of a pitch. Let’s go straight into… the product itself. So, the three things that, given we’re only got 20 minutes left here, there’s not a whole lot I can demo, but feel free to reach out to me, and I’ll show you guys all super deep dive demos of all these. But the three ones that I’m going to show today are going to be our SEO Marketing AI worker. If time permits, I’ll also show our Email Nurture AI Worker. And then our SDR AI worker as well, and then we have a white paper and e-book writer and designer, too. So this is the core of the platform. So let’s go take a look at the SEO AI worker as a starting point. Cool. So, this is one of our templates that’s available. This all starts off with an input where you just give it a persona. So here we’re gonna say, Head of Marketing. Gonna run the workflow. Now, what this is doing is it’s calling a first-step AI worker, which is a persona plus keyword researcher. And here’s what that looks like under the hood. So this is how you create AI workers. The first thing is we gave it knowledge. So, here I’ve given it knowledge about all of our different functions, segments, our messaging and positioning. our product, our offerings, how we help different business leaders execute things, and these are actually RAG databases and vector storage you’re looking at right now, so it’s really easy to contextualize agents here. Setting these up in our UI is actually just like managing a Google Drive or SharePoint folder. You just drag and drop upload documents, and then name them. And then to connect them into AI workers, I can just set up a RAG pipeline by checking a box. That’s it. It’s done. So that’s one of the core things about us. The other component is brain, so I’m picking GPT-5.2 to run this. And then here is our instruction set, right? So… Here, I’m telling it what its role and objective is. Here, I’ve given it a complex instruction set to query its vector memory. Query our messaging and positioning memory, and then go off and do really complicated keyword research across a number of vectors. Here, I’ve told it to output as a JSON array, and then I’ve given it guardrails and fences on how to do that. Skills in there are real-time web search skills, so it can go and do deep research, and that’s it, right? So, that’s the AI worker. Now, let’s see where it’s progressed in its keyword research. Great. So, it got the query from my workflow, it’s done its research, and now it’s come up with 15 ideas for content. So, let’s go back over to the workflow. We’ll skip over these two nodes, this is just some technical wizardry that we put in so it could parse the JSON. Now what it’s doing is it is calling the worker… workflow that is going to research, write, design, upload to HubSpot, and then publish directly into my HubSpot CMS. So let’s go take a look at that, because this is pretty fascinating as well. So here, it’s getting each persona and keyword as an input trigger, and this is calling a blog researcher, writer, and optimizer worker. It’s saving that into collection, so it doesn’t repeat blog articles that it’s done previously already, so this is how it gets smarter over time. It’s writing metadata here. Here, it’s passing the entire article it generates to another AI worker that generates a prompt for Nano Banana to design a header image. Here, it is calling a HubSpot skill to upload the file, for the header image it designs. And here, it is calling another HubSpot skill to go and create the actual draft in our CMS. So… Let’s go take a look at where we are. Yep, so this is running one article at a time, and then sending the next article after 10,000 seconds to the AI worker, so the whole thing takes about 5 minutes for it to run and publish, but we should be able to start to see. Some drafts coming in. Okay, great. Here is… Some example drafts that it’s put in. 5 minutes ago, 7 minutes ago, this thing’s constantly running for us. So, let’s go and take a look at one of the outputs from my previous runs while we’re waiting. So, as you can see, it drops a nicely designed header image right in there, drops the full article, the instruction set for this optimizes it for AI search, for standard SEO search as well. It’s really, really well written, nicely structured, and it handles internal linking as part of its instruction set as well, so it’s gonna link to good pieces on your site. And from its research phase, it is grabbing external links that are relevant and high quality. So, again, you’re building that link repository in there. And what else is also interesting is both the depth of the content, but the fact that, depending on the intent of the article, that it classifies, it’s gonna put one of three potential calls to action right there in the article. It’s top of funnel. We put a call to action to start your certification journey inside AI Academy, which is our, free online certification offering to teach business people how to use AI better. There’s our upcoming webinar where we’ll demo even more of these for 2 hours. Then right at the bottom here adds an FAQ section, which is also really important for AI search visibility. So… This is how we’ve been able to just scale like crazy, and this is how our blog’s performing. So, we’ve been going to market for about 3 quarters now. Every 30 days, our views on the blog are going up by about 50%. And then, if we look at the… The way our traffic is performing. Let’s look at versus last quarter. We’re up 230% on AI referral traffic, and then, as you can see, our organic continues to climb. month over month as well. I think HubSpot tends to misattribute the direct, direct site visits for… as… as not organic, as long as, you know, I doubt, like, 1,500 people are inviting directly to our blog pages. But this thing works really well. We’ve rolled it out to, 5 different customers, and we can get it up and running for you in about… 3 hours for the core worker logic, and all we really need from you there is your messaging and positioning guide, your personas, your ICP, and then some stylistic elements. for how you want your HTML coded, but that is the SEO marketing manager at a really high level. Let’s go and take a look at… the SDR worker next. Okay, so here is the demo version of our SDR worker. So how this works is we’re going to grab a segment from HubSpot. So, let’s say… I don’t know, white papers list. So, now I’m gonna grab the list ID. We do this for the demo version. The actual production version of this just runs automated on triggers, so it’s reaching out automatically as new contacts enter various lists and segments. I’m gonna put that list ID right into here. I’m gonna give it a limit of 5, and then the campaign name is gonna be… demo campaign. So, this is gonna grab contacts from that HubSpot list, and then this is gonna call a… workflow for our SDR worker. So here, it is getting the contact data as input, preparing that, and then it is running the SDR AI1 worker for follow-up, which… this is what that looks like under the hood. One pattern you’re going to see here is the knowledge is always similar across all of these, because we’ll create a unified knowledge system for all your marketing and sales AI workers, so they all are on brand and speaking about you properly. I’m using GPT-5 Mini this time, giving it a quick little roll and objective. Instruction set is telling it. What information it’s gonna get, then it’s gonna classify, based on the job title into a intent category, and then based on the intent, job title, and company context research it’s done. It is gonna use web research skills. to find relevant insights. And based on those, it’s going to write a personalized four-email SDR sequence, that is completely personalized to that individual prospect, following some good best practices there as well. Output format, it is going to do some Simple output formatting for JSON, so it works in a workflow. So, let’s go and take a look at its chats. That’s where you’ll see… So, here’s the first contact they got. This gentleman from, at Bentacar.com, and interested in financial operations and sales. And you can see, yep, more sessions are opening up as it is being told to process additional programs. And then here is the initial configuration of it prior to putting it in the workflow when I tested it. So I gave it this guy. Bryce York, Director of Product at Atari. It did this full research report on this company. signals, like, deep, deep enrichment and research that you probably previously would need to use a vendor like ZoomInfo or Clay or other sources to do. Now you can just do it with an LLM with a web research skill call and a good instruction set. And here is the type of email sequence that it generates. So, really good subject line, and then really, really great personalization, right? So, we often get replies to our emails like this, like, oh, wow, really great research, yes, I’d love to take a meeting. We’re seeing about 5 to 10x over the standard 2% reply rate that you can achieve with outbound, using this approach. Now, let’s go back to the workflow. Yeah, so the reason I did 5 contacts at a time is because it does a lot of research, so it takes a while to run, but as soon as that finishes, it goes and runs over to another workflow. which then actually creates the contact, directly inside of our email sending tool, which we like using Email Bison. So here’s our demo campaign. So, as you can see, it goes and puts in a completely personalized four-email sequence for each individual contact. And what’s really cool about this is, you know, in my previous life as a marketing leader, anytime I wanted to up-level my SDRs outbound, I had to make about, I don’t know, 30 or 40 different segmented Sequences, they were never super personalized. And then managing that operationally was always a nightmare with HubSpot automations and workflows. With this, you can be as simple as just one sequence with custom variables that all of your leads go into, if you like. What we do internally, and what a lot of our customers like to do, is split them out by intent trigger, right? So, we’ll have sequence for different PDFs, like e-book downloaders, for all your web Webinars for all your events. for your demo requests, and then for your segment hypotheses for your outbound. And the reason we do that is that way you can just use the native reporting in your sequencing tool to figure out, like, okay, this sequence is This signal is getting good reply rates and meeting rates, this one isn’t. And then maybe you refine your lead targeting for input there, or maybe you refine your AI worker instruction set to improve the performance of that, but that lets you get good traceability on all of it. Okay, let’s go and… Take a look at our… Email Nurture. or an AI worker, or workers right now, let me go to workflows. Okay. So, here’s one that I just built. Here, I’m going to give it a prompt for, an email… 3-email nurture sequence, and it’s going to write it, design it, and create it in HubSpot automatically for me. So, interest, let’s say, So, here again, I think you’re starting to see a pattern here with how we architect these things. It is calling our email nurture Writer AI worker. Which… looks like this. Again, same knowledge, same brain, different instruction set. So, again, interest classification. Then reading content on our website pages and articles that we have that are relevant to that interest. And then a approach to creating a three-email structure, some guidelines on rules, and some basic formatting. output is going to be that. And then final instructions, again, are going to be your guardrails and principles. No skills needed here. Let’s go take a look at its chat. Okay, great. So, here’s the prompt I just gave it, so you’ve got your three chat sessions here. And it’s created 3 email copies with subject lines, right? So, let’s go back to the workflow. Where’d it go? Great. Okay, now, that passed that entire chat to these email extractors, so they’re each gonna extract one of the unique three emails from there. Then they pass that email copy to a header image designer, and they also pass it to a file name agent. The header image designer passes it to, a nano-banana scale, with a good prompt for it, so it generates a header image. That then uploads the image to HubSpot using another scale, and then it sends it to one more AI worker that’ll create that email in HubSpot with HTML formatting. So, as that runs, I can show you examples of ones that we’ve ran Previously, which I did, because I didn’t know how much time we’d have on this call. So, here are examples from this AI worker. So, it’s got a right from, which is Austin, who’s our content marketing director. It’s got a great subject line, good preview text, and it’s generated a on-brand image header for us, as well as formatted the HTML properly for this, and put in a good CTA. to our end-to-end AI Workforce Solution for Marketing page. So, really think about using this design pattern to create any sort of email sequence or email that you would like. You can do anything from a, email newsletter, to nurture sequences, to pretty much anything you can imagine. I think one of my gaps that I always ran into historically was I never had time to create nurtures to cover all of my segments and interests that were really well written. Now with this, you can just personalize away and, really have no gaps or limitations in your bandwidth or output. I know we’re at 5 minutes here, so… If there are questions, I would love to just open it up to, audience questions or anything like that.

Julia Nimchinski:
For sure. Thank you so much, EMEA. This is fascinating. I have a lot of questions, but first, let’s address one from the audience. Question from Joe. Can EMEA speak to human and loop checkpoints in this process? For example, link validation, source validation, campaign IDs for marketing automation workflows, etc.

Ameya:
Yes, so I think a great example of that is our… SEO Marketing Manager, which the first iteration of it was, Human in the Loop. And really, doing Human in the Loop is as easy as modifying instruction set. So… what I told it to do in the original variation of this was, I added in gates. Now, you know what? I think that might have been in V2. But, there we go. Step 9. Once the user approves, then proceed to step 9. Once the user approves the linked article. Then proceed to the next step, so on and so forth. So, the reason I did this is because when I first built this, I wanted to make sure I nailed the instruction set so it was producing content perfectly every single time before I scaled it. So, that’s the approach I’ve taken with designing all of these, is start with human in the loop, validate, iterate rapidly. And then perfect the design pattern for the instruction set to make the non-deterministic deterministic. And then fully automate. So that’s how we roll it out to our customers as well, and that’s how you can design it yourself. But everything is customizable to however you want to run your process.

Julia Nimchinski:
Love it. I’m curious, how do you see the end user here? Because at a certain point, marketing just became, I don’t know, like, dozens of micro-functions, and, I mean, you showcased it as an example, per use case, from SEO to design, we haven’t touched on that, but I saw that briefly, content marketing, and lots of micro elements, but Are we transitioning into an era of, you know, a marketing orchestrator with, like, basically deep acumen into setting up all of those micro-agents and macro agents? How do you see it?

Ameya:
Yeah, you know, it’s interesting, I really don’t see it as anything different from what we’ve been doing as marketing and go-to-market teams broadly for the last 10 to 15 years, right? So, we got in, CRMs came out, started to get into automations and basic workflows. now there’s those GTM engineers that are using clay tables and all of that to go and do things. So this is just a natural evolution of that, and I think what we’re seeing in our customers is we’re coming in initially and then rapidly configuring our templates to them. So first they just get an understanding of curating knowledge and creating knowledge source as well. Then they get to understand how to use the workers that we’re deploying for them in those first initial weeks and start to get results. As they do that, they start to unpack and understand the design patterns. So we train them on how to leverage the platform, which there’s an agent that helps you build all these agents in there. So they build their own. But really, the ultimate end goal and outcome for it is that every single person on your marketing team becomes an AI-first operator. So, they just go and describe their business process and get really good at documenting it, and then the platform takes care of the technical nonsense around creating AI agents, which, frankly, the technology is not complicated. It’s figured out already.

Julia Nimchinski:
Alright, so essentially, every worker is almost like a hub for your full marketing team. Do you just specialize in marketing, or is the plan to expand?

Ameya:
No, we’re a horizontal platform, so we’ve got, similarly to what I showed you here, we’ve got prepackaged solutions and templates built out for HR, recruiting, finance, we’ve got a really great APAR worker there, customer support, customer experience. Supply chain, shoot, business operations, pretty much anything you can think of. And we can do full custom… custom builds based on whatever you want there as well.

Julia Nimchinski:
Great stuff, Anea. What’s the best next step for everyone watching?

Ameya:
Connect with me on LinkedIn, and just reach out to be there. We also were doing another webinar, which we’ll invite you guys to, but… what I always do is, when I go and talk to marketers, I’ll go help you figure out where to use AI workers best, and I’ll actually just make you a worker for free that’s completely tailored to you, just so you can see that we actually deliver on our promise. So, just come talk to me. I’ll walk you through it.

Julia Nimchinski:
Appreciate the transparency and the statistics and analytics as well. Thank you so much.