[0:00:00]

Let's get into this. We're talking about outcomes today. We're not leading with AI intentionally, we're leading with outcomes because AI is just one of many tools as a part of a larger system, hopefully. Right. To get you to the desired outcome. And I am joined today by some lovely gentlemen. I am Em Wingrove. I'm the Chief Marketing Officer at Aptitude8, an elite solutions partner in the HubSpot ecosystem. And the gentlemen that are joining me today are former co workers, former boss and now a really friend. Current partners, but actually friends. Right, Introduce yourselves please.

 

[0:00:32]

Hi everybody, I'm Connor. I run a company called Happily. We are a HubSpot Native Event Management solution, which is a really odd term but basically we, we sort of make HubSpot into a full end to end event solution for registration, to check in to tracking the deals and the meetings that you got. We have a mobile app that you can scan a badge and sync that lead data back to HubSpot. And so we're doing a lot, we're building a lot with AI, we're building a lot of software with AI. We have a lot of customers sort of figuring out how do we leverage AI. And I think the big trend that we're seeing is as more and more of the digital experience sort of gets really noisy with a lot of AI slop, we are seeing a lot of people invest in in person experiences and so we are helping a lot of customers with that. And there are a lot of cool ways you can using AI technology to be able to have really killer in person experiences. Maybe out having a whole events team, which I think is one of the demos we'll get to talk to about a little bit later. Also

 

[0:01:34]

on that topic, I'm going to be using your AI lead capture tool at an event in a couple of months. I'm very excited for it.

 

[0:01:42]

We just updated a bunch of stuff. If you haven't opened it recently, do update. Because we found out we don't force you to update. And that's mostly fine, but sometimes not good to know.

 

[0:01:53]

All right. I'm Ryan Gunn, I am the founder of Attribution Academy, which does exactly what it sounds like. We are teaching HubSpot marketers how to do attribution. We're going to get into in a demo later how AI can play into attribution. So stay tuned for that.

 

[0:02:14]

Something that we skipped actually that I think is relevant for this audience also is Gun and I are both instructors for the HubSpot's AI just totally neglected, which we do the next session for next week. The last one was the biggest bootcamp that HubSpot's ever done through HubSpot Academy. We have a new one next week. I do not know if registration is still open for that or not.

 

[0:02:40]

Registration is still open for boot camps, but ours filled up in less than a day. There is a wait list, though. I would still recommend join the wait

 

[0:02:48]

list because we're doing a bonus session, I think, at some point, and it'll go out to the wait list first. But we. We also teach the AI Bootcamp with HubSpot Academy, which we know a whole bunch of stuff about HubSpot and AI, which is why M had us here and we totally neglected to tell you about that. But that's also true.

 

[0:03:06]

Okay, so what. So what you're telling me is that there's probably not two other people who know more about AI and attribution and. And all of the marketing, sales and rev ops and all of the things. Well, let's get into it. So we'll set the stage for you. We'll tell you how we're feeling about what we're seeing, what we're feeling in terms of just AI on the market, the fluff, the speak, all of the AI speak. We'll talk about how folks are adopting it and why that's hard. We'll talk about who's adopting it, teams, individuals, what's going on here. We'll talk about outcomes, which is the key driver of this webinar. And then we're going to show you the best part of the webinar is we're going to show you actual solutions that are powered by AI that are really. They're built for the outcome, and you can build them. And they're not that hard. They're not that risky. We can show quick value within your organization. So we'll show you those. Go through some key takeaways and some Q and A and all that good stuff. So with that, let's set the stage. I mean, AI is everywhere, right? Nothing's that clear for me. It's almost getting, like, funny because I'm seeing at this point almost verbatim, like headlines in ads or subject lines in emails. It's a lot of fluff and not a lot of guidance or direction. Are you guys also seeing that?

 

[0:04:24]

Yeah, definitely. It's. It's something where, like, I have friends who are not. Who don't work in the tech space, who, like a friend who's in construction who is hitting me up and being like, okay, what's the deal with AI, like, I feel like I need to learn this for my job. It is truly everywhere. And the messaging about it is everywhere. It's reaching industries. They don't usually care that much about what's going on in technology. So I think the, the, the panic is happening. I think the, the reality is that, like, you're probably, if you're behind, you're probably not that far behind, but now it's the time to start learning.

 

[0:05:07]

My barber asked me about Open Claw last week and I was like, what is happening right now? We're all cooked. I would say that I oscillate between which I think is right, is that every new, every new advancement, every new thing is, whoa, the world is different. This is crazy. Everything is upside down. And then you sort of, I think, orient back to where the shifts are actually happening and what's actually real and not. And I think to, to Gun's point, it is really challenging because the, basically every stakeholder that you might hear from stakeholders, probably not the interest is maybe the right word. Right. Is either trying to tell you that this is going to supercharge their business, their section of economy. You should invest, you should jump on now. Valuations are crazy, you're missing the moment. Or this isn't real and it's not going to happen and it's not interesting. And I think, you know, the truth is somewhere in the middle, where there are things that it's doing exceptionally well. And what I tell everyone, and especially analysts that I talk to that are like, is software dead? Or actually is software going to make everything incredible because of AI stuff? And you should go build stuff with some of these AI tools and you will both very quickly get to a place of wonder and amazement about what is possible, and you'll start to realize where a lot of those limitations are and you'll find out that the truth, sort of, if it's somewhere in the middle. So I am grateful for us to be able to have the platform, to be able to talk about what's real and not real, as opposed to being forced to be like, shareholders say AI is the best. You should accept that it's the best and it's going to change everything. Right?

 

[0:06:49]

And I think your attitude to anyone's attitude toward AI is heavily driven by what that noise is closest to them. Right? So if you have an executive that's on that one end of the spectrum being like, this is fake, it's not real, like, don't even worry about it, then those folks are probably going to have an attitude of like, I don't really want to do this, don't try to make me change. So the spectrum is pretty fast, but I think you're right, Connor, it's somewhere in the middle. Why is it so hard? I will tell you. One of the things that everybody says, but it's actually true is because it's predicated on data. And if your CRM infrastructure is not clean or set up in a way where, you know, a knowledge vault or an assistant can't access that information or that context to be able to drive the AI value, like, yeah, sure, of course. But I think a lot of it is that attitude too. Like, it's not just the data. What, what do you think?

 

[0:07:40]

Why is it hard?

 

[0:07:41]

What do you, what are you seeing? What are people saying?

 

[0:07:43]

It's.

 

[0:07:44]

I, I think data is dead on. We, we harp on this a lot in the AI boot camp, that it's, it's a garbage in, garbage out scenario where if you have bad Data and most CRMs do, then yeah, you're, if the AI doesn't know the difference between good data and bad data, accurate data and inaccurate data, so it's giving you its best guess based on what it has access to, and if you can clean that up, you're going to get better outputs. But the other thing that I think is a factor here is how things are integrated. And often AI is being used as a separate system and not an integrated one. And that creates a lot more manual work. It creates inaccuracies. It doesn't have a complete picture, full context of what's going on. So that's also going to create, you know, degraded results.

 

[0:08:42]

I think that to, to Gun's point, I think you kind of have two different forces going on. I think you have this, you have both fear and innovation, but both of those forces, like you have both of those coming top down and you have both of those coming bottom up.

 

[0:08:56]

Right?

 

[0:08:56]

So like on the, the fear side, top down, there's this like, well, we don't want to use this because we're freaked out about doing the wrong thing and having it be, you know, to customers or we're freaked out about data security or whatever it is that sort of the, the fear and the anxiety at the organizational level is. And then at the same time, you have the top down innovation, which is everyone has to adopt this, everyone has to do this. We're not hiring anybody, we're not doing anything. Everything we buy has to be AI first. And like that comes from a top down basis. And I think at the same time you have the same thing happening bottom up, which you have people who are saying I'm really freaked out about AI doing this and like where's my role? And I don't have a job. And so let me, you know, obfuscate or delay or try to avoid interacting with this. And you have bottom up innovation of people saying whoa, this is amazing. And I can use this to do parts of my job and parts of my work, even though it is not yet systematized in the organization. And I think it is really difficult to align both of those forces because you have both of them happening at the exact same time. And so I think some of what you sort of talked about here in terms of you have independent users, you have independent use cases. I think what we see most often is you have bottoms up innovators that are, that are figuring out how to do stuff really close to the metal. And I don't think that the top down innovation push is working super well right now for most people in most places because it just sounds like you should upend everything you're doing. And people hate change. And it, and sort of what we talked about at the front end is that the, it's really hard to find the signal through the noise and it's really quick when you bump into somebody and be like, oh yeah, see, I told you it isn't going to work for us in this scenario. And I think all of that makes it really challenging to get people to actually use stuff.

 

[0:10:42]

Yeah. Because the top down mandate usually isn't prescriptive. And I'd be curious too in the chat, like what are you, what is the biggest challenge or blocker to people getting curious and wanting to adopt, let alone adopting AI in your organizations? Because to your point, sometimes it's the top down innovation, sometimes it's top down mandate and they have different sort of outcomes or impacts. And it's, it's interesting because I do see such a variety but somewhere in the middle I think is where the most motion or like movement can happen. Like the completely bottom down or completely top down or bottoms up. Top down, excuse me, I think is just difficult because it's further away. Right. The two points are further away from one another. So yeah, let us know in the chat what are the big challenges in getting sort of AI pulled into your organization? And one thing you said, Connor, is like people hate change. I've heard this saying before, people don't hate change, they hate being changed. Right. That, that force coming, whether it's bottom up or top down. Like people don't, humans don't really like that. So I think that's part of the challenge of getting adoption is getting people to want to change, right? To feel like this is their decision.

 

[0:11:51]

One of the things that Connor and I have talked about before is the idea of like doing a pilot or like involving some people who are already enthusiastic about it and kind of showing results that can be achieved there. And then, you know, once you expose that success to the rest of the team, you're more likely to get people excited about it than if you're just saying like, use AI, it's going to make us more efficient. And then they go and they like don't know how to prompt and they put something in, they get something bad and they hate it.

 

[0:12:25]

Right. I think the real like currency here is value. And I think when someone has an AI solution that works for them and their, you know, little self contained daily work stream, I think that creates the best possible fuel to go tell everyone and share and oh my God, this is actually changing my Life. I've saved 17 hours this month because I have been summarizing all of my company notes or what have you. I think that that value is really the key to unlocking adoption. Because if it's just speak, if it's just pressure or urgency from the market or from an executive, this doesn't feel very motivating. But when someone is speaking from the heart that this has changed their life and if we could just scale it to this team and then from that team to the full org, I think that's where the magic can start to happen. I'd be curious, what are some of your like your daily, you know, this is my individual independent AI use case. What do you guys use most often? Share with the group, please.

 

[0:13:29]

How to answer that question. I am fully obsessed with Whisper Flow and any of the.

 

[0:13:40]

I'm so excited to hear about this because I. Are you doing it? I don't use it.

 

[0:13:45]

I have never used it. But I keep hearing, not even joking, like using Whisper Flow into Gemini Claude, whatever you want to do under the sun. And then extracting that out will make you feel like typing messages is just the most Neanderthal behavior. Like it feel like. I feel like I. In a meeting, I'll be like, give me one second, me one second. I'll like mute myself to be like. Because it just like is life changing. So basically the workflow is. It is a. It sounds really silly, but it's a dictation. Tool. The dictation tool does an amazing job of getting like names. If you start listing stuff it'll automatically format it as bullet points for but you basically can shift from and everyone sort of has the experience of that. You talk to Siri on your phone and like what it types out is terrible. This is the amazing version of that and it actually works exceptionally well. But when you do whisper flow and then you also use whisper flow using any AI tool you, you can start from a position of just talking and explaining what you're doing, what you're trying to do or what you're trying to communicate. You add some context and you're like what I want back is this information. And you will just get something so well structured. So a very tangible example for me is we're working on a paid ticketing product. We're actively trying to figure out how we're going to price it and because it's layered on Stripe Connect. So I'm like here's Stripe Connect's whole pricing page as a link. Here is me talking to my computer and just explaining here's what we're doing, here's what it looks like, here are my goals, here's what I want to do, here's sort of the product mock ups and the things that we're doing like help me sort of think through and put together a pricing strategy in this whole thing. And that exercise is so enhancing and collaborative and then we take that and kick that over and use that as engineering spec and it is magical and I can't recommend it.

 

[0:15:31]

HubSpot should build something like that into

 

[0:15:33]

their I think Apple should build something

 

[0:15:36]

like that in Speaking of which, I don't know if I'm allowed to say this or not, but I heard that that functionality is coming to ask Elephant.

 

[0:15:45]

Yeah, it already has.

 

[0:15:47]

They okay, okay.

 

[0:15:48]

I'm alarm okay.

 

[0:15:49]

Yeah. It's on a public website page so I also think the naming for it is adorable. It's called peanut.

 

[0:15:58]

Yeah, I love it. One of the use cases that I employ is just data analysis. Like taking out a bunch of objects that are associated to one another, pulling them into whatever large language model you want and then having it make sense. Like you know, obviously prompting this is the problem that and what I'm trying to solve but it's just so fast. Like if you're just breathe is great

 

[0:16:24]

actually like the breeze copilot is amazing. I I a huge portion of my job is to ask people what is going on with X. It's like a huge percentage of what I spend my time doing. And now I don't bother those people anymore. I go to our renewal deal, and I'm just like, hey, is this renewal deal gonna close? Like, why is this date wrong? And it's like, oh, yeah, Ben already emailed them on this day. They already said it's fine. Like, the subscri is wrong. And Ben, you know, message somebody else on the team and like, here I'm like, oh, great, I don't have to bother Ben now. And Ben appreciates it, but I think that's been really impactful.

 

[0:16:58]

I did that this morning with Breeze. Like, I couldn't find my DQ reports that I had built, and I was looking for, like, another, like, cohort layer of data. So I went to Breeze and I was like, typed it in and I literally just gave. Gave me their report, built it for me, and it gave me a table too. I was like, no, no, no, I don't want the report. I want a table. So, yeah, Breeze is pretty cool. Gun, do you have any, you know, go to's?

 

[0:17:20]

Yeah. My what? Well, I'm actually going to give an example that. That my wife Erin does because I love it, and she uses it, like, every time we go out to eat. She has trained a custom GPT on her taste in wine.

 

[0:17:37]

Love that.

 

[0:17:38]

So she will, like, we'll go to a restaurant, she'll take a picture of the wine list and be like, which one of these will I like? And give it parameters like, I want to be in this price range, or, you know, I want red and not white, and it'll send back something that, like, nine times out of ten, she's like, I love this wine.

 

[0:17:58]

Since you're talking about personal use cases, I may or may not have built a spreadsheet of different dates that I've gone on and had you with some of the data from their profile to have it help me. Because, you know, sometimes when you're in it, you're not as aware.

 

[0:18:13]

Yeah.

 

[0:18:13]

So it's my extra eyes for awareness. That's a good tip for anyone else on the dating in the dating pool. All right, so going from these, you know, independent, sort of contained use cases of leveraging AI, how do we get it to the teams like we talked about? Sharing your wins. What. What are some other ways you can actually get it to scale out? Like, technically? Right. Like we talked about, Like, I got to get the attitude there. People have to want to buy in. But what are some actual ways to scale it?

 

[0:18:46]

A big one for me is defining inputs and outputs making it really easy on your team where they don't have to like craft a prompt, they can just upload a document or like tell it some basic level of information. And the back end context is providing all of the, you know, details of how you want it structured, what, you know, what you want the output to be. Because if you just give it to your team and expect them to write perfect prompts every time that they're. It's not going to happen. Like somebody on the team is going to think differently or you know, want to enter the information differently. So if you really tightly define it then you're going to get more consistent outputs and your team is going to like it better because they don't have to spend, you know, 15 minutes writing a perfect prompt.

 

[0:19:43]

Right. And I think it's yeah. Tighter sort of requirements inputs and outputs. And also it doesn't have to be the V3, V4, like a V1 is fine. Yeah, I think a lot of folks try to bite off a lot more than they can chew from a technical perspective of like solve for everything at once. I just don't think that that's the way to do it.

 

[0:20:03]

I think that point, the value here is also building for workflow. So I'll give you something tangible that we're doing right now is like to get adoption of this, which is we know from all of our attribution data and our HubSpot data that social is a big driver of demand for us. And when we dropped off on social at the front end of the year, everyone was really doing a lot of it. Then we went to holidays, then we came back, people got busy like they just did less. And we saw that meaningfully impacting our top of funnel. And so we were sort of looking at this and saying we need everybody to be a lot more active on social. How do we equip them and enable them to do that? And so we could have gone and said, oh well if you guys go to Ask Elephant or you go and leverage AI tools like you can write drafts and you guys will be on top of this. And like that's sort of just the top down use AI to solve this problem. Instead I think approaching it from a workflow perspective is significantly better. And so what we ended up doing is we are leveraging Ask Elephant and we built a bunch of agents in Ask Elephant that monitor for great customer quotes. They monitor for interesting use cases that prospects are coming in with. They monitor for wins and onboarding or anytime somebody's like, wow, this is Amazing. We extract those, we compile all of them, we push them to a unified Slack channel and then we are using social prompts on top of that to then write first drafts from the voice of people on the team. And the difference here is, is you could look at this and I think the AI maximalist view is just automate all of that and post to everybody and spew out a ton of slop. And we know that both that won't be impactful, but we also know that it's not what our team wants to be putting on the Internet. Right? Like they don't want to be putting that out as their voice. And so instead we now have this feedback. We have a weekly that's scheduled with everybody that go through this feed, get it to write that first draft, schedule out all of that content. And now we're scaling the social output from the entire organization. And the way that we're approaching it is an AI is unlocking that at a level and scale that we couldn't have done otherwise. But we're approaching it from a workflow first view of how can we enable people in this workflow? How can we use AI as part of that solution as opposed to use AI to solve this problem and like automate the entire thing, which I think ends up resulting in a very inauthentic non human experience.

 

[0:22:27]

I don't know, I just thought of too another really good enabler. Thinking about HubSpot and I'm thinking about how their AI features are so front and center across every corner of the entire platform that you're going to see that little pink diamond somewhere. The breeze, little diamond. That's one way, right? Hey, we've never talked about this. I have no idea what it is, but I saw it on my screen and I know everyone else that also looks at contact records every, every day also sees this thing. Right? Like it can start the conversation because my understanding and please do not hang me in the street if I'm wrong about this, but for Salesforce, a lot of, you know, their AI features are behind a paywall. It was like for a while it was all or nothing. And HubSpot doesn't have that sort of culture attitude towards AI and that translates well into their product. But it also means there's like AI stuff everywhere too. So could start the conversation, but it can also maybe be a little chaotic. Yeah, I mean we talked a little bit about this. Like what, what enables the shift. It's small wins, it's showing the value, creating the conversation. Right.

 

[0:23:25]

I, I want to actually go back and take that, that subtitle. Small wins become scalable systems. This is something that I've been running into recently because I've been helping a client out with some, some AI workflows. And one of the things I noticed when I came in is that they were trying to do way too much in a single workflow action or a single prompt. And a big key in terms of making it scalable and making sure you get consistent results is break it up. Like instead of having one workflow action that is a gigantic prompt, break it into five or six steps that are each designed to do one thing and then combine them all at the end once you've got accurate and consistent outputs from each of those. So that's more of just like a tactical recommendation, but it will drastically improve how, how rapidly you can scale this stuff.

 

[0:24:26]

Yeah. It also makes things a little bit more scary when it's all shoved into one area where the functions are happening.

 

[0:24:32]

There's so much room for things to go wrong when it's just one gigantic thing.

 

[0:24:38]

Yeah, I mean I kind of feel like that just about workflows in general. It's like the age old do you do it all in one big workflow or do you break it up into little pieces? And with AI involved, I definitely am going to be on the break it into smaller pieces camp.

 

[0:24:49]

Something I'll share that maybe is relevant that Emmy, I'll just think this is cool. You'll both think it's cool, maybe you won't. I think you'll think it's cool. Something we're doing right now at happily is we're working on essentially post event recap emails and we're, we're sending it to every user so that you went to an event, you did the lead scan thing and we're basically going and saying okay, and what we want to do. Right. Our vision and dream and pre. I think about this from a marketing automation standpoint, right. It's like here's your total number of leads you captured. Here's sort of like basic data. And we're essentially rethinking that to say our ideal email is actually a you just did this event, here's how many leads you captured. Adding sort of surprise and delight moments of, you know, that is if you captured eight leads, that's one less than the number of planets in the solar system and giving sort of like fun number facts and extracting some insights from all the conversations that they had and actually saying these are the most valuable conversations you had. This is what you said your next steps are and sort of here's how everybody on the team performed differently in sort of each one of these sections. And then this is the next event you have coming up. Like this is what your priority should be. We were looking at this and saying how can we take all of the data and all the interactions and the voice notes and the information we have? And to your example of the one shotting, we're actually building that email off of four distinct AI prompts that are happening. Because what we found is that trying to do all of it in one was way too much information and the output wasn't very good. And instead looking at this as like this one email has four different sections and each section we're going to feed to different things and we're actually using different LLMs for each one. So like the, the social analysis element of like, these are the insights from your conversation. Like OpenAI does a really good job of whereas the numeric analysis and sort of like these are the cor next steps you want to do. We're using anthropic for. And by just breaking up each piece of that and looking at that as the sort of like modular AI workflow yields way better results. And I think that that's a very narrow slice of action that I think can be extrapolated to anything else you're thinking about doing, which is the more modular and narrow you can make it, that builds up to that greater whole, the better outcome you're going to get at the end.

 

[0:26:58]

Oh, that's sick. You should. Sorry. My wheels are turning. I was thinking if it could access the CRM data and then also say like based on your average deal size and your sales cycle length, you're going to have an ROI of about X. Yep. Yeah. Push that release. Like.

 

[0:27:19]

Yeah.

 

[0:27:19]

The gap I think is really just that people are especially that top down mandate vibe. It's very feature centric, like AI as a general thing. And then you know, insert like breeze for example. But like what does that mean? What are we really trying to do here? And I think that for me, when I think about, because I'm an OPS girly and when I think about how I solve problems, like a problem is presented in my head, I already have a preset bag of tools that I can pull from to solve that problem because I solve the same similar problems often. Right. But when it comes to AI as a tool it's we don't have that muscle yet. And so I think the best way to build that muscle is to think about whatever the outcome is, not what a feature can do, what are you trying to solve and then think about what the features are and how they might play into a solution. Because most, and tell me if you guys agree, like most of these solutions aren't solely AI driven. Like there's other parts to them that aren't AI. Like automating the email send for example is not AI. So I don't know, I just think that we have to train our brains a little bit differently to think differently, to solve problems differently. You guys agree?

[0:28:30]

Yeah, totally. I think it's when I think about the benefits of AI, the outcomes you can achieve, a lot of it is like let's get the boring stuff off of my plate so I can spend more time focusing on like higher level strategic projects. And if you're just looking at like hours saved or you know, the an activity output you might be missing the like opportunities you're gaining with what you're able to do with that time saved and the additional value that you're able to create there.

 

[0:29:11]

We did just get a question in the chat that's interesting which is how do you handle data security? They're working on AI governance now but the big fear is around ccpa, gdpr, non compliance. I don't know. My first, my first gut to that was you just gotta be careful what data you have inside of your CRM and how you want to use it. Right. Like we have clients all the time that are migrating or that are integrating huge systems. And a lot of times I'll be honest, like it's the marketers that are the culprit of like no, I want all of it. I'm gonna bring it all in and calm down because that's where risk happens. Do you really need it all? So I would say that would be a big piece of it for me. And then making sure that you know you're following the right protocols for getting the opt ins and things of that nature. Any thoughts there Gunner Connor?

 

[0:29:57]

The good news is that if you're using a tool like HubSpot, like they are already thinking about this stuff. They already have processes in place to make sure that you're able to delete customer data. If you mark a property as sensitive then it's not going to be exposed. So the tools you're using within HubSpot you don't have to worry too much about that when you're thinking about integrating with other systems. I would just be careful about what data you're exposing and how you're using it, but there's also, there are tools that are kind of closed systems that, you know, if you buy the, the business version, it's like they don't, they don't use that data. In terms of training, I think for

 

[0:30:51]

the, the sensitive data properties, it's not only that they're not visible or exposed, but I don't think you can trigger anything. You can't action them. Like, you can't, I don't think.

 

[0:30:58]

Right.

 

[0:30:59]

Yeah. So it's not just about the visibility, but also you can't, you know, trigger a workflow or an email to go out based off of some sensitive data point. All right, do you guys want to go through some real. Builds solutions that drive real outcomes?

[0:31:14]

Yes.

 

[0:31:16]

All right, first outcome we want to solve for is a marketing to sales handoff. We love this. Every company has a marketing to sales handoff. And there's a lot of little things along the way, obstacles and things that can make that handoff bumpy or not very successful. So Gun has built a pretty cool use case here. I will jump right into it. Gone. You want to set stage?

 

[0:31:38]

Yeah. So this is not a single use case, but it's actually several. I think it's 4 or 5 in the marketing to sales handoff process that all leverage AI in different ways. So really trying to, you know, figure out how to, how do we make this not just a, you know, a tool that we use somewhere in the process, but kind of a culture of adopting AI and using it to inform strategy and the, you know, create a more holistic system. So starting out, we're going to be looking at attribution. This is kind of the end of the, you know, traditional marketing journey. Right. When it's being passed to sales. Attribution data is really valuable to sales reps in terms of them understanding what, where the lead came from, what they've interacted with. The big problem that I see is that attribution data is really messy and you store that data in a ton of different places. You've got the original source properties, there's drill downs for the original source properties, UTM properties. You may have like a how did you hear about us? Question on your forms. That's storing attribution data and that's not even getting into if you're doing offline events or if you have partnerships. So all of that stuff is stored in different places in your CRM. It's not very easy to just look at all of that data and understand the larger context of like, what's actually important here. Luckily, AI is really good at synthesizing a bunch of data and kind of standardizing it and giving you a single output. And this is a theme you're going to see come up in several of these marketing to sales, handoff use cases. But what we're doing here, this is a smart property and we're basically feeding in a bunch of those properties that are storing attribution data. We've got the original source properties, we've got self attribution data, we've got UTMs, and then we are giving it some guidance on how we like that information to be weighted. So there are different ways you can do that. You could say like certain types of interactions, like attending an event should be weighted more than a website visit. You can wait by recency when the property was, was most recently updated. You can wait by the number of interactions. So if they have, you know, 50 website visits from the same UTM source, you can weight that higher than if you, you know, just have like a single visit. So the, the smart property here is going to synthesize all of that information and it's just going to output what it believes is the most, the most important, the most likely candidate for a source. And the way people are doing this today is usually they're looking at first touch or last touch. And as I'm sure both of you have seen, like, that's not necessarily the most important touch or even unimportant touch. It can tell you some information. But a lot of attribution is a, it's a correlation, not a causation. So if you're talking about, you know, the last touch, that's, that's not necessarily the reason that they converted to a deal. It's, it's just the last thing that happened. So this is hopefully giving you a more informed and more like context aware source to draw from for your attribution reporting.

 

[0:35:43]

It's also what large language models excel at. You know what I mean? Like this is what they're best for.

 

[0:35:51]

All right, so this next one, ICP routing, this is, you know, lead routing is a complete nightmare, especially as you, your company gets bigger, your sales team gets bigger, more complex. You know, if the, the way a lot of companies are doing this is just like if the state property equals X, send it to this rep. And what that misses is a lot of the, the nuance it ignores, like the, the specific skills of a rep or you know, there's a lot of ways that you can kind of strategically align the deals you're Sending to a rep, to that rep and their skill set and their, their work history, the industries that they've got experience in. So what you can do is this is a very similar type of prompt to what we just did with Attribution, but we're using it in a workflow action here instead of a smart property. And you're looking at the, you know, a lead that needs to be routed to sales. Taking a bunch of information about the, the contact who is associated with the deal, what their role is, the company size, industry, if they've, you know, gotten funding, you can do research on the company, you can gather a lot of information and then use this AI prompt to basically assign a tier. And the, the important thing here is you're able to treat differently tiered leads in different ways. So a Tier 1 lead, you may want to send them straight to an ae like they are a great fit based on everything we know we want to get in front of them as soon as possible. Let's like send them to an AE that, that is on the right team to help them out. Tier 2, you might want to send to an SDR, might need a little bit more qualification. Maybe we need to under, they're like on the cusp in terms of company size and we want to understand their budget. And then tier three could be something that you're like, I don't want to waste the sales team's time but just in case let's like push them into a self serve or a PLG motion that allows you to still potentially capture a deal but you're not spending valuable human resources on working that deal. So what the end result of this is hopefully it results in a lot less manual reassignment. You see that a lot when it comes to automated lead routing. A deal will be assigned to someone. The CRO sees it and says no, that's not a good fit. Let's move it over here. This should hopefully reduce a lot of that. And then your, your reps themselves should feel like the leads that they're getting are better fits and they're better equipped to serve them. So hopefully you're increasing close rates there.

 

[0:39:00]

I just had an idea too. What if you could do it based off of skill set or like, you know, I have a ton of experience in the insurance industry. Yeah, I know you can do I think for ticketing now. But anyways, what's next?

 

[0:39:14]

All right, so you know we've, we've done the lead routing, we understand the attribution but we need to actually tell a Sales rep, what's going on with this lead. And if you're just doing the lead routing and assigning it to them, and then they have to go to the contact record and look through the activity, there's so much data there, they, they're not going to know like, what's important and what's not. So we've got several AI actions here. We talked before about the idea of breaking stuff up into, into different actions so you can get better outputs. So we're using one for research, we're using one for, to analyze properties and then we're using one to actually write a brief that is going to be sent to sales. So it's, you're compiling a ton of information. You're looking at like recent marketing interactions, what, what types of content, what like topic clusters they've engaged with, if they've, you know, attended events, if they've previously met with anybody on your team, all of that stuff along with company information, ICP fit, you know, research that we're doing there on the company and then you can summarize it into a brief that you are sending to the, the sales rep when that lead is assigned. And the, the key point here that I think creates so much value is structuring the data into a narrative. If you just provide like a data dump where you're saying, like, these are all the properties that are important. You know, make sure you read all of these. These are all the website pages they visited. You've definitely received a notification like that before and like, your eyes glaze over and it doesn't sink in. But if you're able to structure that data into a story, into a narrative about, you know, the customer journey, it's much more likely that the sales rep is going to read it, engage with it, absorb that information, and then bring that context into a sales call so they can have a more informed conversation. Instead of coming into the sales call and being like, hey, great to meet you, what are you looking for? They can say, oh, I saw your team downloaded this ebook and attended this webinar on this topic. I'm really curious how that ties into your current initiatives. And overall the conversation quality will just be dramatically improved over them. Coming in cold.

 

[0:41:53]

We got a couple more and then we'll move into Connors.

 

[0:41:56]

Yeah, we're getting into some, some of the more advanced, really cool stuff. This is an assistant, a custom assistant. So very similar to like a custom GPT or a custom gem if you've used those. But this is trained on your CRM data So it's much more valuable. This is something that I created for AI powered content recommendations. It's a tool that sales can use during the sales process and it's going to look at, you know, where they are in the sales process, what content they've engaged with before and the other information or you know, industry company size. And you can, it will recommend marketing content for them. So you could send, it could suggest a case study that is relevant to the industry and the products that they're interested in. So this is again another way to just create, you know, better conversations, better follow ups from the sales reps. Love it. All right, last one. This is kind of closing the loop because marketing and sales is not a one way street and a lot of times after a deal closes, whether it's one or loss, marketing doesn't hear a lot about what happened. So this is an agent that goes in and analyzes closed loss deals and it can help inform both marketing and sales in terms of creating new content, adjusting messaging, how we are doing the tiering and the routing process that we talked about in the workflow before. If we see like deals in this specific industry that go to this rep are being lost more often than we would expect, maybe we route those differently.

 

[0:43:48]

I'm blown away by that one about the single threaded deals because how would you know that that's an issue, right? Like I'm trying to figure out how would I have pulled that out of like oh well, all you have to do is look at how many contacts are associated to the deal to be able to infer whether or not it was associated with. Right?

 

[0:44:01]

Like yeah, but that's so much data analysis that a human being is going to have to go in and like export a bunch of stuff or build reports. And this is, you're just saying like I need to know what's going on with my close loss deals. Like how can it help inform what I'm doing in these areas. So this is, it really is going to turn the whole go to market team hopefully into a more of a well oiled machine that has a feedback loop so that it's constantly improving itself.

 

[0:44:38]

Efficiency gains, knowledge gains. All right, let's pivot into the demo that Connor worked on. And this is more around events, which you've teased us sufficiently on. Events are a huge channel for marketers, right. And they usually cost a lot of money whether you're sponsoring one or hosting one.

 

[0:44:56]

So.

 

[0:44:56]

And look, we also talk about marketing. A lot of things in marketing are a little bit of a bet, but if you're going to spend a ton of money towards like hosting or sponsoring an event, you want to make that bet, the odds of that bet as good as you possibly can. So looking around to see does it even make sense, are the right people going to it? So that's sort of what we're, we're solving for with this one, Connor. I'm going to hand it over for sure.

 

[0:45:20]

So this example is around something that we see lots of our customers do, which is basically doing ancillary events to core events that they're, they're attending. So sort of with, with happily, our sort of thesis is if you're going to events and you are having your sales team work a booth, use our lead capture app so you can sort of capture all those anonymous people. But the thing that we see increasingly often, I would recommend everybody, if you have an event strategy do, is do your own micro events around the event, right? So that's like the happy hours, the dinners, customer dinners, executive dinners. And one of the things we really focus on is how can we enable our customers to be able to only be able to extract value from the core event, but how can we enable them to execute all the stuff that happens around those ancillary events which for anybody who's done those, knows it's a lot of work. But one of the biggest and hardest parts of it is we're okay, we're going to this big conference, we know it's going to happen. This is an example for Cyber Security and RSAC. There's a big conference in 2026. And so we've decided we're going to host a private dinner during conference week. But we don't just want to email everybody that's going to be at the event because we kind of don't want everyone to show up. And we need to figure out who we actually want at our executive dinner because we need to curate the room, we need to facilitate conversations, we need to make sure it's really good experience. So the way that we start is that we come and build a data set using HubSpot and it allows us to basically take all those companies that we have analyst somewhere that we know are going to sort of have people in and around this event. And what we want to be able to do is figure out which one should we be trying to get to our dinner. And so here's where AI sort of starts to help us to figure out how to put this together. So step one that we're going to go do is we're Going to build a smart column into this data set. So we've written this whole prompt to say, I want you to go analyze all these companies in this list. I want you to figure out whether or not they have buying signals, whether they have budget, where they have momentum, whether they have urgency. And we want to create this smart column to sort of understand, is this company attending this event? We're going to look for their website, we're going to look for booth numbers. We're going to automatically, automatically go and pull all the available public data to find out whether or not people at these companies are going to be at the events. From social posts, from web posts. Often companies have a list of their events on their site. Smart columns let us extract all that information, and we can kick this back in with a big prompt. And all we really want to know is, but are they going to be there, though? And AI is going to figure all of that out for us and tell us instead of giving that as a research project to somebody. So step one is, big list of companies, are they going to be at this event or not? Once we know that, we're then going to go and we need to figure out, are these people, we should be prioritizing. And so that's where we can introduce conditional columns in this data set for us to then determine with ifs and thens and otherwises, and sort of building a logical tree here to tell us if, if this is somebody we care about. So for this particular targeting, we want to know, do they have funding, Are they attending, are they attending? But maybe we don't have an indicator of whether they're funding or not. And then we maybe have this third criteria, which is like, there is no funding. We have no indication that they're attending. Like, maybe not somebody. We should be allocating a lot of time here. And so this allows us to sort of take this and say, instead of just dumping a list to our sales team and say, go and invite all the people at these companies to our event. Figure out who we care about, why we care about them, and if we should be prioritizing, getting them to actually come to the event that we're doing. And this is something that allows us to run this inclusive of all the data we already have. Right. So take everything Gun talked about already. You can extend that into this. Maybe you want to prioritize people that you've had some sort of sales conversation with in the last year or so. Maybe you want to prioritize people you met at other events. You're using Happily you'd have that data, it'd be super useful for you. But you can start to pull together a bigger picture of who you want to focus on and what you want to do with it. But the AI layer doesn't just start or stop at does start, it doesn't stop at the analysis. We actually want to extend this into what do we do with these insights. We now want to make them actionable. So what we're going to do is we're going to take that data set, we're going to push it into a workflow and we are going to use those tiers that we established. So we're going to say if this is the company in our high intent tier, we want to invite them, we want to work them, we're going to go and have the custom data data agent analyze all the information about their organization and it's going to create customized communications to all of those people that we want them at our event that is tailored to them, tailored to their company need, tailored to our event topic and all ready to go so that we can make sure that they are showing up and excited about the events that we are participating in. But we don't auto send this. This is the other big piece. We're not just trying, we go back to sort of the example we gave before. We don't want to just create content, blast it out to people and hope something comes on to the other side. What we want to do is enable our human sales reps, who have a much better lens on this, to be able to get their faster. So what we're doing here is creating a task, giving it to that sales rep and saying we've identified this as a tier one account, we've drafted communications for you, look at the additional context, tailor it to however you might want to and let's make sure we do this. And this is how you actually are enabling your team with AI Instead of looking at this as just we're going to eliminate the human and authenticity element and just replace it with mass blasting robots. It's a great way to burn all the people that you want to go and connect with that you identified as your tier ones. But we have this whole other audience of people that maybe are in our tier ones. There's a lot of them. We only have so many sales hours in the day. And so what we're going to do here is we're going to set up a prospecting agent. So we're going to go and build a prospecting agent inside of HubSpot. And we're going to use that workflow to add people to that prospecting agents for those. So for those medium accounts where we don't really have the energy and the power to justify going into all of those, we want to be able to enroll them and prospect into them. And so when we build a selling profile using, we're able to configure exactly how that's going to work. What is it inviting people to, what is it promoting, how is it working on it, how are we going to get them to be able to get engaged even if they didn't qualify for coming to our particular dinner. And prospecting agent is really powerful and you can tailor it in so many ways with those selling profiles to make sure that it's talking about this the way that you want it to as you go. And so we're going to automate the lower priority ones that we don't want to spend time with. We're going to do handcrafted, tailored approaches for all of our high priority ones. And we're going to execute this whole campaign in a way that operationally as either the rev ops admin or the marketer, you can set all this up, you can go and enable your team, you can go and leverage your salespeople to start working on things. And this isn't some giant project where you have analysis, you have to put all the stuff together, you have to have somebody go pull all the data. Like I remember, not even I remember doing this 10 years ago, but I remember five years ago when we would want to go do this. And this was like a multi week, multi person project. Try to go and execute. And then you have to meet with the sales team and be like, guys, like please go and invite these people. And the sales just sends an email that's like M Dinner week Vegas. Question mark. Oh my God, no. And so the ability to leverage this to scale up the impact of your team and I think this is one of the elements when people think about, you know, all this stuff is going to get replaced. There is not a finite amount of work to be done. And so what we want to do is use AI to make the work that matters more important and to automate the work that is less valuable and less important and ultimately get more done with the team that we have.

 

[0:52:47]

I'm with you on that. The human pieces are, you know, some of those are not going to go away. And that's the hyper personalization of your comms. You just blast stuff is never going to work. Used to though, gone Are the days. Those were awesome demos, y'. All. I really appreciated you sharing them. I talked a lot about scaling it. I think it starts with an idea that solves for a singular use case that then can expand out into others that impact teams. And honestly, like, I was just thinking when you were showing like the, the attribution one, for example, that doesn't. If I'm the marketing leader and I work with the sales leader, which I'm not even that much need to do that. But like, if I own Attribution, I could go build this and then my whole marketing team can go see it. Like it's in their worlds you can go look at. Right. Like, and they didn't have to do anything, they didn't have to build anything. All they know is that our sources are getting stamped right every time. Right. So it's not even having to do it yourself, but building something for someone else that doesn't impact their, you know, day to day clicks and everything. But actually, you know, their reports go up and to the right more. It can help with scaling for sure.

 

[0:53:54]

One, one thing that I didn't mention in my demos, but I think is really valuable is the, if you have an AI meeting note taker like Ask Elephant or Fathom, they sync a lot of good information to the HubSpot contact and deal records, usually in notes. And that stuff is a gold mine in terms of like creating a smart property or a workflow action that looks at that information and pulls out valuable things. That's something I've, I've started doing a lot recently. And like a lot of good attribution data can happen on sales calls. If you just train your sales team to ask the question, you know, how did you hear about us? You know, what, what led you to jumping on this call?

 

[0:54:41]

Yeah, I would also say just no one asked the question, but how long does this stuff take to build? I worked on some of these on my own, I've built some of them on my own. And you guys can weigh in here, but some of these don't take very long at all. Like the attribution 1, 30 minutes if you know your model.

 

[0:54:57]

Right.

 

[0:54:58]

Yeah.

 

[0:54:59]

That's just one piece of it.

 

[0:55:00]

Right.

 

[0:55:00]

I know there's like five sections of it, but this isn't stuff that takes months and months to build at all. Or we could absolutely share some of the prompts that we used for the sales summaries.

 

[0:55:10]

Feel like a killer downloadable asset that you guys could do on a bunch of stuff is like prompt library stuff.

 

[0:55:17]

Yeah, good Call out prompting is hard. That's like the new skill set everybody has to learn.

 

[0:55:24]

No, you just go. You go to the AI tool of your choice, you whisper flow to it, and then you're just like, write a prompt for yourself. That's good. And then you open a new chat and you put that there.

 

[0:55:35]

That is. That is the. The pro tip of the year is like, use AI to refine your prompts for AI.

 

[0:55:43]

I'm gonna get killed by our AI overlords because I. I have mainly moved almost everything from a personal standpoint to Gemini. And then we. We still do a lot of Claude for dev, but the only time I still use chat GPT because I should swap it on my phone, but I just, like, haven't is when I'm cooking. But all my chat GPT prompts are just like chicken parmesan, no shallots. I just love. Did it come back and just shout ingredients at me. But it does a pretty good job. It gives me a recipe and then I'm like, but lemons? No, no lemons. We don't have any. And it'll do a great job.

 

[0:56:18]

Oh, speaking of recipes, I got a shameless plug we built. We're building a. It drops in a couple of weeks. We're building a full cookbook of recipes of AI recipes that will. They're all outcome based. So marketing to sales, handoff, service handoff, kickoff, onboarding. We've got lots of tasty recipes that will be dropping very soon that you can all go follow. Cool.

 

[0:56:38]

Love that.

 

[0:56:40]

I'll end on a shameless plug note. Awesome. All right, well, Connor Gunn, thank you both so much for coming today. Love gabbing it up with you guys, as always. And better than gabbing is watching what you guys can build and how you guys are growing your organization. So thank you for joining me. Join us next time, y'. All.