Intro to Hunter.io in ChatGPT | Find email addresses in your chats | Hunter.io

Hi everyone, welcome to today's webinar. Um, we are just waiting for attendees to join. So over the next uh two minutes or so, we will uh just introduce ourselves and I would love to hear from anyone who's who's live with us right now. Uh what is your current agent that you're using um for AI? Matt, what agent are you using >> this week? I'm still using news. I'm trying to get news researches uh Hermes Hermes Hermes Hermes Hermes something like that. Um agent trying to get that working. Uh and in the meantime while it's not I'm using chat GPT. [laughter] I haven't quite mastered some of the uh nuances of some of the platforms yet, so excited to dig in. How about you, James?

>> Um, you know, I'm I'm still sticking to Claude a lot. Um, I'm probably a little bit behind the curve on kind of playing around with the new version of chat, but have a lot of my work getting done with Claude. >> Ruy, I assume you built like your own agent. You're not uh >> No, >> no. [laughter] I I actually use [snorts] um a wide variety of them, but mainly it's clogged. Um I'm right now actually playing around with a lot with dropot um to test how far can I go uh in automatizing some of the things we do. Um, but I'm still very stick to claw ba basically because I've spent a long time investing in building my claw co-work uh >> memory let's say. Um, and so now I find it a pain in the ass to actually move to uh something else. >> Well, just wait till tomorrow or next week or the week after. You'll have all new things to think about.

[clears throat] >> [laughter] [snorts] >> Well, that is a a great way to kind of get ourselves into the, you know, the meat of today's webinar. Uh, so today's webinar for everyone who's who's joining us, uh, today's webinar is designed to take you through the process of connecting Hunter to your AI and running your first prompt uh, with a with Hunter inside your AI. uh it it is a great opportunity for us as well to hear from yourselves in terms of what uh AI you're using not only in terms of the tool but if you're using any specific agents uh please do let us know in in the chat you'll find the chat uh and the questions tab in the right hand side of your browser um please as we go along let us know any comments any questions any feedback that you have based on uh what we're talking about today the goal is to really help everyone uh to understand the power of of AI um with Hunter inside it. And so without further ado, um I'm joined today by both our CEO Matt Tharp and our product manager Ruie Costa. Um thank you both for for joining us and taking us through today's content.

Thanks James. And please, I would ask people just throw things in the chat. Like, do not worry about timing. If it's you're reacting to something five slides ago, please still put it in there. We want to make sure that we're we're tying things to the most important bits. So, uh we've got a lot to cover. Um James, I'll just uh chat about this briefly. So, our agenda, uh, we're going to talk a little bit about like why is a like what are some of the things that you need to think about with AI? What should you know? Look, we're also we've just said a bunch of AI generically, so we're going to clarify what those terms mean for us today so we're all on the same page. Um, how are you going to be able to use AI with Hunter? What have we figured out? What works? What, you know, what doesn't? Um, and then some ways to get started and then of course QA along the way, but a little bit hopefully at the end. um just in case. So with that, let's start with this because uh James and I had a funny little amusing topic where I said, you know, you know, all of these slides reference AI, but that might mean very different things to very different people on the call. So maybe we should uh talk about that. So we put this together just as a quick little look, for this meeting today, we're just going to say AI a lot. We're not going to try to clarify each nuanced [clears throat] little portion of this and I've tried to simplify the visual as much as I think is humanly possible without losing all

the meaning. So let's just do a quick primer on what we're talking about so that we can get rid of all the terminology and move forward. The first thing is at the center of this and this is what matters for today to be clear. At the middle of this is Hunter's API. And as you guys I hope know, Hunter has made every single part of the application available via API. And so that means if you want to essentially run your entire workflow end to end and never log into Hunter, I think you can almost do that completely now. Um, and if not, give it a week. It's very, very close. And I've done a lot of crazy things using AI and haven't had to log into the app. So I'm telling you, it's very possible and it's very exciting. So at the center is the AI or the API. Now around the API is something called MCP. And really all that is is a translation layer so that AI or LLMs can talk to an API. It's really all it is. It's it's like a little translation wrapper around an API. And so that means if you're using chat bots, Kimmy, Claude, Chat GPT, the the standard kind of off-the-shelf uh chat bots, those are going to use you can use our off-the-shelf apps or MCP, which is far more robust um to basically do anything you want through Claude, Chad, GPT, or anybody else. Now that's different than the agents because sometimes you might have something like Grockbot uh Gemini you might be using news um these are

agentic platforms or agents and you can also get some of that functionality from like chatpies dots uh and and claude uh is still I think a little bit behind but there's about 20 different ways that you can schedule things like an agent would uh using claude using anthropic. Now, you might also be thinking we're referring to what's in Hunter because we do have a Hunter agent um that's coming, but specifically we've released the first part of that for most of this year actually, which is the AI assistant that lives in discover and helps you do a lot of the stuff that's in discover that sometimes isn't super obvious. For example, my favorite is should I use industries or keywords or both? Great question. This is what the AI assistant's job is, is to kind of like allow you to pick one and then help you identify whether or not maybe the other option would have been better or has more or less or what have you. The goal here is to give you the flexibility to get what you want regardless of the platform. So if you're using CRM, they have their own AI agents. Those AI agents can communicate with our MCP. So if you're using inadin, if you're using Clay, Salesforce, HubSpot, doesn't matter. All of those have agents and all of those agents can talk to MCP the same way. They don't even need to be natively integrated like our Hunter agent does. So when we talk about AI today, essentially we're just talking about this whole thing. Um we're talk we're going to talk about individual aspects of it, but I wanted you to kind of get a sense for look, we're not going to drown you in

terminology. This is the end of that. We're going to now move on and get into more of like what I think is actually important for you to know. And with that, I'm going to turn things over to Ruie. Ruie, take it away. Thanks Matt. So before we actually go and connect anything and show you how how to drive this uh I wanted to start by actually showing you why this matters. Uh so we asked AI emails addresses uh sorry people's email addresses to AI just like you would and then we check the results and we use nine models for AI apps and this is basically what we found in this experiment. So on the left I'm asking Chach GPT to give me uh the email address of a particular person uh at at the company and you see that chat GPT there immediately immediately gave me an answer. So it answers right away Jordan.ly um and the the domain address it looks great but actually if you take that email address and you go to Hunter and try to verify the email address uh that email is invalid. So it doesn't exist.

So the benefit of then connecting Hunter to to shot JPT is shown on the right. So now I'm using the same shot GPT but this time it is connected to the Hunter MCP. And now I do the same question and what I I get I get a person I get an email and I actually get the proof. So Hunter found the email on on the website. He saw that it there recently and it's verified. So it's the same AI on the left. It's basically guessing on the right. It's actually checking to to know uh if the information is is actually true. So this is just one example. We actually took this um this example and and run a much larger experiment on it to try to understand how much does this uh affect uh your outreach efforts. So we tested extensively. We picked 125 uh different companies, real ones, big names, small names. And for each of them, we asked for two or three people. Um we also asked it in three different ways. I think in total it's yes 9 915 uh different prompts and then we checked for each of those what was the result that we got and the results are quite uh astonishing let's say. So in blue uh for each of these models uh you are seeing uh for the the the context that we ask for for how many did the AI actually gave us a result. So for example in chat

GPT we got a result for 42% uh of the prompts we we placed. However only 19.5% of those emails were actually valid. So that that's in total for the total requests that we made that's less than one in five of the emails that we've asked we actually got a valid one. The we tested for multiple different models. Um and the results are well chatty was actually the best one. So the other ones are even worse. Uh and the wor the most challenging piece is that for many cases uh the AI it doesn't say I don't know. it actually tries to guess. But when you put all of this together, that's the big number, right? 76% of the emails we actually got from the AI models were invalid. And um they were basically guesses. And why does sorry before that the first thought we had when we looked at these numbers we thought about does the is this related to the fact that larger companies have more uh information out there so they are easier but smaller companies are more opaque from a public web perspective and that's why we get bad results. That's not the case. actually uh the ratio of um invalid emails was pretty much consistent across all the different types of companies we had on

on the query. And why is this important? Right? So why why do must we care about the fact that AI is giving us wrong emails? first and be foremost like inbox placement. If you are using a lot of these let's say guest emails uh that AI gave you that are invalid you will likely go beyond uh 2% of bounce rate and that will put you already in a spam blacklist. So whatever you do after that that email that you're sending will be it's spam the best sequence best copy most personalized still your recipient won't get it because it's spam. Um it's also about time right so if you are using invalid uh addresses you spend a lot of time preparing those sequences um optimizing the copy trying to optimize the time to send the message but in the end nothing happens because again you you went to an invalid address and of course you lose revenue in the end right the the people you want to talk to uh never see your email because the email you got is wrong um and then there are other competitors of yours that are actually using the right uh email addresses that actually get to talk to those people.

So the difference happens when you actually add Hunter to uh one of these AI tools to basically combine the power of the AI agent with solid data that we have at Hunter and the results are fantastic like um we run the same test exactly the same test uh with SHA GPD uh with the model GPD 505 um with Hunter connected and it's interesting because in this case The best result was obtained by uh Chachi P plus Hunter was even better than Hunter. So we got the right uh email address for 78% of the time. Recall that before, let me go back. We were talking about getting 76% of the emails wrong and now we are talking about having 78% of the emails right. So in a nutshell uh and before handing over to you Matt in a nutshell that's the message right if you are using AI to in your pro prospecting process you should connect uh to Hunter to actually ensure that you are getting valid accurate data so that you don't jeopardize your outreach efforts Matt >> thanks Ry sorry it's the tiniest mic microphone icon, you know, like clicking it was like surgical precision. Uh, the

fun of UX. All right. Um, thank you. Look, before we move off of this, um, I think this is a really important slide and I think it's easy to gloss over this or interpret it differently and I I just want to double click on it for for a moment. Um, oop, sorry. Uh there's I think if you've ever sat on a prior webinar, sorry for the stammering, you know that every industry is a little bit different and so whether or not you can find something in this industry or that industry at the same rate really is it's something that fluctuates quite a bit. Big companies, lots more data. Small companies, a lot less data, as Ry said. But what's really interesting is that you can get a a net positive. So, it's it's like not just that it's as good as Hunter alone, but that it's this much better. It's partly because it's able to do some things in the what I would call the guessing process that are actually really meaningful to probably everybody on this call because at some point you're going to be asked about data waterfalls on your team. You're going to be somebody's going to bring up clay.

Should we do clay? Should we look at something like this? The same problem that creates negative downstream ramifications in your waterfalls is also what's shown here. That 76% of hallucinated emails is the biggest core problem of all of these data sets. Because if you think about a waterfall, it's it's only asking for a result that then gets verified at the end. And that means if you put the cheapest provider at the front, which mathematically would seem to make sense, it's also most likely the one that's purely guessing, probably is also the worst at guessing. And worst worst case is it's actually using AI to try to do the guess in the first place. That's what email finding really is, right? And so what is really important to capture here is that it says frequency where valid email addresses are found. All of this is predicated on the idea that the first step is getting the right email to begin with. That's why we focus so much on verified emails because you can't really tell until you verify the email. So, you might look at this and say, "Well, a 64% find rate isn't great.

78% find rate, maybe that's not great. I'm used to seeing higher numbers." Yes, it is true. If we just guessed emails and stopped there, we could have a much higher number. But that's not the point of this. The point of this is to get better accuracy out of your systems. And this is what we're excited about is that the combination of Hunter and additional systems can actually really create a much higher accur more accurate yield for your data. You put that through any pipeline and you're going to get better results. So let's talk about what that actually looks like. So how is AI lead gen different? Okay, Matt, Ruy, you guys are telling me you want AI to be a part of this process. What does that actually mean? Where do I put it? Look, you know the part up front here. It's just about telling the system what you want in words instead of actually trying to do it using clicks in a UI, but there's really four areas that you get to spend more of your time when you do it with AI versus not. So the first is list building. What you do is you tell it what you're looking for. you give it your ICP or if you haven't done an ICP exercise, you give most of the information that you would use in an ICP exercise like size of company, geography, various attributes about the businesses that make them more likely to convert. Some of this is in your data and you may not have done this process.

So, one of the things that's great is you can ask AI to do this process for you and make it that much more robust. contact discovery AI can just do that because of the first thing you define the list it can then go do the work and it can experiment with filters and industries and do some of those things without using up your time and I think that's one of the biggest benefits what this is meant to do is increase the value of the results you're getting while decreasing the amount of time that you have to put in to get those results that's why we're so excited and it's so powerful now on the verification side AI is just going to do this right you guys used to this with Hunter. We already do this. It's just a thing you're going to make sure, but it is an opportunity to go through, eyeball the list, double check it, make sure all the verifications are accurate. Enrichment, you don't really have to do this in Hunter. Luckily, we have Hunter automatically enriching, and we have for years. That's not new. That's just part of being a Hunter customer, being on the platform. But things like email drafting, this is one of the the hardest parts of outreach. I think if you ask a room full of people who do cold email every day, which part do you are you the most worried about? And I would say most of the hands are going to go up around writing the email. And the nice thing about this is it's way easier to get a hundred different variations of emails using AI than it is to just come out of your own brain. I do this all day every day looking at so many emails and even I can't really come up with that many

ideas. you kind of get locked into your first couple of ideas and it gets really hard to get out. AI doesn't suffer from that problem. Well, you can work around it more easily. And then there's the analysis, right? If you're using any of these systems that I had on that earlier slide, these things are all tied into more data than just what's in Hunter. And that gives you a better set of insights to determine how should you continue to improve moving forward. Now, all of these things by themselves without AI are incredibly timeconuming. but also highly valuable. The biggest difference between the high converting uh sequences and the low converting is this. It's being able to really deep dive on each of these steps in the process, which is now infinitely easier with AI than it used to be. So, not only is it fewer steps that you have to touch, but it's fewer hours touching those steps. And so what what that actually helps you do are things like look if you were trying to find target companies before it's a lot of manual work. I've been doing it for 20 years. It's not a ton of fun. And you kind of learn some tricks over the years. But what you can gain with using AI to do it is that things that would have taken you years to learn or hours to do is much faster. You can kind of set it, come back later, it's going to tell you the answers that you need to know to make decisions and then you make decisions and move on. It's almost like everybody becomes a leader, you know, everybody becomes the boss [laughter] when they're using AI. And if you do it right, it's just bringing you decisions to make. And I think that's

the powerful part is that's really where your human brain should be applied, not clicking icons in a UI, right? And then lastly, of course, uh verifying and enriching that can be more automated now, so you don't have to worry about it. You can add secondary and tertiary checks just to make sure it doesn't get out of hand. And then of course, you know, writing the emails themselves. So all of these things you can think about like here's an example of something that you could do when it comes to researching the lead or writing the email itself. If you were to do it manually, you go to their LinkedIn, you look up their profile, you manually read like I'm not going to go through everything you'd have to do manually if you wanted to, but it's it's a lot. It's timeconuming and you know there's tons of data. Now you also know what you want from the data. So it's not just that there's infinite data available. It's that there's probably 10 things you actually care about. You can send AI off to go do those things for you and then it comes back and again now you just incorporate that into your decision-m and so all of this becomes easier becomes more robust and less time consuming. But it's not set it and forget it because that's the problem that I think a lot of people fall into when they think about like AI SDRs, which if you haven't heard about it, I'm sure you will. It's essentially the idea that it's like an an SDR in a box. It's run by AI. It connects to an inbox, etc., etc. But the problem with that is that you're

sort of out of the loop. You want to know what the message is. You want to know which message is working. You want to know what the difference is between these companies and those companies. all of this becomes uh really critical. So here are the here's eight steps here's outreach activities there's eight of them that all can be improved if you do those steps with AI. So when it comes to like defining your ICP, um you could just take your CRM data, give that to probably any of the AI platforms, although I would use one of their more expensive models and say, help me understand, cluster these customers, break these customers down into a set of segments that that describes the customer group very well with similar attributes and common, you know, common aspects to their profiles. like you can describe in words what you're looking for and it can come back with a really robust ICP analysis and segmentation report. That's something that when you feed that in will make your messaging better.

Identifying decision makers, not every company has a CMO. So if you go into any sourcing platform and you say, "Oh, give me all CMOs, blah blah blah, do this, do that." It's going to skip a lot of people, a lot of companies because they might not have a CMO. Maybe they had a VP of marketing, maybe they have a head of marketing, right? You need to have the the opportunity to kind of go through and find these things. And this is something that AI can do nearly automatically. Um, I actually have moved to a point now where I do almost all my segmentation analysis with AI first. Um, because I have a pretty clear understanding of what questions to ask to make sure it's coming back with good results. Enriching context easier and faster. Now, in fact, most CRM have a have their own MCP, have their own agents who can talk to our MCP and our agent. And so, like a lot of this now can be done behind the scenes between systems in ways that you don't even really need to be directly engaged in until you get the results back. Um, scoring leads, prioritizing leads.

Honestly, who has time to sit there and look at a list of a thousand leads and try to prioritize those? Not very many. But it's beneficial if you can. It allows you to spend a little bit more time on the most important ones. This is something that is now extremely easy that was incredibly difficult before. The last four are very similar. Tracking buying signals, using those signals to identify things to do next, uh verifying lists once more, doing like a final check before you send. Um I've talked about the writing and then using performance data to improve. You can say, "Hey, AI agent, look at my CRM. Look at all my cold outreach, don't necessarily worry about the direct attribution, but let's see which of these is actually the best audience or how can I refine my message based on what we've seen are the final results." Um, if you have a salesdriven process, this is great. If you have a marketing driven process, you probably want conversion data, right? Same idea, but being able to incorporate those things into your outreach analysis allows you to experiment faster, experiment more deeply, and run way more data through the system um to make your next decision. It's it's like having a group of analysts just riding shotgun on everything you're doing. Now, I uh I think it's probably good to pause there for a second and just see, James,

how are we doing on questions? Do we have some that would be good to address before we move on? >> We we do have a few. There is one question that came in from uh Rachel and Dan plus one this around uh whether this is GDPR compliant. >> Yes. Yes, it is. Um we have a whole privacy section on the website. Um a whole GDPR section. If you go to the footer, you can find a ton of information on Hunter. Um I can't speak for every AI platform. Um you would have to look at those individually. uh that I can't ask I can't answer in real time which combinations remain GDPR compliant. Um and so that's something that that you'd want to look into, but within Hunter, yes. >> Great. Thank you. >> Absolutely. All right. Now, hopefully somebody on here besides James was thinking to themselves, well, wait a minute. Didn't didn't Hunter tell us that something like 70% or something like 69% of decision makers don't like AI written emails when they can tell that it's AI written. And it's the last part that's actually really important. It's not that it's just written by AI. Honestly, if you show five emails to people, if they can't tell that it's written by AI, you've won. Okay? But if they can, that's like doubly bad. So it's it's not enough to send a bad email, right? If it's a bad email and they think AI wrote it, that's like 2x bad. So, what we want to do is use AI

to make sure that the a that their email sounds well written, but also that the personalization attributes that are incorporated are are used wisely to look like a real email to to leverage all the information that AI can help you gather about your prospect. That's the other thing is like AI can go out and do a full debrief, a full research project on every single prospect if you want. Depends on how much you want to spend on tokens, but if you want a full enterprise sales class deep dive on a customer that's about to receive something, you can get it. It's where do you want to put that information and how and how refined is it? But look at what happens when you can incorporate just one more custom attribute. Right? Right? So, if you're talking about personalization, no custom attributes, 50% less than if you add two. When you go into editing, if you fully automate a sequence does okay, but if you edit by hand, almost a 20% improvement instantaneously.

So personalization is important. A lot of that information is publicly available and AI makes it easier to pull those things in and incorporate them into your emails. And if you do it, you can see anywhere from 20 to almost 60% improvement in terms of your things like your reply rates. It's really, really valuable. Now, here's an example if you just want to see what that looks like in my inbox. I can attest that the one on the left is much more what you're going to see. And it's a quick question. I hope this finds you well. It's going to use, you know, 15 minutes for a quick call. It's going to use so many of the like cut and dry straight out of Reddit, you know, cold email techniques. The problem is we've seen in the research as the cold outreach volume goes up that the reply rate goes down. So sounding like a cold email is probably not going to be your best friend. instead something that's like, "Hey, Sarah, I saw you're hiring for these things." That usually means what's happening as you're growing. We help you by reducing the amount of time. We help you. We record the best demo so a new hire can get productive in their first week and you can automate more of that process. Do you want a two-minute walkthrough? This is a much better version. Smaller ask, much better value communication. And the only difference between these two is one was using a basic template, something we've seen

everywhere, and one actually asked AI to just do a better job. I'd have to find the specific prompt. I'm sure that you can get that from James if you ask in the chat, but it's a great example of how much better a message can be. And there's nothing about this that particularly suggests that an AI wrote it. So, it's very clear. It does its job. Feels like a pretty good thing, a pretty good email pitch. doesn't sound like every other cold email I've heard. And nobody's going to read this and immediately go, "Oh, yeah, AI did this." And so that's really the gate that it has to pass. So what can you do with it? I'm if you are still awake I apolog GPT and say hey I got a list of 200,000 leads can you break that down into lists of 50 it'll do it so that's a much better experience when it comes to segmentation enriching leads you don't have to worry about that but that's something that can be better verifying lists. A lot of this is built into Hunter, but as you go outside of Hunter, there's ways that you can extend these even further.

So, without further ado, I'll let Ruie talk about how you can actually make this happen now that I've thoroughly convinced everybody that this is what they should. [laughter] >> I will actually end up by showing some of that, but I will try to keep it very very simple in the end. Um well basically four steps uh you connect connect a verified source of data uh hunter and I will show you how easy it is to do that. Uh you can define exactly uh how the AI can spend credits in doing whatever uh you asking the AI to do. Super important you give it context. So you tell them what's your ICP? What are you trying to accomplish? the more information it gets from you uh the better as Matt mentioned for example if you plug it into your CRM you can even learn uh what is the most successful customers you have and from there infer the strategy for you and then finally um obviously that a lot 99% of this is prompting um and if you don't have any ideas on how to start we actually have um a prompt library where you have tons of examples uh very very very tail for different use cases where you can actually uh lean in, use them, start to play around and get used to it.

Um, the more you do it, the better. The more you iterate with it, the better is also the experience I've been having well since day one. So, the first step, super easy. This works uh if I'm not mistaken to every uh AI out there. You just go to and I will show you um go to under.io aents. You will have a prompt there that you can copy paste into your agent. Just press enter and it will guide you through the connection into uh for example chatpt with with Hunter. So that's pretty much straightforward. Then a question I hear a lot is uh will AI spend all my credits? So to be clear, AI will use the the credits you have in your plan. uh same rules uh so discover so prospecting companies is free of charge finding in emails is one credit verification costs half a credit enrichment costs 0.2 two credits. So, nothing changes there. We only charge you if we actually give you results. Um, most importantly, the AI will always ask for your approval before spending credit. So, you are always in control uh every time you actually are spending something with with the AI. Um, as I mentioned, a super and that's the reason why in the beginning I mentioned that it's been hard for me to move away from cloud its context. The more the AI

knows about you, the more he knows about what type of product are you selling, uh whose comp what type of companies are you targeting, what type of value do you offer, what has been working, what has not been working, what has already been won. The more he knows about this, uh the better he will be able to assist you in every of of the tasks you have. So just to give you um three examples of prompts you can start to play around with after uh connecting Hunter to for example ship PT which is the one I'm going to show you next. Um the first prompt basically cleans the list. Emails go bad over time. People change jobs for example. So this prompt will basically take your list and rechecks everything older than four weeks. Split good from bad and tells you the cost of it first. So before actually proceeding it will tell you how much it will cost so that you you want to approve.

The second prompt is an example where you can actually leverage the fact that AI has the ability to go uh to the web and get information for you. So you can give them like what are my three best customers? AI can fetch that information, identify patterns, identify what's common between those customers and try to to help you then building a list of similar companies to those where you already been successful. Um, and finally, again, among hundreds of different examples, we picked this three, but finally, uh, another angle you can use is, as Matt mentioned before, um, looking at who's replying to your sequences, what's working, what's not working. Try, uh, also to, um, check if, uh, check if this is the right moment to engage to someone that you have already sent an email in the past. Improve your copy. So there is a lot you can do uh right there. So now let me try to share with you a few steps. So doing a demo I was just sharing this with Matt and James before we start. Doing a demo with an AI agent is always fun due to the unpredictability of it. But uh wait oops I pressed the wrong button. Just give me a second.

Uh yeah, for some reason. Ah, there we go. Now you should be able to see my screen. I hope so. Yes, there we go. So, as I mentioned to start with, um you just need to go to uh for example enter.io.mcp. If you go there, uh you just press the onboarding prompt. I will show you what that prompt is for ChachiPT, but it's basically something super simple. It's to instruct the AI agent to go to our web page and fetch the instructions uh to actually complete installation. Um then we do this. So it's basically reading now the uh instructions we have on our website and it will complete everything for you. Um, well, as usual with AI, it's not super fast, but we will get there.

There we go. As you as you can see, like we also have this information on the web page, but you don't need any API key. You don't need anything. You just need to go here which is the app store for shot install plugin and that's pretty much it. Then you approve the connection and it's pretty much done already. So now just one small detail. Uh usually you need to start a fresh a fresh um a fresh prompt for it to be able to uh identify that um you have already connected Hunter. Even if I type Hunter, you already see it there. So now you're already connected to Hunter. That was it. Uh so Shotg already has access to your Hunter account and is able to use uh the Hunter features to uh validate your data. So now now let me try to Yes, there we go. Instead of typing every time the the copy.

So I'm going to play you an example. Um this is actually a live example. This is actually true. Um so my sister and I uh own a property, an Airbnb basically u where uh one of the goals we have right now is actually to expand to the American market. So, what I'm telling the AI first of all is some information about me and I'm actually placing the the website of our Airbnb rental. Why? Because now the AI can go to that website and fetch more information and grab more context. So, I'm enriching the agent with uh more information about my ACP so that he can expand it. And then I'm basically telling him this is my goal. So, it's I want to to grow in the American market and find uh US traveling agencies and I want now I'm calling Hunter. I'm not sure if you notice how I did that, but it's basically at Hunter and you basically calling Hunter. Um, and I'm asking Hunter MCP to basically help me identify agents to reach out to. That's basically it.

Now, we just press to see what's going on. uh and it's um going to first fetch my website, fetch information about me to get the right contacts and then evolve to uh finding the actual companies with Hunter. So now it's calling Hunter as you can see he's already performed um a Hunter search and he's looking at the results and the hopefully uh fast he will show you the next step uh which is yes this is beautiful like right there is a massive power of using AI uh plus Hunter. So, first of all, it went to Hunter to perform a discover search to find the right companies. And then uh because I gave him the context of my property, he was able to identify that this is a property in Portugal. Um this is a property in the uh wine uh region.

And so likely my best ICP for for my goal which is to find travel agencies is actually to prioritize agencies that sell Portugal iteneraries wine tourism and families experience because I will skip you the details but that's actually very very very accurate to uh what we are looking for and based on that and based on the search he performed in Hunter he was already able to find um some interesting um agencies that we might want to partner with and he's already giving me uh some positioning. For example, he was able to understand that the best position for us is to be authentic by riverside near Porto and all of that. So right there is already a massive uh value in using AI as Matt mentioned. So if you did this in a platform, you you apply the filters, you check the companies one by one, you try to understand which ones fits the best.

Here AI did that all for you with just a single product. Moreover, if you want to uh you can still jump into uh enter application and see the results and then uh continue to operate there. But I will stick to uh using chat GPT. And now the next step is basically now that I found uh uh my ICP let's say my ideal uh customer profiles. The next question is who are the people that I should should be actually um being reach outing to. By the way, I always start with perfect thanks because the moment the machines dominate the humans will I want them to remember that I was a nice guy to them. Um but proceeding um so as you can see right now uh based on everything he knows he's already helping me identify within those agencies what within those target companies what are the type of people that I should be actually reaching out to and it's pretty much accurate uh the outcome and again simple prompt already uh is helping me trying to identify what should be my priorities uh within niche of the agency it has identified before who should I target first and so on and so forth. So the next step obviously is trying to find um sorry let me copy paste this is trying to find the actual people to reach out

to the actual emails. So what am I doing is I'm asking Shachi PT to help me find the the contacts of these people and then using Hunter to verify the emails to avoid the problem we just mentioned before which is getting invalid emails uh that AI elucinated. So hopefully this one will go fast. Let's see. So all the contacts is already uh in the AI. He's only looking for contacts within the agencies that he has uh shortlisted here. Basically prior already prioritized and we'll to take take a few minutes. By the way, I usually find cloud faster, but that's my experience. I don't know about you, Matt. >> Depends on which model I'm using. Um, yeah. >> Yeah. I put it to medium, but it should maybe make it even smaller.

>> If you go up, it'll generally just go slower. [laughter] That's what I found. So, if I need a fast answer, I slide the bar down and try to get a fast answer. But mostly, I'm just like, >> go run this process in the back. >> And and here we go. Like during the process, he already found went to the web, found a few emails. Now is the moment he's trying to verify the emails. And as I mentioned before, this means email verifications mean you will spend some credits within Hunter. So the agent will not perform that until you approve it. So you allow it and then it proceeds. But meaning that you are still in control of what's happening um to your credits within Hunter uh every time. And if I can add real quick, Ruby, >> when you're in when you're doing anything outside of Hunter, >> um, but it's using Hunter information or accessing your Hunter account, it uses the same credit pool, uh, credit limits, credit volumes, the things that are already predefined in your account. So, you set those things up once um, in your subscription, and then all of that gets governed by your subscription automatically. So, you don't have to worry about that. And as as Ruy pointed out, it'll prompt you before it goes off and spends your credits for you.

>> Yes, that's um a guardrail that we put. There we go. It's now giving us the results. Um yes, so it was able to find a few valid emails. So, as you can see, um all these emails that he found for the target companies are um valid and validated by Hunter. He's still on doing more searches. He also guessed a few emails and he was able to to check that um they were invalid. So to put a pause there if we were not using Hunter uh you would likely get those emails uh and use them and you will bounce and you by that you will increase the likelihood that you'll be landing in spam. So now as a final step we did that we use enter and and the shachi pt together to actually help me finding the emails for the target uh travel agencies that I'm looking for. So the final step is basically to what should I tell these people right help me writing my uh my outreach campaign. So to keep it hopefully fast, I will just stick to you to the top two contacts he found and ask for the first email to see what it does. And as you will see, um you will basically be able to use the information that Hunter gave you about those those emails, go online and check whatever you can learn about the company and the person to try to draft a very

personalized uh email message. Like this is this is for me in terms of the outreach copy the most powerful aspect of using AI. As you can see, it's already giving you a super personalized uh email. You notice that this person actually moved to a company uh is trying to find new opportunities and is writing in the email with a very succinct pitch to it. Then for the second person I asked an email um is grabbing the context of uh wow that's actually very cool because these two properties are very close to my property. So this is actually very very relevant. Um, and again a super personalized email that you get as an output. And just to conclude, if you want to, you can then just send these emails through Hunter. Again, it will not send the emails until you actually approve. It will basically create drafts within Hunter, but you will get um you will get those emails sent or those sequences. Here I'm just talking to two persons. So if a larger list you will create a sequence. As you can see he sees that I already have a sending account connected within Hunter.

It will create um the the sequences and and send from there. And that's basically it. So I try to keep it like super super simple simple example. But now you can imagine that you can expand this very widely uh and you can make it even more personalized. The thing I wanted to highlight three things the context. So the more he knows about what you're trying to sell like give them your give the AI your website give him like what have you been successful with? Give him your CRM. The more context you get the better. By connecting to Hunter, it will mean that uh the likelihood that you will get valid emails is much higher. Actually, very much higher. Um and then just iterate with it. Check what he's giving you. Give him feedback as well so that he can learn throughout time what works best for you and you will get a significant improvement to your results in a way much faster compared to today.

Thanks, Ruie. Well, uh, James, uh, do you want to mention the starter kit? Uh, I think if I think you've already shared it in the chat, and I just wanted to make sure you could. There you go. Uh, thank you. Uh, the link is shared in the chat. James posted it. But there's a bunch of different prompts. Like the goal was for us to be comprehensive, give you prompts for just about everything, but not make you feel overwhelmed with a library of 2,000 prompts. Uh so you could kind of drill into each one of those uh that are listed here for whichever part you want to experiment with um and start uh and start exploring them. All right. So, now we get to quick Q&A and uh let's start to take a look at some of the questions that have been posted. Uh we have a two-parter from Patrick. I'll grab that one, James. Um and Ruie, if you want to look at the next ones question part one, I use a personal Yeah. So, this is from Patrick. Uh thanks, Patrick. I use a personal chat GPT that knows a lot about our business and how I sell. my company has introduced Gemini because of concerns about confidential company information being used to train AI. Yeah. Uh great question. Um a common concern uh where even if you go in and I know I've certainly done it where I toggle that little button that says don't share my data. you know, I use a pro account, so you know, I could do things like that.

And even then, there's something about chat GBT. You you just like it leaves you questioning whether or not it's really being safe. I don't know. I think a lot of people are like that. A lot of businesses are doing kind of what you're suggesting, Patrick, which is like moving to like a a Google Gemini or something. Um, this is a great technique if you can do it, which is to teach one agent your preferences deeply, especially when it comes to writing >> and then to have that kind of sit there and run sidecar and do some of these things for you. Um, I don't know that, look, I'll be honest, I don't know that Gemini is great at this. I don't know. I haven't really tested it. Um, I know there's a lot of different ways to tweak better answers out of each one of these, so I can't be totally familiar with all of them just yet. Um, what I can say is we use Gemini internally as well. Um, we also use Anthropic and Chad GBT. So, we've got pretty good visibility across the major ones inside our walls as a company. And uh I think Gemini is pretty good at some aspects of it, but my candid belief is that it has not always been my favorite when it comes to copywriting. Um and so what I tend to do is, you know, if I need to use the company version, by the way, Claude versus Chad GPT is very much the same way. If you ask Claude to generate images in anything, it's pretty terrible at it. And if you ask Chad GPT to generate images, it's actually pretty

good at it. And so, for example, we're mostly clawed here, but whenever I have to do something visual, I just ask Chad GPT to do it. Every business industry is a little bit different. I think the LLMs are kind of like this as well. I we have some folks on our engineering team that are in love with Grockpot. So, like it's I think depending on what you're trying to do, certain models do seem to have some benefits and some some uh differences. I don't think there's any reason why you shouldn't be able to build a more secure version of your writing assistant though, Patrick. And so I think that's doable. To more directly answer your question, I'm trying to use news research to try to see if that's a better way to incorporate a custom model, multiple normal LLMs, and then to be able to route projects to different ones differently. Um, I've tried a couple agentic platforms. Haven't loved all of them for this. Um, I don't know if news is going to work either. But to answer your question thoroughly, I think that's the other thing is like you can use one of these agentic platforms to sort of route jobs to different LLMs based on the task. That's what I'm seeing work. Um, I just don't have a great recommendation right now for which one I think is awesome at this particular job, Patrick. So, I hope that helps. Uh, hope that's thorough. Uh, part two, I'd like an still like an AI assistant learns my sales approach.

Yeah. And then obviously just make sure you're I would say two things just to be safe. Um, don't ever let it have any transactional data. Um, you know, there's too much good information in that. That's probably dangerous. Um, and the other is I would say sensitive information business. Yeah. I I think the more you can keep your sensitive tools. So like for example, we use BigQuery and we're setting it up so that it pseudonmizes or anonymizes the results when we look at them. So we're not sharing personally identifiable email addresses through a process when we're running it in Claude. So like that's an example where I use AI to tap into something like BigQuery. And what it does is it intentionally in BigQuery we set something up so that it masks the data that comes back out. That's something that you can do to make this more more safe and secure, Patrick. And so I think there's a few best practices like that that that are out there um that that I can recommend and we can take a look at those.

>> But ju let me just chime in on on something there which is um Patrick mentioned also the topic of continuity. So moving from one to the other how can we pass the knowledge? >> Yeah. Uh that's the reason why I started with co-work because it uses the memory files that I have locally on my machine and I can move them around. But honestly uh you can actually use your existing uh agent to build your memory files extensive memory files in memory directory so that you can take that file and move it to another platform. It's not perfect, but it's usually good enough for you to move into a new uh agent and without losing uh much of the context that you had had in with your previous agent. So, can take a few minutes to do that because you need to iterate a little bit to make sure that all the memory that you actually want uh to pass to to the other agent is there. But once you do it um you move this file to the new uh agent that you you want to use. You iterate with it a bit to fine-tune some of these learnings. But this is pretty pretty much it. And within I would say 10 15 minutes you are able to actually start using the new agent already with enough context to to keep moving forward.

>> That's great. All right. Yeah. You're welcome Patrick. Kevin, so for us tech geeks, do you recommend we do this kind of custom process with an oh with an AI chat session? It's been a while since I used chat as opposed to a clawed code session or do you think there's leverage in getting an AI agent to run through the whole process? Um I'm in the middle of experimenting with both. Uh Kevin, having it go through the whole process and then at the end get an end result and then go back through and review it versus have it go through the process and stop each time it makes a series of or it needs a decision. Um I guess it depends on how much of your day you have available to do this. I tend to have to do it in bursts. And so for me, I like for it to go run and do some things in the background, bring me a set of decisions when it can't do anything else, and then I come back when I have a burst and I give it some feedback and clear a bunch of queue, you know, clear a bunch of tasks in the queue, and then I let it go off and do some things again.

I have found that seems to deliver a better end result, but it does drag out the process, Kevin. So I would say it needs more experimentation and your results might differ a bit but historically I have preferred that particular method especially depending on if you're earlier in the process or later like if you already know your message if you've got your message locked in and you know it converts well and you're not doing a lot of experimentation there um I think it can do more right if you're not trying to really deeply refine the lists in really specific ways but you're just kind of looking for a better way to break it down into smaller lists. I I think you can let it freelance on that. So I I think it depends on how picky you want to be in the process. Um that's I realize a bit of a vague answer, but I I it's accurate. All right. Sorry, I'm trying to keep up with the questions, so give me just a moment. Oh, James has one here. Is the Hunter AI chat a callable tool for chat GPT?

So um that's a tricky one. So the enter AI chat is not a callable tool for chatpt but under MCP is. What what does that mean is that you cannot within chat GPT ask our agentic functionality that we'll have on the our application to work for you. but actually don't need to. You actually need the data that enter has and that's why you should use the MCP as I've shown you before and that's that's all you need. Um so to that to that question the answer is no. Uh but Hunter as a data provider is uh callable tool uh for CHP as I've shown you. Uh, somebody really asked while you were demoing, why did you ask the AI not to give you emails? >> Oh, to keep it to keep it brief. I was trying to keep it brief and fast. That was it.

>> The because if you had asked it to do the other, it would go off for quite a while, right? Yeah. It would spend time refining the list and and doing all that. Yeah. >> So, so, so the the the initial prompt I gave him with a description, if I just say, I own this property. I want this type of I want to find agencies in the in the US to help me expand my business and I press enter. It does the entire thing. It goes from prospecting the companies uh identifying among those companies you found in in Hunter which ones are most relevant based on my uh my own property find the people uh suggest me copy suggest me campaigns he does everything and that's why I had to ask him one to be succinct and please don't give me the emails right now otherwise we'll just be looking at at chat GPD for five minutes typing stuff >> this is great Um, Enmal I'd like, if you don't mind, uh, my email, I'm going to post it in the chat. Uh, you had some interesting questions that were going in kind of a different direction than what we were covering here on the webinar, and I don't want you to lose answers to those. So, what I'm going to ask if you could uh email me directly, I'm happy to answer those questions um and give you a proper response and not something quick as we're going through this because I do I get where you're headed with your questions and I'm happy to to answer that.

Ruy, do you want to take this one while I'm reading the next one? >> Uh biggest question is how to use my AI with uh finder and filters to use as few credits as possible. Well, the the fundamentally the answer to that is use your AI to be um as precise in terms of what you use your ICP. So the companies you want to target. You don't need 2,000 companies. You need very specific small sets of companies all small segments uh using the termin the right terminology where for those small sets then you you actually find one to two person for each of these companies that are the right persona. So the right as I've shown in my example the right buyer for your message and then within that you build company lists you build the personas you build the profiles and then within Hunter you actually can move on to find people um that are pinpointed to to those groups. That's the most effective way I believe that AI can help you there is by allow you to do the segmentation identifying the actual right uh ICPS right companies and the right personas meaning the right pro people profiles that you should be looking at uh prioritizing among them to make sure that you uh are actually spending the credits on those that

matter the You know, there's well, while you're on this topic, there's uh one of the great things that I love about any of these AI platforms is that things that used to take me like two teams of people to get done are now something I can do completely self-service, which makes decision- making faster. When you're doing things like sequences and campaigns and things like this, you that's what you want, right? You want rapid cycles, rapid loops that allow each time you for to for you to learn something and take that into the next one. That gets you the fastest conver, you know, the best conversion rate improvements and the best yield from from what you're doing. One of the best ways is to ask AI for you, hey, there's two ways we could segment. Either we could build a list of 20,000 people with small segments and put them all together. Kind of like a bottoms up approach. Let me find a small pocket.

Let me find a pocket. Let me find Put all these little audiences together to create a big audience >> or top down. >> Hey, I want to get 200,000 leads in SAS. Okay, great. Now, let me break that down into smaller chunks. The nice thing is that you used to be limited more by systems than you are today. If you wanted to do either of those two approaches, they would be fairly time consuming for you to do historically. And now either one of those. And so what you could do is you could actually give AI uh whichever your preferred me is, you could give it both jobs and say, I want you to figure out which of these approaches is better for my business. do a tops down top down segmentation approach or a bottoms up segmentation approach and let's see which gives us the best segments that we can talk to. That used to be really hard, difficult or expensive and now it's not. And so you can really play around with things like that to try to see which of those give you better fuel to work with in your outreach process. Um that's kind of the thing is it used to be like everybody was trying to find the one thing everybody should do, right? It's like we would have come into a webinar like this and we would try to give you like three things to go do immediately that would apply to everybody.

We also know that that's what leads to not great cold outreach outcomes. The more relevant and the more personalized they are, the better the results. And what we're showing you today are all the ways that you can leverage these AI tools to do much more sophisticated things than you would ever be able to do on your own without having to spend a ton of money or hire an agency or lose the learning by handing it off to somebody else and hoping they're doing it right or an outsourced consultant or something. This allows you to keep that knowledge and keep that learning and drive those cycles as quickly as you want. that is preferred for for a lot of people. It doesn't mean you have to do it that way, but I think that's the cool thing is AI is creating a lot of flexibility for people to do things inhouse that they might not have otherwise been able to do before. And when it comes to segmentation, it's time consuming and kind of a pain. When it comes to email writing, it's timeconuming and kind of a pain. And so the things that have historically been barriers to better success here are getting easier. And I think it still requires your knowledge, your expertise.

Um, you who knows your brand and knows your market better than anybody. It still needs that, but it can do a lot of the other minutiae for you, which just makes life, frankly, a lot easier. Um, James, are there any other questions that we should try to tackle before we wrap? We are a little over. So, thank you all for coming. Um, there's nothing that I can [clears throat] necessarily see right now that is pressing to answer. >> Okay. >> Yeah, >> great. Uh, so I'll give a couple minutes. I'm happy to stay on a couple more minutes if anybody does have any follow-up questions they want to add. Um, for those of you that need to leave, thank you so much for being here. Um, Ruie, James, really appreciate it. Uh yeah, we'll just hang out and answer questions until uh we run out of time or until people run out of questions, which may have already happened. Thanks, Patrick. Appreciate it. Thanks, Kevin. Thanks for joining. If you guys have ideas about specific things you'd like us to do webinars, also please email after or put it in the chat. Um and uh almost happy Halloween.

>> [clears throat] >> True. It's almost there. >> It's almost there. Thanks, James, for posting all the emails. Thanks everybody. >> Thanks everyone. All right, I think we can close it out. >> James, any last words before you click the button? Before you hit [laughter] end? >> No. [snorts] Uh thank you everyone for for attending. Uh if you have any questions that we haven't been able to get to, we have taken note of them. Uh but please do reach us uh via our email address that's been put into the chat.

But for now, thank you all so much for joining >> and anybody who doubles their reply rate, we will feature you on the next webinar and you can tell everybody what you did. >> Let's do it. >> That would be awesome. >> That would be great. Let's do it. Thanks everybody. Bye. Bye.