Julia Nimchinski, CEO HardSkill Exchange
Our community was really excited about this one. AgenticGTM.
Welcome, Alina, cofounder and co CEO of Chili Piper. Our pleasure,and, yeah, an all-star panel here.
Welcome to the show. How about we just start with a quickquick round of introduction, everyone?
Alina Vandeberghe, CEO Chili Piper
Thank you, Julia, for having me. I'm very excited to have Liza and Matt and Kevin, Jonathan, and Kelly.
We, gonna talk about use cases that are working, that we've implemented internally, that have produced results, and I'm very excited about that. What I would love to do as part of the ins introductions as well with everyone, if that's okay, is, just name and what the company does in, short, inone short sentence.
And I would also be super curious what everybody understands by, AI agents, how you look at it and, like, just this short definition. Because I feel that different people understand different things, and we can start with that.
So I'm Alina Vandenberg.
I'm cofounder I'm the technical cofounder at Chili Piper. I love building and breaking things.
That's where my most joy comes. And, as soon as I saw that we can do so many cool with AI, I got everybody in the company excited as well.
I made it more of a game. Everybody can volunteer and buildAI agents on our GTM flow.
And, there are all sorts of rewards for it. Our company is all focused on demand, demand gen conversion, how to optimize the buying funnel for better conversion rates.
And for me, personally, an an agentic, flow is different than a regular workflow that you'll create with Zapier or something like that.Well, now with Zapier, you can also create agentic workflows.
It's the ability to, make decisions, in the in the processand, learn in order to get to a goal. So the there there's more to the automation because there's also decision making in the process that's,independent from a human.
I would love maybe, Jonathan, you want to go next? Because that's who I see next on my screen.
Jonathan Kvarfordt, Head of Growth, Momentum
Sure. Alina, big fan, and thank you, Julia, for letting me be a part of this week.
Really exciting. So my name is Jonathan Kvarfordt.
Some people call me coach. I am the CEO of GTM AI Academy,which is all about on demand and live, education for how to use AI.
And I'm also the head of GTM growth for Momentum, which is what we call enterprise listing platform, as I'm sure we'll get into that herein a second. I would actually agree with you as far as your, your definition ofagents because I know I'd in the email I sent you, like, some people say agentsis like a chatbot.
I lean more towards the actual agency where you have this,technology that can make decisions for itself and execute. So I I agree with you on that one.
Kelly Hopping, CMO, Demandbase
Hi. I'm Kelly Hopping.
I'm the CMO at Demandbase, and certainly agents are a big part of that. So Demandbase is account based go to market platform, which is all, you know, data and insights.
It's AI powered to drive the right recommendations on audience and message and and next steps and things as you're going after target accounts. I think the agent space on ours, we've been talking so much about this lately in terms of what is really the value prop.
We went through a bunch of messaging pillars about agents yesterday, and we we said, is it really, like, what's the hook? Is it about being faster? Is it about being better? Is it about being, getting too strategic faster, more efficiently? Is it about, like, just the mundane going away and getting sort of burdensome tedious tasks out of the way. And I think what we kinda came down to is it's really about unlock for us, it's about unlocking the value of of the platform.
So for example, account base can be super complex, to put into place, to get adoption, to get the whole team running on it. And if you can get to value faster, meaning adoption is faster, your account list are setup faster, your customer journeys are set up, your campaigns are done faster,all of that because they're all powered by, these AI powered insights, then you can actually benefit and drive revenue faster because you've, you've unlocked more value in the platform.
So to me, that's where agents really come in is about how dowe take the things that usually get in the way of productivity and getting those automated so that we can focus on on the really strategic bets we wanna make.
Matt Darrow, CEO, Vivun
Great to be part of this panel today. I'm Matt Darrow.
I'm the cofounder and CEO of Vivun. We've raised over $130M fromAccel, Menlo, and Salesforce Ventures, and our customers are enterprises, that you guys know in the audience, ADP, Snowflake, Dayforce, DocuSign.
We've built the world's first AI sales engineer to give 24/7coverage to your entire go to market team so they can move a whole lot faster on their own. So my definition definition of an agent is actually pretty simple.
An agent can reason like an expert to actually complete work without being prompted. That's how we see it.
That's what we're building toward as well. And, we've deployed all sorts of agentic use cases, at Vivun.
And I'm really excited to share some of those learningstoday with, the rest of the group in, real tangible sessions about what worked and what didn't, and hopefully people can learn from it.
Kevin White, Head of GTM Strategy, Common Room
I lead the go to market strategy at Common Room. Common Room is a is a platform for, for go to market, efficiency or go to market pipeline creation.
Essentially, we capture a bunch of buying signals, tell you the person in the account behind the signal, and then enable action on top of that. And you can see how it kind of applies to AI as capturing all this dataas the foundation for AI and agentic things to work on top of all that.
And then as far as the definition of agentic, I feel like the jury's still out. I really like Matt's simplistic definition or simplistic in a good way.
And, we're workshopping this this, analogy at Common Room where, we think of an agent or, AI as, like, a chatbot where it's like the the analogy is how you would tell a teenager to clean a room versus a housekeeper.In a in a teenager, you say, you know, clean the fridge, vacuum this, like, doall these things as, like, one by one.
Whereas a you you tell a housekeeper, like, oh, just clean the house and it it's it's there and it's done. So it's like agentic is a form of and now the analogy is, like, it's doing a job for you and reasoning.
So I think everyone's kind of, like, had this similar response. So yeah.
Alina Vandeberghe
And last but not least, the queen of, using agents for our strategy, Liza, I want to go to next.
Liza Adams, AI Advisor & Fractional CMO, Growth Path Partners
Alina, you're so kind. I'm so humbled.
So Liza Adams. I'm an AI and executive advisor, and also fractional CMO with Growth Path Partners, and I inspire go to market teams with what's possible with AI and help them with their transformation.
And with regard to the definition of an AI agent, in its most simplistic form, I believe that it is AI that does things on our behalf autonomously. But the purest, would say that it does things on our behalf autonomously in five different layers.
So it sets goals, it plans, it executes, it learns, and ita nalyzes. So if it does all those things, then that's a pure agent.
However, in today's environment, we can find different companies and different groups that do a subset of that. So they might primarily be doing the execution part or execution and a little bit of planning, and that's still called an agent.
So I believe that the agent term now has a spectrum of capabilities along those five things that I mentioned.
Alina Vandeberghe
So my hope is that at the end of the session, we all come up with new ideas to implement in our GTM funnel and things that are working and things that we've tried and they we already flopped on them. So it's not worthtrying again, at least not in the next three months.
So on practical use cases, for me, the biggest difference has made and last year, we only had two marketers that booked 1,400 sales-qualified opportunities for our team, which is a mind blowing number for coming from just two people. And the reason why we were able to do that is because weput all our data into Snowflake.
And so all our CRM data or our marketing data, and we were able to query and personalize a lot of our, funnel. And I'm gonna give example of two workflows.
I have, like, the entire Notion doc with what we've built,but I'm just gonna give exam two examples, and, I'm I'm going to go around and get two examples from each one of you. The first one was around account scoring.
So for us, being able to go to someone with our website and see what their buying funnel looks like, what the who they're selling to,whether they have a contact us form, and all the details that show us whether there is leakage in the funnel would allow us to create an account score thatwas, that allowed us to focus all our efforts to specific ones that we knew that they're gonna convert fast. And then the second piece was, providing anSDR with all the detailed information to reach out.
They would say they would take screenshots of the form. They would say what what the message would look like, what the chat would look like.
So the SDR, would they be able to craft super personalized,messages based on what the agent or had already discovered that was broken on their on their, buying journey? So that was also a big amplification for forSDRs. Alright.
I'm going to go next. Maybe, Jonathan, you, since we go in the same order, you'd give us your two examples also.
Jonathan Kvarfordt
Sure. So the first one is with, with Momentum.
So we have an agentic workflow where, when we have identified the biggest criteria that salespeople are supposed to meet at acertain point of the sales stage, AI has the ability to analyze the conversations real time so we can get all the information from the from theconversations, analyze the the exit criteria as you go through stage by stage.And then once the things are already met, all the CRM data and all the exitcriteria for whatever stage, it will automatically move the stage as it goes.
So it's going through the processes the salesperson goes through. And then for those people who, you know, don't trust the AI to do whatit needs to do, there's a human in a loop factor where it will send a messageto the opportunity owner saying, hey.
You've met all the opportunities. Do you wanna move thisf orward? Press this button.
You know, just click that button in Slack and move the move the deal forward. So really nice to be able to let the salesperson kind offocus.
The other one's kind of similar to what you said, Alina, and that's using Emplr with, various amounts of tools. Emplr.
ai, where we can use things like we have, like, data coming into Slack, and then we have data from Clay and other enrichment sources. And then this agent understands when someone comes in from, our B2B, for example.
We get that Slack channel. We can understand what's going on.
They can analyze when it's visited the website from our B2B, and then analyze that with, Clay Data and any of the enriched thirdparty stuff, and then use that information to then go through a a sequence of,creating the messaging, reaching out on LinkedIn, doing the whole process usingall the data points they they require. And then the next level of that is theycan analyze the person, title, and industry or company and then look out andsay, okay.
What other companies are like this and are other people who are like this to then cross over to, similar look alike type companies, whichis kinda cool. So those are my two.
Alina Vandeberghe
Alright. So it's also similar, at the top of the funnel also for the account scoring and the optimization of the of the close rates.
Kelly, you want to go next?
Kelly Hopping
Sure. Yeah.
Ours, probably similar. I think, Clay is a big one for us,for our SDR.
So similar to what Jonathan talked about, just supplementing and augmenting the data gaps that we have that not only help our SDRs be more effective and get them to the right data that they need to go, after the accounts, but also, like, what do we kinda say? Builds our own data. So itupdates within our own contacts and updates within our own sort of data hygiene to make sure that we're kinda cleaning the data in the same at the same time.
So both helps enable them, but also make sure that the next time it sort of cycles through. So that's been really helpful for us because our sellers have been able to to get to, contact information faster, which is always sort of the goal.
So that's a big one. Another one that we have is, I don't know if I'd call it I'm I'm interested to see, by the way, and how the the industry moves on agents versus, say, AI powered workflows.
Are they different? Are they the same thing? Is everything just gonna kinda get rebranded as an agent when that's just because it's cool and sexy, but at the end of the day, it's just AI driving a workflow? So we useJasper, on the content team, but we've actually ended up using it across theorganization. We run about 50 different, workflows that come sort of prepackaged with Jasper.
But I we use it for everything from taking, you know, a single thought leadership. We don't use it for original content creation as much because we sort of like owning the voice and things.
But we use it to take a piece of content like a an article,and we can use Jasper. It does these workflows that automatically converts itto a LinkedIn post or automatically converts it to a blog or to a social postor a quote or anything else it's able to extract.
And it's and the the brilliance of it that's different than,say, a chat GPT is that it has all of our, demand based, data in it. So it's got our ICP definition.
It's got all of our buying group messaging. It's got all of our, how you know, our our buying guides for customers.
It's all the things that you need that are proprietary. So it's probably got a hundred documents that are all about our own messaging, ourown positioning, our own tone of voice, our own brand standards, whatever itis.
And they have trained the answers back. And so everything is well informed on our own products and our own messaging.
It's able to to populate all of our content. So that's beena big one.
And then, of course, we're rolling out, a whole bunch of agents from demand base, just to help you get up and running on demandbase faster. So, that's just a few.
Alina Vandeberghe
I'm really happy to hear that.
Jasper does some some amazing things with AI, on their ownGTM as well, and they use Sleepypacker to optimize all the ways in which they get prospects to, buy faster. And I'm so happy that you're using them.
I would be super curious, and I'm sorry to put you on the spot so we can, get that later. I'd be super curious if you can have an example of what you have created with Jasper, like, a post so that, I'm Icurious on the structure and, like, the format of it.
Kelly Hopping
I'm happy to find one for sure. The what what we did, which was kinda crazy, is that we loaded in like, I have an a podcast.
So we loaded in, like, every episode of the podcast, and it was used to create my voice from that, like, just the way that I talk and the way that I communicate. And then I overlay that too with my post.
So, sometimes I'll say, hey. Here's an outline of what I wanna talk about.
Use kelly. ai voice to actually write this thing and then create an article out of this.
And it's just fascinating. So, yeah, I'll pull I'll drop an example in if I can find one.
Alina Vandeberghe
Ok. Matt, I know that you've done a ton also.
Matt Darrow
Okay. Well, I could talk for, like, an hour on this.
I'm not gonna monopolize it. I'll give you guys three examples, Alina.
And then if you wanna dig deeper, go for it, and I'll stick to, like, where we focus and the outcome, and then we could talk talk about,like, the tech behind it: demand gen, sales engineering, legal.
And each one of the three, like, on the demand gen side,like, we had a traditional sort of SDR structure that you guys might have come across in any other B2B company where there's an SDR leader, there's pods ofSDRs that are rolling up to the sales team, and we basically dismantled that whole function. And we're able to reduce that down to a handful of folks that work directly with marketing and a whole host of different systems that actually have allowed us to blow all those historical pipeline numbers out of the water with a very, very different composition of, people and technologies.
So it's, like, one thing we could go down. The second thing, sales engineering.
This isn't a Vivun commercial. Yes, we built an AI sales engineer. That's our innovation.
But the thing of, like, well, why you'd care about this is,like, our reps don't go to enablement tools to ramp anymore. They're always ramped because the AI sales engineer knows everything about our products, our competition, our customers, all the new use cases, so they can move a whole lot faster on their own.
That's just been like a crazy dynamic where you don't needthe same ratios anymore and your reps can be more independent. And the third thing is legal.
We do a host of different things with tools there where our turnaround time on red lines, like, our sales team has never experienced that before. And that actually, like, when it's crunch time at the end of a quarter,as much as we all know in B2B, how much you want some level of linearity that never happens.
It always hockey sticks at the end. And being able to quickly turn and move through contracts have given us a big edge where we can pull forward deals because of how quickly we can work through the legal side.
Those are things we've already deployed across the board,and we have a whole host of other things in recruiting and support that we're looking at too. But those are big areas that we've invested in some of the changes we've made and happy to talk about, how we make any of it work too when the time comes.
Alina Vandeberghe
I love the legal, use case. I use it a lot myself for legal,red lines.
And for enterprise, it makes a lot of sense that you're using that for the reps. For the first use case, I've always had a love hate relationship with the idea of removing SDRs from the flow because on one side,you're removing people and, that has a my, the mom in me feels bad whenever we're talking about that.
But on the positive side, the way I always, envision is that SDRs can get into roles that are fulfilling, like account executives, and and,it just means that they want to do low, level of work, and and it's more fulfilling. We can we can dig further into that, but I would like to go first to Kevin, and then we can go back to the tech stack and then all of that.
So, Kevin, you want to walk us through some It's a use case?
Kevin White
Yeah. Nice segue into into my section here too because, al lof our our use cases are more around, like, the the SDR and AE side of things.
And I think to your point of, you know, taking away the mundane tedious task, like, that's what we've been focusing on enabling ourSDRs with, with agentic workflows or whatever you wanna call it. The first thing we do is, take away all the the process of, process and the hard work of,doing an account research plan.
And so we kind of cut we were able to cut that time from anhour to, like, sixty minutes just using AI and and workflows, and we use Common Room for for that. And the other thing and I don't know if this is actually,like, agentic or AI.
Maybe it's just a powerful workflow, but it it's kind of slick how we do it where we have, account IP tracking on our website. But whensome account hits our website, we don't have the person, the the contact behindit.
And so what we'll do is prospects the five use AI to prospect the five most, relevant titles, and then, sync that, with data in ourSalesforce and then and assign an owner to it and all that kind of stuff and then send it out just an automated outbound email. And so that whole thing just happens in the background without any humans in the loop, and it's assigned to the right owner in the when when we get replies from it, it just lands the SDRAsinbox.
And so that whole workflow is is automated where, you know,in the past, like, a year ago, it would take, you know, a Slack alert,reviewing the account, all this kind of stuff, and and now it's all fully automated. So, pretty slick use case, I would say.
And come we use Common Room for pretty much the whole thing not to be that annoying guy. I'd be slinging my own back here.
Alina Vandeberghe
I'm happy that you're removing friction. Not making your buyers work for the harder to get to your tech.
Liza, you have a different kind of, you have different kind of examples than the rest of us because you were working, with with different clients. Do you want to give us, like, a couple of examples?
Liza Adams
Yeah. And I just saw Julia's question here about she's noticing that the panel has, noted, analytics for the sales process.
Right? And my my, use case actually is more on the righthand side of the bow tie. So rather than the acquisition phase, it's really more on the use, retention, upsell, cross sell because we know how hard it is to acquire, and we are dealing with, you know, less resources, less budget, and it's a lot cheaper to grow our existing base, than to get new customers.
So my use case is really more around upsell and cross sell where we we have a client, couple of clients that are really leaning in on that right side of the bow tie where we're looking at the usage of different features on a SaaS platform. And depending on the features that they use andthe amount of usage of those features, we try to determine which ones are, the best opportunities for an upsell, which ones are the best opportunities for cross sell.
So all that data is then fed to AI. AI then makes a determination which accounts are, better suit are best suited for an upsell cross sell or simply an awareness of our road map.
And then depending on whether an up they're an upsell or across sell opportunity for us, we then customize emails, and and an outreach essentially for them to, do that motion. We also use AI to do an in app message for those cross sell and upsell opportunities.
So I think it's slightly a different use case because we're dealing with clients that don't have a lot of resources to acquire new business, so they're trying their best to keep who they have, and improve market, wallet share with with those customers. And I think my second use case is a bit more broader than what I just said where we've actually transformed a really lean marketing team.
Absolutely wanting to do more with less. That pretty much is the mantra nowadays, more with less, but still achieve the goals, where we've transformed this team into a 45-member organization where 25 are humans and 20are AI teammates.
So the AI teammates were actually built and are being managed by the humans to do very specific tasks. So they are actually on a onan org chart, and these AI assistants are reporting to the humans that built them.
So, in product marketing, we have, a pitch deck creator in campaigns. We have a performance analyzer in content marketing.
We have a topic generator. So there's about 20 of those that are in support of the humans doing the mundane and repetitive tasks.
Alina Vandeberghe
I love the cross sell and upsell use case. For now, we're also doing a lot there because, our expansion use case is is big.
But I find that the account scoring modeling that we file for the top of the funnel is sufficient for us to do the automatic plays. So if we identify certain accounts that fit certain criteria, we put them into advertising.
We actually run advertising campaigns against our customer base also. I don't know if some of you are already customers, have already seen my campaigns maybe.
But we actually multithread several buyers into that funnelas well because our user is not our buyer. So it's very, important for us to do those multithreaded plays, not only sales like Julia is asking.
We run a lot of advertising campaigns, into the funnel. You wanted to say something, Matt?
Matt Darrow
Yeah. I wanted to maybe continue on that question too around, you know, why did we all focus on sales or at least this discussion forit.
And to to Liza's point, I I don't think that's the case.Like, even for us on the engineering side, like, if your engineers aren't usingCursor, right, they're being left behind.
That’s another great example of an AI driven tool that's gonna give them massive productivity and development. And I was talking to Julia about this yesterday.
I had, a simple matrix that talks about how quickly impact AI's like, how quickly AI is gonna impact your role and effectively maybe, youknow, put you out of a job as a human or teach you that you need to learn something different. And if it's just a simple grid between the complexity of your role and the level of live interaction that you have with somebody,that's, like, a really simple matrix that you can just think about.
So if you're really high complexity, really live interaction, you're probably using these agents to, like, augment and help you do things. But if you're low complexity, low live interaction, you can just sort of live up to what Gen AI is all about, which is disruptive labor, like getting rid of some of these roles.
I think that's why there's so much noise in the AI SDR space because it is a perfect zone for that sort of that bent of disruption. So, forus, you know, this might sound a little complicated, but when we made that change from the big SDR org to what we run now, you know, we'll have Warmly to help us with interesting intense signals that feeds into Common Room that helps us out with segmentation.
That Common Room hooks into HubSpot or Gong Engage that allows us to then go and do the really type of targeted outreach.
And then, ultimately, that helps us drive traffic to ourwebsite, which is Webflow, which we've had Chili Piper embedded in Webflow,that this whole consortium allows us to take what we were doing all this inbound and outbound for and just sort of automate the hell out of this wholes tuff. Now I think what's interesting and, you know, Alina and Kevin, we're I'm customers of your guys and happy with your guys' service, but I think what's interesting now, you look at where OpenAI is going with their sales agent and what 11x is trying to do, and there's people in that space are trying to then consolidate and wrap all this stuff up.
And we didn't feel like those tools is like the one stop shop to handle everything in that sense was really there yet, but I could definitely see, like, next renewal cycle, like, those things are gonna be more mature potentially when the time comes to.
Alina Vandeberghe
There's a for sure, there's a lot of, in in this space,there's a lot of disruption, and, there's a lot more, gray area between thetools and who what does what and who does it well. I think that, as it comes to the SaaS industry, OpenAI is going to disrupt many companies, especially for simple use cases, especially for SMBs, and that's a good thing.
I think that AI is here for intelligence, not yet for wisdom, but for intelligence definitely, and, it helps us do our jobs, better.Kevin, you wanted to add something?
Kevin White
Yeah. I just wanted to add something to Matt's point of, youknow, maybe the foundational models are maybe, like, entropic or or OpenAI won't get theirs where, you don't need that a platform or a tool like Common Room or Chili Piper or whatever.
But, I think one part that's missing from those things fully taking over SaaS, right now is just like they don't have the connectivity to the data warehouse—Snowflake--, or your CRM or all these different, like, datapoints.
And so being able to aggregate that and connect to all the different things, the extensibility issue is, you know, certainly something that is, like, required to get the get the right data in because, like, the the quality of the outputs is just as good as, like, the the data you feed into the model. So, like, I think that is a really key place of where where you can actually make a difference versus, you know, versus, like, an ineffective AISDR is not gonna work if it's just, you know, crawling LinkedIn for what's public when it's not, like, in using your internal first party proprietary data.
Kelly Hopping
Yeah. I think the data makes a huge difference.
Right? I mean, if you're not powering with the right, I would guess, if you're not unifying your data together and using it to power,these agents, then, you know, it's just I mean, it's like any part of AI.Right? It's garbage in, garbage out, and the automation doesn't help if it endsup being crap.
So I think that's interesting. I did was thinking about Matt's comment.
Do y'all think of these agents as being, it sounds like,cost savings and when when, they were talking about the the mix of sort of human and AI org chart and things like that. Like, in my head, yes, I'm sure there's cost savings that will happen.
But I don't I never I don't want to think of agents as replacing human jobs. Instead, I feel like I wanna use the humans to get the more strategic work out of you, and so that we can actually just move faster because we're able to spend more time doing, the good meaty strategic stuff.
And we're just sort of, you know, outsourcing or whatever to these agents, the more mundane. So I never think of it as cost savings.
I think I think of it as revenue acceleration. But I don't know what your thoughts are on that.
Alina Vandeberghe
I love that reframing. And I do imagine that at some point,if OpenAI gets access to the data, so it has access to your CRM and it has access to Snowflake or or anything like that, they could still do many of thethings that us in SaaS do.
But there is an extra layer of expertise because these agents can only produce as well as the wisdom behind the people that programand structure the agents. And when you've been in the industry for, like, ten years, it's hard to replicate that by the OpenAI team on what the best way to structure these workflows, but the best way to to get most out of it.
And that expertise and wisdom of humans is still we're still better at it, especially in certain, spheres. I don't know how things are going to evolve past the ten year mark, but I I would imagine that for the next five to ten years, that wisdom is going to remain for, for us to to to have an to to be able to get the most out of it.
I would also be curious now to get into some of the flops as well, some of the things that you've tried and you thought, oh my gosh. This is gonna work, and it completely failed.
And I remember one of you had interesting examples before. Ithink it was you, Jonathan, on the chat, flow or maybe it was you, Matt.
I don't remember.
Matt Darrow
Yeah. We're always experimenting with this stuff, as well.
And, also, just to double back on on Kelly's point, which II liked on the agent value prop case, like, I always found, it's gonna spike on one of three dimensions. It's either faster, cheaper, better.
Like and it's okay. Which one of those is, like, the agent really gonna highlight? Now we're always in search for these things.
So two two things came to mind. One was, like, we're always trying to figure out on the video and image creation side.
Like, can we off board a lot of the creative work that wedo? And, and we haven't really found anything that we love there yet because what we found is there's ways that you can do things internally faster. But when it comes to, like, the actual creative production asset that you wanna putout in the world, we just find ourselves, like, toiling in the last 10%.
Like, god, this would have just been a lot faster to do this end with to end with end to end with a human to begin with. And then like the secondary thing that we had tried was some of those a little bit more out therelike AI website agents and bots where a website visitor will just go and likesort of converse with this thing that's sitting on your website.
And, and and the problem with that is, they're just, like,wildly inaccurate. So whenever we had somebody, come in in that realm and then get sent to sales, the salesperson is coming in that conversation worse off than they would have otherwise because this sort of inaccurate bot whose brain was the basis of the LLM, which is the the wrong brain to put in an agent, is actually sort of skewing the initial conversation.
So those are two that didn't really, work for us in terms of trying to get, like, full production image video marketing assets. And then also thinking that, you know, somebody could, like, interact and get a good legup, for a sales conversation by interacting with, something, on a website.
Alina Vandeberghe
It depends a lot on the industry and the use case, and I can see, Matt, if in your case, it didn't work, there are certain elements that do perform really well, especially on B2B or B2C websites. It it can do really well.
On B2B, there's been a lot of hit and miss. We've, refrained back from releasing our own AI agent on chat because of that.
There's still a lot more RAG experimentation that needs tobe done before it's like, okay. I trust it.
We're not at that point yet, but I think that in the next three or six months, it's going to, get better. Any other flops?
Jonathan Kvarfordt
Yeah. So mine was, I was making an SEO agent team, so I hooked up relevance to make into SEMrush.
And was just working on experimenting to see what kind of content it would do. So I had I had this agent team where I had a manager agent and then a backlink agent and then a research agent and a keyword agent and a writing agent and then an editing agent, and they all work together.
And and it broke several times because it was just creating content that, number one, was not even close to remotely what I was trying to get to. I was just accessing the wrong data.
So it's just, you know, one of those things a few months ago, I was experimenting with all this. I was trying to get it to work that it's it's obviously really important to tinker into play, but also iterate and test like crazy because all my good successes have come from failing many, many, many times before I get to the right result.
But as an example, our flop, that is the output was, with the name of momentum, there's a lot of different momentums. There's, like, this momentum medical thing, and so it made this article about medical.
I was like, what what in the world what in the world is this? And so I just started loosening on the topic, which obviously wasn't good. So, yeah, that was mine.
Alina Vandeberghe
It was brave of you to try to, do your entire team of agents talking to each other. I I mean, eventually eventually, it's gonna work, but there's more tinkering for sure.
Kelly, some flops from you?
Kelly Hopping
I mean, other than you said creative, I'll say I don't know if you've ever done headshots with AI, but I look like a big boardroom Barbie or something. Like, it's so bad.
Like, it makes me have, like like, I look like I weigh,like, ninety seven pounds, and, they're horrible. So that's the only thing I will say is that that one is my biggest flop.
Outside of that, I think it's, it's more of the change management side, I would say, that we have the challenge on. It's not that theAI use case doesn't work.
It's sometimes is that the org isn't ready to use it. So for instance, we're when we added the clay piece to the SDRs, the behavior of getting them to look at the the full account data differently, or to know,like, way to get the full value of what the the account data was showing, Ithink, is a change muscle.
And so adoption of that agent running in the background, Ithink, has been the bigger challenge, versus the agent itself flopping. So that's kind of the big one.
Sometimes we have a little bit on sort of voice and tone we have to play with, that that gets us, there. But I think, we've seen a ton of success running it in, inside, like, our platform and things like that we've seen.
I think it's a third party integrations we're still trying to work through. But, yeah, stay away from.
Alina Vandeberghe
For sure, the buy in from the other teams to use, the agents to in their favor, it takes a while. They have to see the benefits, and you have to adopt them.
Kevin White
I was gonna say I tried one of the AISDR tools that I, won't name, but I'm a total believer in the technology. I just don't think it's quite there yet.
What I found is that, you know, it's, I had to prompt it to death so that it wouldn't, like, hallucinate or say, like, oh, I hope this email finds you well and stuff like that. And so, essentially, like, the amount of prompting that I had to put into it, it would come up with the same exact template every single time and, like like, two or three differentiated best messages that go out to the same exact person or the the same exact, like, prospects.
And so I was like, I should have probably just wrote this email myself or wrote the sequence myself and then, like, use use the, youknow, sequencer, like, outreach or something. So yeah.
And I I tried it out. I think I sent, like, 3,000 emails from it.
We're targeting, like, a smaller smaller, unrisky audience for us. It wasn't Square or ICP and just, like, didn't get any responses or anything like that.
So I think we'll it'll evolve and we'll get there, but, like, it just it that didn't work for me.
Alina Vandeberghe
It didn't work for me yet either. My best performing emails are still our templates that, I know that they're true and tested, and I still experiment on low score accounts.
And, it's just, yeah, still still more work to do.
Kelly Hopping
It's one of the challenges we have. We had considered whether we wanna build, like, an AI SDR capability into our product.
Our competitor has it, and we were, like, losing some dealson it. And then what ended up happening was because they had it and we didn't.
And then what would happen is they would go over and they would choose the competitor because of the availability of that feature. And then they boomerang back, not too long after because they're like, well, we chose for this thing, but that thing kinda stinks.
And we're like, yeah. I mean, so there's also, like, the ability, like, to just figure out, like, which of this AI is like, which of these are really gonna stick around? I read a stat the other day that said,like, 35 or 40% of AI companies won't be around in a year.
I'm sure a whole new batch will be doing something else, but of the current batch that are available. And I think it's because the, the concept is there.
I think the quality is where, is where they're gonna kinda make or break. And so just interesting to kind of follow that journey and seeing, like, how much you can believe in the marketing of AI versus the actual delivery of really solid, content.
Alina Vandeberghe
I feel the same pressure in the banks, cycle. Their competitors that have certain things with AI, and I know for a fact from their customers that it doesn't work.
But it's hard to convince someone that what I'm saying is true unless they actually experience it, and I have a lot of boomerangs in the same way. Liza, you want to give us some flops?
Liza Adams
Oh my gosh. I love this topic because I have many flops, but I won't tell you about my many flops.
You know, as they say, with success, we win, and with failures, we learn, and that's how I learn. Sorry.
I have many, many failures. But I'm gonna +100 what Kelly said about change management.
Right? Like, AI, I think the AI adoption is less about, technology, less about innovation. It's truly more about the humans.
The humans are the toughest part of this. And if we can't meet people where they are and be respectful and graceful about how we bring them along, that is going to backfire tremendously.
I think trust building is is so critically important. But more specifically on the flops, you know, my initial thought on AI, this wasabout a year ago, was, oh, this thing should be able to do math super well.
Right? So, heck, you know, it's AI. It should be able to do math.
But it is a large language model, and it's good in language,and it's not good at math. So, you know, for very simple math, it's it's okay.
But, you get to 30,000 rows of Excel, based on customer feedback at a huge event, and you try to analyze all that data. It will givey ou really good insights.
But for it to give you very precise calculations and graph sand stats that are accurate, it's not gonna go very well. So, that was a lesson learned, and I think we just need to really reset our expectations that if it'ssomething more nuanced and it's more around language, then I think AI is reallygood at it today.
With math, really complex calculations, big spreadsheets,it's not today, and the qualification is today because it continues to move forward in advance. So at some point, this will get super good.
Alina Vandeberghe
Yeah. It's progressing a lot faster, but on the simplest things around math, it's definitely challenging.
I'm gonna go to the last topic, which is what's a tech stack that's enabling you to have best results. We've covered some of the toolsa lready, but, I would love to dig a little bit further.
For me, the most valuable in my tech stack right now is Snowflake and Gemini. I have everything in Snowflake.
I don't need for need Salesforce anymore. I don't even need HubSpot anymore.
It's becoming my truth, and on we create most of our workflows with Gemini. So for instance, right now, I have all my customer data in Snowflake, and the queries that I do are around what messaging is winning,what's winning against competitors, looking for customer testimonials, lookingfor customer data.
Everything is in Snowflake, and I query it. And the kind of insights that everybody gets in the team, the product team, the design team,the It's, like, mind blowing to me that we get access so fast.
This would have taken us, like, I don't know, the the datateam months and months of research, and now we have that access super fast. Sofor me, these are the two tools that are most valuable in my tech stack rightnow.
Alright. We go next maybe to Jonathan.
Jonathan Kvarfordt
Oh, man. I got a humongous list, but, I wanna go back to Google, Liza, whose white paper talked about the three main components of agents.
Wanna make sure I call those out, which is you need an LLM,you need a software, you need some sort of orchestration layer. So for me, Imean, obviously, I use the main three are obviously OpenAI, Claude, and thenGemini for my, LLMs.
For orchestration, MomentumMax is like an orchestration layer, so obviously that's that's one. Emplr is another one.
I use Relevance, Make for Automation, and then tapping into things that perplex your other APIs. And I'm seeing more and more tools becoming more available with giving their APIs open so you can use things likeSEMrush and other tools.
And I can I use, I'm a our team's a customer of Chili Piperas well, so we use some of the, data bank and and things we get from ChiliPiper, which is awesome? But but I would also suggest if I can, it's really easy to get overwhelmed. I would say kinda like you, Alina, of just picking two or three core and pushing those as much as possible because it's really easy to get overwhelmed with all the technology out there and just kinda limit yourself to the the ones you get really familiar with and get powerful and then add them on carefully based on the outcomes you're trying to achieve.
Alina Vandeberghe
Oh my gosh. I completely agree.
It's overwhelming. Like, it feels like every minute there' s a new tool that does something, and every second, something else that does something else better, and it's like, oh, it's very debilitating.
Matt, you want to go next?
Matt Darrow
Yeah. I'll keep this one short because I shared with you guys, well, the tech stack on the demand gen case and the sales engineering case.
The one that I'll just put out there to Jonathan's point is,I mean, everybody in the company has access to enterprise, OpenAI. That that's just a given.
It was a big part of the culture change, like Liza andKelly, what you're saying about change management. It's really hard to drive change management.
Not everybody have access to the tools. And even if everybody has a very, very different use case, it just gets people in the motion that using this tool is gonna just be part and parcel for your job moving forward as well.
So that's, that's been critical for us for a variety of reasons.
Alina Vandeberghe
Seems like you're doing a lot. I think, we would all love more in-depth implementation from you, Matt, especially on the on the external support.
Jonathan also on the all all the things that you as soon as he nails the on the agents talking to each other. Kevin, you want to go next? Or maybe Jonathan.
Jonathan Kvarfordt
I’d say, it works now. It didn't then.
It works now. How about that?
Kevin White
Yeah. Maybe I can provide, like, a different take on techstacks.
I mean, I feel like there's just there's so much out there,and you can get over overwhelmed, on, like, connecting all this stuff and the nalso having to maintain and manage it all. And I see some of the the charts and flows that people are using.
I'm just like, wow. That is, like, way too complex for,like, a small early stage start up to to be putting all these these tools together.
We call it we have a term that we use called Frankenstackwhere, you know, we see that, and it's like, oh, you just do this with I thinkI think getting back to the the the dream actually is just to simplify it incredibly would be to just, like like you said, Alida, just dump everything into Snowflake or dump everything into a system of record and then just have the AI work on top of that. Like, maybe that's where we'll get in, like, five years.
This is like these are two these are there's a text back.It's like two different tools.
But, yeah, I mean, we use typical stuff, Common Room. We just we have our own product.
We use Salesforce. We use HubSpot.
For analytics, we use, like, metadata meta Metabase, stuff like that.
Alina Vandeberghe
We we've if we have all our Gong transcripts in there, it's so different, right, when you have this kind of intelligence and emails and call recordings for each one of accounts, especially kind of understanding which ones are converting better, why, what are the characteristics. You can do all sorts of inferences super fast.
It's, like, mind blowing to me. It feels like I have extra brains.
Can I also give a plug oh, sorry?
Kevin White
I should also give a plug for Chili Piper. I didn't mention it, but we we we use Chili Piper for for like, it's been a game changer for our SDR team, from managing inbound.
Alina Vandeberghe
I'm so happy to hear. I'm so happy to hear.
Kelly, you are saying that you were using Jasper. Is there anything else in your tech stack that you're, very happy with?
Kelly Hopping
I mean, certainly, we run DemandBase on DemandBas. We run it internal, but for our sellers and our marketers, which is, a big piece, because that's all AI powered.
We use Workato, which has some AI components into our revops capabilities. We also use Stencil on that side for marketing ops and the way that we, communicate emails out, integrate with with, Slack and all of those things as well.
We use, Momentum is one that I really like. I don't know if you guys use it with, oh, it says look at that moment.
DemandBase uses Momentum. Somebody just told me that.
Yes. We do, which is great for me because it basically sortof pulls all those Gong calls together, into a summary and says, hey.
Your sellers are doing great at crystalizing this valueprop, but they're really struggling with pricing. Are they really struggling to talk about data insights? Are they really you know? And so we it helps us kinda troubleshoot at a thematic level in a really quick way.
And I get that, Slack update all the time, and I love it because then I can kinda figure out where to point with the product marketing team to kinda help enable our sellers. So that's one that's worked really well for us.
Alina Vandeberghe
I'd be happy to hear. And what about you, Liza?
Liza Adams
Yeah. So on the tech stack side, it's not an easy answer because the answer for the right answer for one company is not the right answer for another.
Right? Because we have different goals. We have different existing tech stacks.
We have different strengths and weaknesses. So what I generally guide people to do is start with what you have.
And for the most part, what you have will most likely have some AI capabilities in it. And if it doesn't, it probably has it in the roadmap.
So whether it's your MarTech stack, sales stack, CS stack,or I mean, like, take a look at what you have because to Kelly's point, a huge part of this is change management. So if people are already familiar with, youknow, the technology, the the, likeliness that they will use it is a much, much higher.
So start with what you have. And then the second thing is have at least one or two foundational models, paid versions of it.
So whether it's Copilot, Chad GPT, Gemini, Claude, whatever you've got, because it will, allow you to see what is possible. And a lot of the innovation are happening in those foundational models.
And then thirdly, I guide people towards alright. You're looking at your existing tech stack.
You've got one or two foundational models. If those still don't address the need, then start looking at new technologies out there, new capabilities, startups perhaps.
But be very mindful about this 14,000 vendor tech, space that we're in. Right? There will be consolidation.
There will be mergers, and some of them will not be around.So we need to really be mindful of, you know, evaluating them to ensure that it aligns with our needs, evaluating the leadership team, the financial viability of that company so that we're doing business with, a company that will be around, longer term.
And then more personally, I shared on the chat, you know, mymy role largely is a strategy. It's a lot of, analytics, a lot of thought, a lot of thinking.
So my tech stack tends to lean on thought partners. So the large language models like Claude, Chat, GPT, Gemini, and I actually use them simultaneously, pit them against each other.
When one tells me one thing, I feed it into the other and get its its opinion. And it kinda feels like I've got a team of, 10 instead of just me and, you know, a couple people that I brainstorm with.
Alina Vandeberghe
I like that you're making them compete for, for your attention. We have, nine minutes.
Matt wants to add something, and I wanted to talk about just one last word on the type of use cases that you're excited to try next, some of the workflows that you're, that you're planning. Matt, you wanted to,
Matt Darrow
Yeah. Well, it was it was kind of actually a lead into that, Alina, because I didn't wanna end this discussion without going there, because, actually, we could use some help too from this panel’s, maybe expert point of view, is, there's the dark side of AI as well.
And one one one place that we felt that ourselves is in hiring, and recruiting. And and what'll happen is, like, you open a rec and arole, and there's so much of an inundation on AI applicants and resumes that it's becoming very, very challenging to sift through candidate pools and profiles because people are using AI to apply to jobs en masse and then effectively cook up skill sets that mimic your job descriptions.
And where we've seen this, even I wouldn't have expected this, even in certain interviews, folks are able to use AI to have a digital doppelganger representation of themselves, and they're not even on camera. They're they're behind the scenes talking through the digital doppelganger,behind the scenes talking through the digital doppelganger, researching questions as they go.
So this is one of our areas of, like, where we wanna go and use AI next is actually in this entire applicant screening candidate hiring pool, to actually start to cut through, I would say, some of these malicious cases that we're we're seeing in some of these cases. And I'm wondering if you guys also if there's preferred tools that you've already run across because we're just starting that investigation.
But I think on the hiring side, everybody's going through it, and, it's getting really, really messy really quickly.
Alina Vandeberghe
Yeah. For sure.
I was on some calls also where I could tell that the candidate was just using AI to respond my questions. And since then and thathappened maybe, like, eight months ago.
And since then, every time I get on a call with a candidate,I ask them to show me my their hands. I was like, hands showing.
I'm with my hands. You're with your hands.
Everybody's open. It's only our brains that we're gonna use because I want to hear what you are about, not what AI is about.
Yeah. I I'm curious what others, have done on this.
Jonathan Kvarfordt
I was just saying in the chat, this is a new tech called Spark Hire. It's awesome.
It is exactly what you're talking about as far as analysis and understanding both resume, tones of speech, all the analysis you need to have to help you hire the right people and have, assessments on either side. So it's not just conversation based, but also assessment based.
You can get a better idea of what they are and what they're capable of. So really, really good tech.
I'm not affiliated with them. I just like it.
Alina Vandeberghe
Alright. What other things are you all, are you all, curious and excited to try?
Kevin White
I'll go. I've been I've been tinkering with the, on the creative side, on the, like, page end videos.
Actually, he used I I use it I use it for promoting, these events that I join. It's kinda fun.
I haven't found, like, an actual core use case for it. It's more just like, wow.
This is cool tech that, like, it it it's like an avatar of yourself, essentially. And so I think there, you know, there might be some unlocks there.
I feel like the dream that I have it is also kind of a flop,but, like, the dream that I would love to get to is just, like, upload my brain or my content into, into some sort of, like, data repository and then say,like, oh, write a post or write, you know, write something about this and,like, it actually outputs something that sounds like it's in my voice. But haven't quite got it to that level yet, and, it's always, like, diminishing returns trying to, like, go in and edit with the the output of AI.
So I'm very excited for something that actually works like that, but, TBD if it's gonna happen or not.
Alina Vandeberghe
Yeah. My avatar is super weird.
It doesn't have my accent at all. I've tried a million things, and it doesn't make the mistakes that I make and the ums and ums.
It's, like, super weird. And the other thing is that I have a book and, like, a million posts that I've created, and I try to rag it and try to create it in my own voice.
And it just sounds like a robot. You know? It just doesn't sound like, the human I am yet yet.
Liza, you wanted to add something?
Liza Adams
I really wanna try, chat GPT operator. You know, we saw the demo of it, you know, it ordering pizza.
It, reserving, you know, restaurants and and buying ticket sat games. The reason I wanna try it is I think it will change the way we market because now we no longer just need to market to human beings, but we will need to market to the AI agents that serve the human beings.
So this is, so you kinda look at it as a mirror. Right? We, as go to market leaders, where our teams are using AI to create content, but that same content will be judged by people that are using AI and the AIs thatserve them.
So it's like AI on both sides. So I think there's a lot of implications to the go to market engine and how we will cater to both humans and AI, because what's important to humans, like, you know, personalization and experience and empathy and authentic, stories will not be relevant to AI agents.
What's gonna be relevant to them will be quick access to the information, really structured information, metadata, all sorts of things. SoI'm looking forward to, testing it to see how it how we might adapt, with that kind of environment.
Alina Vandeberghe
I relate a lot with that. I hope that in the future, bots only talk to bots for the boring stuff.
And us to humans, we keep the interesting topics. I I have a personal anecdote.
So I have an aura ring that measures my stress, and I have a therapist. And whenever I talk to my therapist, my you can see that my stress is actually decreasing when I'm talking to him.
Whereas if I talk to Chat GPT about my traumas, my stress remains the same. So I think that humans are still good, and I hope that in the future, again, bots only talk to bots for the boring stuff.
Julia Nimchinski
Alina, what a treat. Thank you so much for the deep dive. And, let's use it as a shameless plug.
Tell us more about Chili Piper, what's the evolution of the platform,and what have you been building all this time?
Alina Vandeberghe
Oh my gosh. I'm building so many things, and I get so excited, but my team hates me for it.
Right now, the thing that I'm most, passionate about is all the things that we've built for ourselves in Snowflake that we have access tothat enables us to accelerate the personalization of, accounts at scale. I want to give back to our customers, and some of the first things are things that I need myself very badly.
I organize a lot of events, and I want to put that on automatic pilot. I'm also creating a lot of personalized landing pages for allmy accounts based on insights that I get on Gong calls because different egments react differently, different personas react differently to themes, different industries react different different sizes, different and creating all of these things at scale is, like, super exciting to me.
It's going to be interesting because the space is kind of getting all everybody's, like, on everybody else. My hope is that we can, Idon't know, be one happy family and all the tech companies can, like, I don't know, maybe consolidate at some point and just we all have one simple option as buyers? But, yeah, my my, roadmap is full of AI agents.
Julia Nimchinski
What are your thoughts on OpenAI releasing all of the sales agents in Tokyo or whatnot?
Alina Vandeberghe
I'm really excited that they're pushing this, and they're showing what can be done, and they're making more mainstream because,otherwise, I I'd have to do the education all by myself. And the fact that OpenAI shows that this is a possibility to others, it, opens up the gate for us.
I think that for us on the SMBs, and on the simple use cases, we've never been able to help anyway. So the fact that they're aiding on that for me is just it's an accelerator.
Julia Nimchinski
And what's your prediction for this year?
Alina Vandeberghe
Because I'm so technical andI get excited about all these things, I imagine that everybody else does and everybody else is building things and everybody else is getting optimized,workflows, but, I realize that I'm in the minority. I hope that people can get excited about it as as much as I am and understand and create more efficiencies in their team so that they can spend more time doing things that are rewarding and that are bringing them joy.
And that is, to me, a catalyst to our best work is to keep us focused on the type of things that we're really good at and the type of things that get us excited. And that's a rare, rare thing to to find, but I hope, that's what AI gives us.
.avif)
.avif)
.avif)
