CMO Roundtable — AI-Native Distribution for CMOs

1417
04:01:09.230 –> 04:01:21.090
Julia Nimchinski: Thanks so much again. And we are thrilled to present Carol Dietrich, our favorite CMO advisor to Lovable, who took a plusie in public. Welcome to the show!

1418
04:01:23.500 –> 04:01:24.610
Julia Nimchinski: Carlo?

1419
04:01:25.710 –> 04:01:40.490
Carilu Dietrich: Sorry, just getting off of mute here and video. So excited to be here, Julia. Thank you for having us. And huge thanks and round of applause to Julia for amazing work to bring so many fantastic people together and to challenge us to think forward. So thank you, Julia, for everything you do.

1420
04:01:40.600 –> 04:01:50.610
Carilu Dietrich: I’m so excited to be here today with a super killer panel of folks to talk about agents as the distribution channel, and the distribution channel for growth.

1421
04:01:50.610 –> 04:02:14.280
Carilu Dietrich: So, traditionally, when we’ve talked about distribution, it’s been our historical channels of email, advertising, partners, websites, but now, agents are becoming the channel. The channel through which we’re selling to a buyer on the other end who’s researching, comparing different options, and in e-commerce, even potentially starting to put

1422
04:02:14.280 –> 04:02:22.770
Carilu Dietrich: put things into carts and buy them. And so, we’re starting to see this emerging field where agents will become the buyer,

1423
04:02:22.940 –> 04:02:45.049
Carilu Dietrich: specifically going out and researching on behalf of their agent operator, and agents as the operator within a company, creating the answers for the company to talk to buyers. And what’s gonna really get crazy here shortly is when agents are talking to agents, and humans aren’t even necessarily in the conversation.

1424
04:02:45.200 –> 04:02:58.170
Carilu Dietrich: It changes everything about the way we think about creating our websites, the content we share, the way we nurture leads, the way we track who knows what, who knows about us already, so…

1425
04:02:58.170 –> 04:03:05.590
Carilu Dietrich: marketing is going to be turned on its head here shortly, and I have this amazing group here to talk about some of the implications

1426
04:03:05.590 –> 04:03:10.040
Carilu Dietrich: Of what we’re seeing already, and give you some of our predictions for what’s coming next.

1427
04:03:10.040 –> 04:03:20.209
Carilu Dietrich: So, first, let’s just get a roundup of each of our speakers and some of their background in this AI and agent space. Marcel, do you want to start us off?

1428
04:03:20.880 –> 04:03:33.399
Marcel Santilli: Sure. Hi, Marcel Santilli here, CEO and co-founder at GrowthX.ai. Prior to GrowthX, I was CMO at a few different companies, like Scale AI, DeepGram.

1429
04:03:33.400 –> 04:03:57.599
Marcel Santilli: HashiCorp, Service Titan, leading growth there. And, a little bit on GrowthX. We build AI power with experts in the loop systems for companies to grow their content and get more visibility in AI. We are lucky to get to work with companies like Lovable and RAMP and Webflow, Sentinel-1, and a bunch of others, and so excited to be here.

1430
04:03:58.320 –> 04:04:00.839
Carilu Dietrich: Great, thanks for joining us. Lisa, would you like to go next?

1431
04:04:01.990 –> 04:04:21.470
Lisa Sharapata: Sure, so I’m Lisa Sharapata, and I’m currently the VP of AI and Go-to-market strategy at Metadata, but I’ve been in the B2B tech space for well over a decade, and in marketing, like, twice as long as that, to age myself. And, it’s been…

1432
04:04:21.470 –> 04:04:32.879
Lisa Sharapata: Really cool to see how things have evolved. Metadata’s been around for about a decade now as well, and is a, ad… digital ad platform.

1433
04:04:32.880 –> 04:04:41.779
Lisa Sharapata: But seeing how AI and the agents that they’re building is really shaped, like, just completely changing the shape of how

1434
04:04:41.780 –> 04:04:58.869
Lisa Sharapata: you know, I’m going to market, and how marketers, you know, can do their digital ads and be thinking about orchestrating the entire go-to-market strategy has been really cool, so I’m excited to share some of what I’m seeing. Great, thank you. Omer, do you want to go next?

1435
04:04:59.340 –> 04:05:06.450
Omer Gotlieb: Sure, and thank you everyone, excited to be here. Omer Gottlieb previously co-founded Tutango, the custom success platform.

1436
04:05:06.500 –> 04:05:17.449
Omer Gotlieb: I’m the CEO, co-founder of Salespeak, and in a nutshell, we’re trying to improve the B2B buying experience by providing an intelligent conversation layer between a buyer and a company.

1437
04:05:17.450 –> 04:05:33.129
Omer Gotlieb: So, if a buyer, by the way, a human being, really, or maybe their agent, if they interact today with a company, they need to, you know, stick with a static website, or with an inexperienced SDR, or with a stupid email, and we enable those kind of conversations to be much more intelligent.

1438
04:05:33.580 –> 04:05:34.400
Omer Gotlieb: And thank you again.

1439
04:05:34.990 –> 04:05:37.530
Carilu Dietrich: And Kevin, do you want to go next?

1440
04:05:37.920 –> 04:05:47.039
Kevin Marasco: Sure, thanks, Claire Lou. Everyone, pleasure to be here. Kevin Murasco, four-time CMO, currently Chief Growth Officer at, Tebra.

1441
04:05:47.170 –> 04:05:57.969
Kevin Marasco: helping independent, healthcare practices apply HR, which is a massive opportunity in and of itself, because I can’t think of an industry in the U.S. that’s more

1442
04:05:57.970 –> 04:06:07.639
Kevin Marasco: ripe for modernization than, healthcare and helping supercharge, healthcare providers with so much burnout. And, if not more…

1443
04:06:07.640 –> 04:06:23.639
Kevin Marasco: even more excited, is about applying AI to totally transform how we do all aspects of go-to-market. You know, in my few decades, I’ve never seen anything disrupting so fast or as furious as what we’re seeing right now, and really excited to be here.

1444
04:06:24.450 –> 04:06:26.270
Carilu Dietrich: Thank you. And Tuba?

1445
04:06:26.900 –> 04:06:51.239
Tooba Durraze: Thanks so much for having me. So, my name’s Chupa, I’m the CEO and founder of Amoeba AI. We’re a newer symbolic AI platform that’s building the new brain for your business, so we have agents that get deployed on top of your data to take a look at what’s happening in your business and to give you recommendations before they actually, offset your business. But by background, I’m a data scientist, a previous VPO product, I qualified, and

1446
04:06:51.240 –> 04:06:52.690
Tooba Durraze: Very happy to be here today.

1447
04:06:53.530 –> 04:07:15.369
Carilu Dietrich: Great. Like I said, such a killer panel. Some of the brightest minds I know that Julia helped me recruit, so thanks for joining me. Let’s just talk about what we mean by agents first, because I think there’s really this, agent washing going on. You know, if everything’s an agent, is nothing an agent, or if every LLM is an agent, does it denigrate what we actually mean by agents?

1448
04:07:15.370 –> 04:07:18.679
Carilu Dietrich: Who wants to weigh in first here?

1449
04:07:18.680 –> 04:07:20.239
Omer Gotlieb: I wanna, I wanna take it for a…

1450
04:07:20.660 –> 04:07:38.209
Omer Gotlieb: Because I want to take it for a second, because I think I agree, I think, you know, there’s a lot of chaos and unclear about what is an agent, what is the agent in marketing. I attended Inbound, the conference last week, and every booth there was, we were doing agenting marketing, and nobody really understands what it is.

1451
04:07:38.270 –> 04:07:53.019
Omer Gotlieb: I don’t think that’s that important, but for this discussion, I would frame agent or an agent marketing into some kind of an intelligent entity that can actually run workflows and take actions for you.

1452
04:07:53.190 –> 04:08:04.690
Omer Gotlieb: So that’s a combination of some things, and of course, the implication, you could say that almost every product may or may not have an agentic component to that as well, but completely chaotic right now.

1453
04:08:05.130 –> 04:08:30.130
Carilu Dietrich: Yeah, and so not just researching your product and LLM, but having more autonomous actions that it’s taking, and more orchestrations, and even deciding things on its own. So, you know, in e-commerce, we’re already starting to hear about people having agents which can shop on their behalf, add things to cart, and potentially even check out. In B2B, we’re in a much different space, although B2B is always

1454
04:08:30.130 –> 04:08:35.130
Carilu Dietrich: following, a fast follow, to B2C. But,

1455
04:08:35.130 –> 04:08:52.799
Carilu Dietrich: you know, many of us on the phone are already seeing a ton of agent traffic to our websites, and Marcel, that’s one of the areas that your company focuses on. What are you seeing in agent shopping? How are the agents performing? How are,

1456
04:08:52.800 –> 04:08:56.849
Carilu Dietrich: How are these, leads that are coming to websites converting?

1457
04:08:57.540 –> 04:09:05.100
Marcel Santilli: Yeah, like, so the way to think about it is, your first sales call for B2B, your first sales call is happening outside

1458
04:09:05.100 –> 04:09:29.830
Marcel Santilli: of your website now, and your conversations, because in AI answers, people are spending on average of 15, 20 minutes, kind of researching, and it’s almost like these people are ingesting that information as the gospel, right? They’re thinking of it as, like, expert advice, so it’s almost the equivalent of word of mouth. So then by the time, if you do show up in those answers, and they come to your website, they’re converting, and we see with some of our

1459
04:09:29.830 –> 04:09:54.630
Marcel Santilli: customers three to five times higher. So the value of that traffic that does end up coming to you is such higher intent that it’s so much more valuable. And it’s not like search is going away, and there’s no traffic there. And so I think it’s similar behavior to e-commerce, except in the e-commerce, you’re checking out on a card, and people are aggregating and curating items, and then you’re finishing the checkout, whereas in B2B, because the process is longer, they’re doing a lot

1460
04:09:54.630 –> 04:09:56.749
Marcel Santilli: More of the research they would do post-sales.

1461
04:09:56.750 –> 04:10:20.140
Marcel Santilli: earlier on, and then they’re giving a lot more context on who they are, their company, and everything about them, which means companies have to figure out how to build this really high-quality, credible content engine that ends up getting cited and show up in these answers to be part of it, right? And the quality bar is even higher, and the long tail is even longer, which now means you have to hit both the speed

1462
04:10:20.140 –> 04:10:24.200
Marcel Santilli: Volume and quality in order to win. Yeah.

1463
04:10:24.200 –> 04:10:42.089
Carilu Dietrich: But, some of the results are showing up really quickly. I mean, one of the things that’s so amazing about the new agents, I guess the new ways of indexing is that some of your new content can show up in a couple days, whereas in SEO, it used to be 6 months sometimes before you changed, right? The agents are moving quickly.

1464
04:10:42.340 –> 04:11:02.339
Marcel Santilli: with Augment Code, one of our customers, I think that within an hour of publishing something, we saw one of their answers changing completely. The context, the way the wording was kind of done. And so, the way I like to think about it is we’re like an inception engine, essentially, except that inception engine actually has to build credibility with your customers as well, because…

1465
04:11:02.340 –> 04:11:04.420
Marcel Santilli: You can’t say the wrong things either.

1466
04:11:05.250 –> 04:11:19.850
Carilu Dietrich: And Kevin, you guys are doing all sorts of Q&A work and have a bunch of agents that interact with different customers that come to the website, including, like, a virtual human, a superhuman, right? Or are you still experimenting with that?

1467
04:11:20.440 –> 04:11:24.429
Kevin Marasco: We are. I think for us, I mean.

1468
04:11:24.670 –> 04:11:37.679
Kevin Marasco: from a B2B standpoint, like, I first thought about it as, maybe not, like, agent-informed buying, or thought of it as not agent buying, but actually agent-informed buying, where, kind of to Marcel’s point, like.

1469
04:11:37.680 –> 04:12:00.489
Kevin Marasco: agents are, you know, the whole top of funnel is being disrupted, so it’s thinking about, like, how agents are transforming the consumption side, but then also, we have to be thinking about and changing how agents change the distribution side, right? And so for us, that’s automating through, an automated system of agents, how they engage buyers through all channels, including

1470
04:12:00.490 –> 04:12:12.870
Kevin Marasco: You know, first of all, identifying targets, but then engaging through email, chat, and then, voice and actual interactions on the website. And so, we’re testing, you know, all of those channels, really.

1471
04:12:14.110 –> 04:12:31.920
Carilu Dietrich: And that leads us right to our next question, which is, as agents, instead of agents as the buyer, agents as the distribution. Tuba, you are doing a lot of work with roadmaps that are including agent-to-agent information requests, and you’re kind of leading in this space. Talk to us a little bit about agents as the distribution.

1472
04:12:32.070 –> 04:12:40.120
Tooba Durraze: Yeah, we’re an intelligence platform, and a lot of folks want to kind of see the intelligence being converted into some sort of an action.

1473
04:12:40.120 –> 04:13:05.079
Tooba Durraze: And then they don’t want to be the ones who are orchestrating that action, right? As long as they’re trusting the insights, etc. So we get requests on agent-to-agent communication in terms of orchestrating something in other platforms, or in terms of sharing information as well. If I think of, kind of Lisa’s product, as well, like, potentially, like, creating campaigns and metadata, etc. So more and more, we’re going to see that you use agent-to-agent interaction

1474
04:13:05.080 –> 04:13:08.230
Tooba Durraze: For the grant work, as long as you’re approving the recommendation.

1475
04:13:09.560 –> 04:13:18.919
Carilu Dietrich: And Lisa, you guys have rolled out an MCP offering recently, and are kind of pretty deep in this agent world. Like, tell us more how you’re thinking about it as distribution.

1476
04:13:18.920 –> 04:13:21.380
Lisa Sharapata: Yeah. So, less…

1477
04:13:21.730 –> 04:13:40.949
Lisa Sharapata: on the consumer side right now, it’s more agent-to-agent, like you said. So, thinking about things like, how do I analyze current campaigns? So, there’s multiple agents that it actually takes to do that. So, looking at performance, looking at different experiments that have been run, different budget groups, different audiences.

1478
04:13:41.100 –> 04:13:48.199
Lisa Sharapata: To identify, kind of, what, kind of, the best use case would be, and then you can…

1479
04:13:48.360 –> 04:13:58.559
Lisa Sharapata: you know, inside of Claude today because of this MCP server connector, which I’ll be showing after this. It’s kind of complicated, right? Sounds complicated.

1480
04:13:59.140 –> 04:14:13.669
Lisa Sharapata: these agents are all talking to each other, so I can create a new campaign with that information, I can build new audiences, I can dedicate budget, and I can deploy it right from Claude today.

1481
04:14:14.120 –> 04:14:28.240
Lisa Sharapata: like, and another example would be we have a bid agent that sits inside of our platform, but it works with LinkedIn. So, if you turn it on, it’s figuring out the LinkedIn algorithm, it’s adjusting

1482
04:14:28.470 –> 04:14:48.380
Lisa Sharapata: your ad spend within the limits you’ve defined, and it’s autonomously making decisions based on the criteria that you’ve set for it, on your behalf 24-7. But you… it’s like a toggle on, toggle off. You set the parameters, you can change the parameters, and then it’s making decisions for you.

1483
04:14:49.500 –> 04:15:08.499
Omer Gotlieb: And, Carilo, I wanna… I wanna double down for a second on the genting distribution, and I want to warn the audience from a trap that a lot of companies are falling into. I don’t know about you, but when I open my email every day in my inbox, every day, there are 15 messages that look the same. Omer, quick question. Omer, can you join my podcast?

1484
04:15:08.500 –> 04:15:19.730
Omer Gotlieb: When I look at my LinkedIn post, I see comments that it’s clear the AI actually wrote them. I think some people say, hey, it’s very easy, let’s take an AI agent, you know, buy it or develop it, that can

1485
04:15:19.730 –> 04:15:22.379
Omer Gotlieb: Produce some content, and let’s just put it out there.

1486
04:15:22.500 –> 04:15:34.089
Omer Gotlieb: And the problem is that 100,000 people are actually doing that, not in a very intelligent way, and we, if we wear our Byron hat for a second, are being stammed and flooded. And so I think…

1487
04:15:34.150 –> 04:15:42.730
Omer Gotlieb: I think one of the most common traps that I see is people use AI, but not intelligently. They use AI to simply

1488
04:15:42.930 –> 04:15:59.190
Omer Gotlieb: automate broken processes, and I think the best example that I have personally is, you know, think about those AI SDRs and things like that, you know. Did you ever want to speak with an SDR? Why would I want to speak with an AI SDR? So why are we taking an SDR and amplifying it?

1489
04:15:59.190 –> 04:16:04.479
Omer Gotlieb: We want to do something different, and I think AI has the opportunity to do that, so when you think about

1490
04:16:04.550 –> 04:16:19.320
Omer Gotlieb: using agentic distribution, it’s not about the automation. Automation is an important part of it, but I think it’s much more about the intelligence that you put inside, the experience that you put inside, and I think we can do great things with it.

1491
04:16:19.320 –> 04:16:34.369
Tooba Durraze: Can I add a lens to that? I think I’ll push back on receiving massive amounts of, kind of, AI inbound, etc. The piece that you’re missing is the agent on your side, essentially, so you could say, hey, I’m only interested in, say, like, speaking on podcasts.

1492
04:16:34.370 –> 04:16:41.869
Tooba Durraze: We never… we never closed the loop, right? Essentially, that way. If it was agent-to-agent, I bet it would be really useful, because you would only

1493
04:16:41.870 –> 04:16:50.650
Tooba Durraze: relevant to you. But to your point, we rushed to automation, and we, like, forgot to kind of bridge that gap.

1494
04:16:51.300 –> 04:16:52.079
Omer Gotlieb: I agree.

1495
04:16:52.600 –> 04:16:52.950
Marcel Santilli: Think…

1496
04:16:52.950 –> 04:17:07.749
Carilu Dietrich: It’s funny, not on the agent topic, but one of my old colleagues is the VP of Marketing at Deal, and they did their ad campaign with Veo, a totally AI-generated video that looks so amazing.

1497
04:17:07.750 –> 04:17:22.270
Carilu Dietrich: And, she’s like, we did it in just a couple days instead of weeks. And I ended up talking to her about it, and they have a pretty extensive video and design team that has a lot of specialty. There was a producer, and they looked at costume design and

1498
04:17:22.270 –> 04:17:47.249
Carilu Dietrich: characters and storyboards, like, they went through a bunch of thinking processes to then use VAO in two days to create something amazing. But I guess, kind of coming back to what you’re saying, Omar, what it really reinforced to me is that the AI can be so powerful if it’s built on top of people with knowledge doing something interesting, right? Right? Like, no one wants to replicate the annoyance of, like, forms and an uninformed reach-out that’s, like, gets you everything

1499
04:17:47.250 –> 04:17:48.150
Carilu Dietrich: wrong.

1500
04:17:48.150 –> 04:17:58.619
Carilu Dietrich: But if it was, like, you know, some of these, forward-deployed engineers that developer, companies are employing as their,

1501
04:17:58.620 –> 04:18:11.180
Carilu Dietrich: as their SDRs, who are, like, instead of, like, trying to qualify you, there’s, like, hey, you know, do you want some sample apps? Like, here’s some common problems people have. If you have any problems, like, setting up, I’m here for you. Like, if we can really, like.

1502
04:18:11.180 –> 04:18:25.709
Carilu Dietrich: create the ideation, to create the agent that’s actually really helpful, it could be delightful, and then humans can be there at the right time. You know, the dream is better than just, like, the perpetuity of horrible agent emails in our inboxes.

1503
04:18:26.770 –> 04:18:39.709
Lisa Sharapata: Well, and just to circle back on what I was trying to get across, too, is that these agents can get you so far, and actually, you could even just say, build it and deploy it, right? But the human in the loop

1504
04:18:39.740 –> 04:18:53.269
Lisa Sharapata: is, I think, very relevant still in the example I was giving as well, where you’re looking at what was analyzed, and you’re still making strategic decisions. And, you know, I can still push back and say.

1505
04:18:53.270 –> 04:19:05.749
Lisa Sharapata: okay, well, I really need to generate this much pipeline in this short of time, and I only have $20,000, not $30,000, so can, you know, you help me reimagine it based on that, not just…

1506
04:19:05.780 –> 04:19:07.590
Lisa Sharapata: like, running…

1507
04:19:07.750 –> 04:19:26.999
Lisa Sharapata: you know, wild, with no guardrails. And same thing with the bidding. No one wants to sit there and bid on campaigns 24-7, trying, you know, moving it up a dollar and down a dollar, so I think that there’s really great use cases that can help free up capacity to be

1508
04:19:27.000 –> 04:19:31.940
Lisa Sharapata: More creative, and more thoughtful, and try different things, and…

1509
04:19:32.330 –> 04:19:39.720
Lisa Sharapata: and then, you know, automate things that, frankly, like, no human could really even do at that scale. So…

1510
04:19:39.930 –> 04:19:43.700
Lisa Sharapata: you know, back to what you were saying, Omar, too, is it’s not just

1511
04:19:43.900 –> 04:19:50.289
Lisa Sharapata: you know, throw an agent on that and spit out 5X more emails to people.

1512
04:19:50.540 –> 04:19:55.550
Lisa Sharapata: That’s, I don’t think, a responsible use of distribution.

1513
04:19:56.800 –> 04:20:10.730
Carilu Dietrich: So, there’s these two aspects we talked about when we were prepping about agents as the interface, like the buyer who’s coming to our website, and agents as the operator, and the operator being the one who’s, like, compiling the information. Kevin, you’ve done some of the more advanced

1514
04:20:10.730 –> 04:20:29.200
Carilu Dietrich: operator agents that I’ve seen. I know you have one for competitive battle cards, you have one for customer references, you have some that are tied together for a number of different steps that inform each other. Like, talk to us a little bit about agents as the operator inside the company doing the selling.

1515
04:20:31.080 –> 04:20:36.720
Kevin Marasco: Yeah, I think, you know what’s key is… is not just…

1516
04:20:36.860 –> 04:20:45.000
Kevin Marasco: we have, like, base-level automation, and I think the trap here is just doing… just automating something, right? Making it a bad process. How do we take…

1517
04:20:45.220 –> 04:20:57.300
Kevin Marasco: agents and apply it to up-level what is the art of the possible, and what you could do. And that, with the increased, like, context library and the speed to be able to analyze, execute, make decisions.

1518
04:20:57.300 –> 04:21:06.969
Kevin Marasco: And give agency, to the agents. It, it, unlocks, on the operating side, some pretty, pretty exciting things, so…

1519
04:21:06.970 –> 04:21:21.389
Kevin Marasco: I think for… for me, it first started where we always want more resources, right? And what could we do with more resources? And you could apply it to whatever area of your funnel or growth lever you have. And, you know, historically, I think, oh, well, we had

1520
04:21:21.390 –> 04:21:31.200
Kevin Marasco: you know, another, you know, product marketing team was 3 times as big, we could create more battle cards and enablement content and help, you know, our sellers, sellers convert more.

1521
04:21:31.200 –> 04:21:44.460
Kevin Marasco: well, now with agents, we could take that work and we could scale. But what the important part is, we gotta scale it, like, in an intelligent way, do the right things. And, like, like, Omer’s, like, SDR’s example, yeah, you could just take an agent and

1522
04:21:44.460 –> 04:21:53.919
Kevin Marasco: send more emails faster, but I think that’s… that’s the trap we have to avoid. Like, how can we make that experience better for the buyer, where

1523
04:21:53.920 –> 04:22:08.450
Kevin Marasco: what maybe was the job of the historic SDRs we’ve known over the past couple days, but supercharge them with information to be hyper-consultative and guide the buyer through, their buying journey and add value to the experience. I think

1524
04:22:08.450 –> 04:22:16.779
Kevin Marasco: Like, that’s been, for me, the trap is, like, it usually starts with, hey, it unlocks this incredible amount of scale and leverage, but…

1525
04:22:16.780 –> 04:22:32.690
Kevin Marasco: then also have to, like, jump to, hey, how do we do it in a smarter, better way to make a better, you know, buying experience and a better, you know, selling experience, too. But we’re, I’ve really tried to take a lot of the jobs to be done and try to figure out, yeah, how do we scale it, but also how do we do it better?

1526
04:22:33.260 –> 04:22:42.810
Carilu Dietrich: It’s funny because, you know the Gartner hype cycle curve, where you get so excited, and then you enter the trough of disillusionment, and then you come back up on the… I don’t know.

1527
04:22:42.810 –> 04:23:07.759
Carilu Dietrich: hill of value, whatever it’s called. Like, one of the things I’ve heard is people got so excited about their agents, they went out and built all these agents, and they have all these agent team members, and then, the GM of AI for Microsoft and I were speaking at another CMO event last week, and she was saying, after they’d created so many agents, they ended up in this, like, agent overload spot, where, like, there were too many agents, and it was too confusing, and the agents were…

1528
04:23:07.930 –> 04:23:23.840
Carilu Dietrich: you know, kind of overwhelming, and then they ended up kind of coming out of that trough of disillusionment back to value to kind of rationalize. So, it does feel like we’re in the, like, agent propagation phase right now, where we’ve created agents for everything, and then maybe it’s going to come back and rationalize.

1529
04:23:23.840 –> 04:23:36.229
Carilu Dietrich: And one funny example, I know, Marcel, you guys have an agent that’s a fact-checking agent, which literally goes to fact-check all your agents who might be doing things wrong. You want to talk to us a little bit about developing that?

1530
04:23:36.470 –> 04:23:42.839
Marcel Santilli: Yeah, like, I think one thing that is so important that a lot of people miss is starting with outcomes.

1531
04:23:42.840 –> 04:24:06.770
Marcel Santilli: and working your way backwards. It doesn’t matter if it’s agent, deterministic AI workflows, or automation or not at all, it’s like, you need to figure out how to get to great. Like, going fast in the wrong direction is actually worse than going slightly slower in the right direction. So it’s like, agents doing the wrong things are off. And if you don’t have a way to evaluate what the agents are doing, the agents are not going to do well, because they don’t know how to even, like, self-criticize and self-correct

1532
04:24:06.770 –> 04:24:30.979
Marcel Santilli: and follow, like, a bigger planning if you don’t have the right context. So in this example of the fact checker, we have this workflow that breaks down a piece of content into chunks, and then analyzes what parts of the chunk have different passages, and comes up with questions to go answer, and then does deep research in parallel with all those questions, and then comes up with a confidence score, or whether or not those statements are true, or if there’s any misplacement or inaccuracy

1533
04:24:30.980 –> 04:24:54.569
Marcel Santilli: or better ways to do it, and then it fetches the writing guideline, and then auto-corrects anything that has a strong confidence that it’s not factual or that it’s misleading in some ways, right? But that’s only one step of the job to be done, and it’s all based on the definition of quality. So we have a customer in the healthcare space, there’s a lot of medical aspects to it, you know? A company like Sentinel-1, or Abnormal Security, or Augment Code, they’re more technical.

1534
04:24:54.570 –> 04:25:06.170
Marcel Santilli: There’s other aspects to it, and so it’s not about having one agent that then abstracts the whole thing away, it’s figuring out what are all the steps, that… to get to great outcomes and outputs.

1535
04:25:06.170 –> 04:25:23.469
Marcel Santilli: And then figuring out where you have enough confidence by having experts in the loop that then you can delegate to an agent, because then you know how to evaluate. And so I think that kind of just shifts where humans are going from not just the output bar raisers and process architects, but also getting the right inputs.

1536
04:25:24.700 –> 04:25:38.799
Carilu Dietrich: And Omer, we were talking, before, kind of, like, leading on from that, that there’s also this transition period. You know, agents aren’t the only ones coming to our websites right now. Humans are still coming to our websites, too. In fact,

1537
04:25:38.860 –> 04:25:48.699
Carilu Dietrich: from a website traffic… well, sorry, I guess that stat doesn’t apply as well. But, how are you thinking about this parallelizing,

1538
04:25:48.870 –> 04:26:01.809
Carilu Dietrich: experiences for humans and experiences for agents. Are we gonna end up with, like, hidden things on our websites, or parallel websites, or… or how do we navigate when the distribution channels are, multiple right now?

1539
04:26:01.810 –> 04:26:10.009
Omer Gotlieb: Yeah, I think it’s a great question, and definitely we are in a transition phase. I’m sure that in the future it will be full agent-to-agent, because

1540
04:26:10.010 –> 04:26:32.839
Omer Gotlieb: why not? It’s gonna be very, very smart and easy to do, but I don’t think the technology is there right now. I do see even now, you know, we offer our customers a way to actually have a shadow website, so one website is fully designed, optimized for human being accessing you, and the other one is, you know, when ALM is accessing you, and not always it’s the same thing. ALM is asking a couple…

1541
04:26:32.890 –> 04:26:45.410
Omer Gotlieb: In a human being, you know what they’re asking, you understand what a human being, LLM, you don’t know a lot of things, but you need to present other things. The way they actually consume the data is different. Now, again, in the future, it’s going to be an agent, and…

1542
04:26:45.970 –> 04:27:00.389
Omer Gotlieb: whether there will be a website or not, nobody knows, right? If, you know, if I have an agent and you have an agent as a company, then why do we need a website? Again, a website will be used for another use as well. So, I do think that…

1543
04:27:00.600 –> 04:27:12.589
Omer Gotlieb: It needs to be a differentiation between how do we provide information, communication, conversation, whatever we want to call it, to a human being, versus how do we want to provide that

1544
04:27:12.720 –> 04:27:27.780
Omer Gotlieb: communication and conversation to a technology, an agent. I think both of them needs to be different than what we have today. Again, you know what? I spoke with many marketers recently, and my best advice, which is so nice, but it’s actually working, is

1545
04:27:28.240 –> 04:27:36.659
Omer Gotlieb: really designed for the buyer. Designed for the customer. If we understand that the experience needs to be in favor of the customer.

1546
04:27:36.800 –> 04:27:45.550
Omer Gotlieb: then we’re gonna win, and we’ve done, you know, even me in my previous company, you know, designing websites with forms, that’s not good for the buyer, it’s good for me, right?

1547
04:27:45.650 –> 04:27:59.739
Omer Gotlieb: putting a threshold that you have to speak with a salesperson, a junior salesperson, in order to get to the right salesperson, and then only he will call the SE if you are qualified. That’s not good for the buyer, and the reason we’ve done it is because that was the only way. And I think…

1548
04:28:00.150 –> 04:28:05.130
Omer Gotlieb: The amazing opportunity in AI is really bringing all that kind of

1549
04:28:05.530 –> 04:28:10.350
Omer Gotlieb: Amazing service, amazing intelligent, amazing experience, up front.

1550
04:28:11.270 –> 04:28:21.420
Omer Gotlieb: And if you think about it, that’s the way I think you need to design your website, your content, your email, anytime you… anytime you interact with a buyer, you can actually enhance those things.

1551
04:28:21.830 –> 04:28:26.509
Omer Gotlieb: And it’s interesting you just say that, because I do think it’s changing the…

1552
04:28:26.530 –> 04:28:34.780
Carilu Dietrich: way we think about our org design. So, for instance, you know, there’s these automated, SDR… SDR

1553
04:28:34.780 –> 04:28:59.320
Carilu Dietrich: buyers, or, you know, superhuman SDRs that can talk to you at any time. But they’re really focused on qualifying someone, or the SDR experience. They’re, like, replacing a role we had, which was an SDR. And there’s this one company called OneMind, which is… Kevin is trying them out, and I’ve been really intrigued by, because it started out as an AI SE, a sales engineer, which was able to help assist in selling, which

1554
04:28:59.320 –> 04:29:23.540
Carilu Dietrich: has always been a really, like, tight, role, because it’s such a specialized role, and they’re so in demand, and they’re kind of expensive, but this SE could talk to you at any time of day and have all the information. But what, OneMind found out was that this SE did a great job of talking to customers, because it had all the information, and it actually does a great job of qualifying, whereas some of the qualifying tools qualify, but if they run into a customer.

1555
04:29:23.540 –> 04:29:27.460
Carilu Dietrich: The customer has to bounce out and go back into, like, a customer support chat

1556
04:29:27.580 –> 04:29:45.250
Carilu Dietrich: funnel. And, you know, when you can have this agent that has infinite knowledge, you know, kind of like the download to Keanu Reeves in The Matrix, you can have one agent that talks to anyone and, like, routes them to the right place and gets them the right information, which kind of totally changes the…

1557
04:29:45.250 –> 04:29:55.169
Carilu Dietrich: The mindset of, the way our teams were organized into, like, new processes is just, like, serve the customer and give them all the information they need without the gates.

1558
04:29:55.170 –> 04:30:18.639
Carilu Dietrich: But I think what’s terrifying about that from a marketing perspective is, like, how do we track that? Like, how do we know what our revenue is gonna be, upstream? Because it’s like, people are gonna come in, like, 95% qualified, and you’re gonna be able to see a script, maybe, of the superhuman, but then, like, the final closing is gonna be so small, that you don’t know what’s working until you don’t have the revenue.

1559
04:30:18.640 –> 04:30:21.250
Carilu Dietrich: Kevin, you came off mute. Do you have a commentary on this?

1560
04:30:21.250 –> 04:30:21.779
Carilu Dietrich: My ramp?

1561
04:30:21.780 –> 04:30:31.399
Kevin Marasco: Yeah, I got the, you know, it’s a group of marketers, so there’s an answer, and of course, it’s an acronym. It’s AQL, an Agent Qualified Lead. You know, it’s just…

1562
04:30:31.400 –> 04:30:31.930
Omer Gotlieb: Oh, buddy.

1563
04:30:31.930 –> 04:30:41.749
Kevin Marasco: Or a BQL, a bot-qualified lead, yeah. But what does that mean, and how does a AQL compare to a traditional, you know, MQL.

1564
04:30:41.750 –> 04:30:42.310
Omer Gotlieb: Yeah.

1565
04:30:42.310 –> 04:30:42.930
Kevin Marasco: And SQL.

1566
04:30:42.930 –> 04:30:59.240
Omer Gotlieb: I would even argue, do we need to qualify? The reason we had to qualify is because only qualified customers, the buyers, get the best service. What if we can provide everybody the best service? Again, the challenge would be completely agreeing with forecasting, because

1567
04:30:59.240 –> 04:31:04.409
Omer Gotlieb: Qualifying… Qualification is also good for forecasting, but if you put forecasting aside.

1568
04:31:04.410 –> 04:31:10.230
Omer Gotlieb: Maybe? Again, haven’t thought about it before this conversation, maybe the notion of the qualified lead doesn’t really matter.

1569
04:31:10.570 –> 04:31:17.609
Omer Gotlieb: I mean, what does matter is, okay, do we have a good conversation with the customer? Can we provide value for them?

1570
04:31:17.890 –> 04:31:20.009
Omer Gotlieb: And can we actually move them further?

1571
04:31:20.580 –> 04:31:29.879
Omer Gotlieb: Again, I do agree that things will change how we measure things, how we focus things, and I have no way, no idea to focus how they’re going to look like, but they will be different.

1572
04:31:30.110 –> 04:31:33.360
Carilu Dietrich: Tuba, did you come off mute? It looked like you had something to say.

1573
04:31:33.360 –> 04:31:56.680
Tooba Durraze: Yeah, I… I think… I agree with Umar. I feel like it’s a computation problem, like, we can’t really compute across everything, we don’t have enough capacity for that, so we, like, create these funnels. But you’re right, I think our qualification criteria would either not exist or be super, super open if we get into a world where it’s just agent-to-agent selling. Like, there shouldn’t be this idea of, like, what is the difference between a bulk qualified lead and a human-qualified lead?

1574
04:31:56.680 –> 04:32:02.180
Tooba Durraze: We would do that, obviously, because incentive structures, when you get paid out on leads, have to be aligned in that way.

1575
04:32:02.180 –> 04:32:09.320
Tooba Durraze: But I think, I think in the future, yeah, some of this qualification criteria will, like, fall off. I totally agree with that.

1576
04:32:11.260 –> 04:32:36.200
Carilu Dietrich: It’s so crazy to imagine everything we know being totally different. It’s like, how the body regenerates all of its cells in so much time. It feels like the marketing tech stack is changing, the marketing processes are changing, the way buyers buy is changing, the way we evaluate things are changing, you know, potentially the evaluation, topics that we use. Like, in this world of agent-to-agent, it seems like trust and safety and government

1577
04:32:36.200 –> 04:32:38.480
Carilu Dietrich: governance are gonna be really important. Like.

1578
04:32:38.480 –> 04:32:50.159
Carilu Dietrich: you know, in B2C, you might let an agent buy something because it’s a small amount, and maybe just on a credit card where you could cancel it or something. Are any of you guys kind of thinking forward to this, like.

1579
04:32:50.160 –> 04:32:53.640
Carilu Dietrich: Trust, governance, space.

1580
04:32:54.510 –> 04:32:54.900
Marcel Santilli: reality.

1581
04:32:54.900 –> 04:33:05.610
Lisa Sharapata: Carol Lou, I have a question to flip back on you, and it kind of pertains to what you’ve been thinking about, too, at Metadata, but… so you’re, fractional for lovable.

1582
04:33:05.610 –> 04:33:16.020
Lisa Sharapata: Lovable has basically done what Omar and Kevin were talking about, like, bringing everything to the surface, right? The product is there, it’s very product-led, you can

1583
04:33:16.020 –> 04:33:29.899
Lisa Sharapata: play around with it, there’s no gates, but it’s like, you know, you have to unlock, I guess, certain levels to get deeper into it, be able to use it to deploy a site or push something live, like, you have to have enough credits.

1584
04:33:30.020 –> 04:33:32.759
Lisa Sharapata: And you have to have trust

1585
04:33:32.759 –> 04:33:55.719
Lisa Sharapata: to do that, right? And, like, that’s one of the things I’ve been playing with, like, free trials, for example, but someone’s gonna have to share their LinkedIn information, or share their account for their Gmail, or whatever it is, in order to, like, let the agent go. So, I’m curious, like, your thoughts on that, given that you’re really leading in that area.

1586
04:33:56.830 –> 04:34:16.249
Carilu Dietrich: Yeah, I mean, I think trust becomes paramount and potentially a differentiation between different companies, and we’ve certainly seen that with Anthropic’s market positioning coming out of Cloud… I mean, sorry, coming out of OpenAI to create this kind of ethical, alternative and different features that reinforce trust.

1587
04:34:16.250 –> 04:34:24.949
Carilu Dietrich: So… And Levable’s done different things, like we had rolled out an agent feature, earlier this year.

1588
04:34:24.950 –> 04:34:49.939
Carilu Dietrich: that helps, reduce errors by 90% by kind of checking over the code and checking over your requests to make the agent work better. So again, kind of a check bot. And then also a security agent that, like, reviews everything and makes recommendations about how to make your apps more secure. And for people that don’t know, I’m sorry, Levable is a vibe coding platform, one of the leaders, where you can take an idea to an application, just by chatting with AI.

1589
04:34:50.180 –> 04:35:10.869
Carilu Dietrich: And it creates these really gorgeous visual sites that then can be full-stack apps and be deployed, and people are starting businesses, or prototyping, or creating entire products for companies. So yeah, I mean, I think trust and reliability is kind of a centerpiece of AI agent offerings, and we’ll continue to differentiate

1590
04:35:10.869 –> 04:35:13.709
Carilu Dietrich: And always the challenge with that is that

1591
04:35:13.710 –> 04:35:25.790
Carilu Dietrich: you know, it takes a long time to build up trust and not a lot of time to lose it if you do something wrong. And so I think that’s why we see so many different security companies and AI security companies

1592
04:35:25.790 –> 04:35:36.140
Carilu Dietrich: really having astronomical momentum, because AI creates so many new vectors for attack and surface areas, that really need to be protected.

1593
04:35:37.950 –> 04:35:42.330
Omer Gotlieb: I think trust is a very serious temporary problem.

1594
04:35:42.570 –> 04:35:50.080
Omer Gotlieb: And the reason I’m saying temporary is we all remember, you know, can I trust, to put my data in the cloud, right? Financial data in the cloud!

1595
04:35:50.340 –> 04:35:59.830
Omer Gotlieb: it’s a non-issue anymore, and this will be a non-issue. It might take longer or sooner, I don’t know. But right now, I think there’s two types of trust issues. One.

1596
04:35:59.980 –> 04:36:18.260
Omer Gotlieb: as business executives, do we trust the agents that we actually, you know, build or buy in order to do a good job for us, or put in front of the customers? And even Marcel had a good example. They did not trust their internal agents, they built another agent to actually make sure that they can trust it, which makes complete sense.

1597
04:36:18.290 –> 04:36:35.419
Omer Gotlieb: But I also think there’s some trust issue from your customers. I mean, can I trust this AI that I’m actually speaking with? I think that’s also a done deal. I mean, I don’t know about you, but personally, if I had the choice between getting my information from

1598
04:36:35.450 –> 04:36:42.130
Omer Gotlieb: a salesperson, and usually these are junior salespeople at the beginning of the process, versus an AI,

1599
04:36:42.419 –> 04:37:00.399
Omer Gotlieb: my first bet was gonna be on the AI. Now, I know I’m an early adopter, but I think more and more people are going to be educated by big companies that, yeah, AI can actually provide, you know, a better solution, a good solution. So I think the trust right now is mainly business owners, can I trust

1600
04:37:00.470 –> 04:37:06.060
Omer Gotlieb: the AI to do a good job for me, but it will… it is a temporary problem.

1601
04:37:06.060 –> 04:37:24.759
Tooba Durraze: I feel like it’s inherently a human psyche problem. Humans have a hard time giving up control, so the guides that under trust, quote-unquote. I do think it’s going to be a little longer than it used to be, because you’re not talking about systems that are non-deterministic, right? And every… all of these systems have drift. How many times do we wake up, like.

1602
04:37:24.900 –> 04:37:33.150
Tooba Durraze: you know, one day, and we’re like, okay, well, the output suddenly changed. So you’re gonna have to design these systems around trust, and I’m…

1603
04:37:33.150 –> 04:37:48.840
Tooba Durraze: I’ll be curious to see where the onus lies. Does it lie on entirely the customer side? Does it lie entirely the builder’s side, on how you reveal, like, certain things within your product and how things are functioning, how agents are functioning, so people can trust that something good is happening behind the scenes?

1604
04:37:48.840 –> 04:37:59.330
Tooba Durraze: But I feel like this is going to be a little bit longer than my financial data in the cloud, just because people don’t like the notion of, like, not knowing how exactly something would map out.

1605
04:37:59.490 –> 04:38:03.180
Omer Gotlieb: Well, financial data in the cloud took a few years, so I think, yeah.

1606
04:38:03.189 –> 04:38:08.449
Tooba Durraze: I think this is, like, 5 to 10 is my, in my head, but yeah.

1607
04:38:08.680 –> 04:38:32.270
Carilu Dietrich: I think one of the things on the trust side is there’s the two sides. They’re like, is the brand trustworthy, right? Like, can our apps be… can Lovable’s apps be secure, and thoughtful, and scalable, all of those things. And then there’s the trust on the consumer side, and one of the things that I’ve been surprised by is how authoritative LLMs are in telling you an answer, especially in product comparisons.

1608
04:38:32.290 –> 04:38:46.599
Carilu Dietrich: Like, they’ll tell you who’s the leader, and then you, like, click on the little link and look at it, and you’re like, oh, this is a secondary site that the vendor put up that compares themselves, which creates, like, slightly less,

1609
04:38:46.599 –> 04:38:56.929
Carilu Dietrich: research support than we might have had when we were looking through Google link responses and SEO, and we kind of did our own self-filtering to look at the validity of different websites.

1610
04:38:56.930 –> 04:39:00.959
Carilu Dietrich: Marcel, this is kind of your world, because…

1611
04:39:00.960 –> 04:39:19.919
Carilu Dietrich: the way all the content you’re generating that shows up in LLMs, from our marketing perspective, we want it to, like, win the LLM and be authoritative. And then also, from the consumer side, we want to be thoughtful about how we’re evaluating brands. Like, how are you building trust into the content?

1612
04:39:19.919 –> 04:39:26.569
Carilu Dietrich: In a way that, even when people kind of click through to really evaluate, they’re still trusting what you’re doing.

1613
04:39:27.009 –> 04:39:39.539
Marcel Santilli: I think this is, like, at the core of what we do, but also at the core of what everyone has, as marketers, have to figure out. And so, like, I’ll give a few examples. One of our customers, they’re in the women’s supplement space.

1614
04:39:39.539 –> 04:39:55.399
Marcel Santilli: started by, a well-known founder, and, they have a chief medical officer who has a background in endocrinology and a holistic doctor, right? So we had to set the right context, we had the fact-checkers, you had the… all these steps, but then, we…

1615
04:39:55.399 –> 04:40:19.509
Marcel Santilli: gave her a piece of content for calibration, this chief medical officer, and she left 30 comments in the document, right? And they were all nuanced medical things. So then we had to reverse engineer those comments to understand, like, what was about this? It was like an oversimplification of the, you know, how this ingredient plays into your hormones, or something like that, right? And so a lot of the trust and quality is not achieved from zero to one.

1616
04:40:19.509 –> 04:40:37.219
Marcel Santilli: you know, when I was at Scale.ai, we’re doing a lot of work for the Waymos of the world. You don’t go from, like, level one autonomy to self-driving Waymo in one system, in one shot, right? And it’s all about calibrating and setting the right context, and then you have to ground that information on something. Like, when we were developing all these integration pages.

1617
04:40:37.219 –> 04:41:02.079
Marcel Santilli: for Webflow, and you go to the Stripe integration page there. Like, that has to be grounded on their docs. It has to be grounded on something in order to know if it’s factual or not, and that’s how you build trust. But if you don’t have confidence in your grounding and what your source of truth is, then you can’t actually go after it. And then the other piece that I would say as a practical piece of advice for everyone is, like, use multiple systems. So, for instance, like, when we’re doing a deep research.

1618
04:41:02.079 –> 04:41:26.949
Marcel Santilli: using EXA AI or Tavly to do the deep research. We might use perplexity’s deep research as a fact checker, or vice versa. And that way, you’re using multiple layers, but also don’t try to abstract experts in the loop too early before you gain that confidence, you know? And so the way we look at it and measure it is how many human interventions are needed in order to get something out, like a piece of content, or a page, or something that’s published.

1619
04:41:26.999 –> 04:41:36.549
Marcel Santilli: And if it goes from 30 comments to 1 to 0, then you can start to build confidence in those systems. And you need to do that internally first, as well, yeah.

1620
04:41:38.080 –> 04:41:52.269
Carilu Dietrich: agents talking to agents, checking agents, agents as the distribution channel. What have you guys seen, as being some of the most impactful, like, practical examples? I think, like, we risk being really abstract.

1621
04:41:52.270 –> 04:42:00.600
Carilu Dietrich: In our conversation, but practical impacts that you’ve seen so far of agents in your companies or in your marketing?

1622
04:42:01.010 –> 04:42:19.090
Tooba Durraze: Can I give a funny example, because I’m not a marketer by background? When I first started this company, I was trying to do my best, as much as I could do, with some, like, LinkedIn marketing, and then I was like, I don’t feel like I’m doing great, but I feel like I’m doing okay. Took all that data, tossed it at Amoeba, at Nurismolic, and

1623
04:42:19.240 –> 04:42:38.650
Tooba Durraze: it gave me a lot of really good feedback on how, like, I was basically creating a lot of boredom for my, like, end users, and the way I was talking about things repetitively, it’s not the right kind of fit for the target buyer that I was trying to sell it to. All things that are very obvious to marketers, but maybe not as obvious to folks who are outside of that.

1624
04:42:38.650 –> 04:42:56.369
Tooba Durraze: And then we kind of evolved that into, like, can you run an audit on my LinkedIn, as an example, like, ever so often? And we’ve seen a big difference in sort of, like, how that’s materialized in, like, pipeline for our company. So I think that’s an example of, for us, like, a lot of times folks will say, like, I don’t have

1625
04:42:56.370 –> 04:43:10.110
Tooba Durraze: maybe sometimes enough data to get started with agents, etc. That’s an example of, like, even with a little bit of data, if you get started, the agents will, like, kind of… can help you guide your system in a way that you build better proficiency or competency.

1626
04:43:10.300 –> 04:43:12.079
Carilu Dietrich: I love that.

1627
04:43:12.690 –> 04:43:13.810
Carilu Dietrich: Other examples?

1628
04:43:13.810 –> 04:43:36.110
Marcel Santilli: other one is when we started hiring and accelerating hiring for GrowthX, like, we started getting almost 1,000 applicants a day on some roles, and so there’s just no way to tackle it. And so one of the ways we did it was to kind of decompose the process and recompose it back up, and so we just built a workflow that kind of goes through and gets people to apply and write their answers and some of the questions.

1629
04:43:36.110 –> 04:43:37.610
Marcel Santilli: And then if they pass that.

1630
04:43:37.610 –> 04:44:01.860
Marcel Santilli: a process, there’s a recruiter in the loop that looks through the best 20% of candidates, passes them through to a video interview, and within that video interview, then we transcribe and do the same process, then they go through an assessment. And even though that process might sound like a bad experience for the candidates, we actually got a lot of feedbacks, because it’s like, we’re asking them for 30 minutes of their time, but in exchange for that, even though they don’t talk to a human right away.

1631
04:44:01.860 –> 04:44:10.650
Marcel Santilli: By the time they do get to talk to us, they’re already in the final steps, and so we can go from application to offer in, like, days instead.

1632
04:44:11.190 –> 04:44:19.539
Marcel Santilli: And so, like, that was just, like, a way to be able… that would have just been impossible to do without actually AI and workflows and agents as well.

1633
04:44:20.630 –> 04:44:32.019
Kevin Marasco: One of my favorites right now is, using agents to two-fold on the sales side for coaching, and then on the marketing side, the market intelligence that comes out of that, and I used to

1634
04:44:32.080 –> 04:44:51.060
Kevin Marasco: you know, 5 years ago, binge, gong calls on 2X to just… to understand what buyers cared about and all that, and now what we’ve done is trained, agents on our sales competencies, what matters for, you know, what our best performing reps look like, what, our, the SPICE methodology that we use.

1635
04:44:51.060 –> 04:45:02.390
Kevin Marasco: And, then taking every single call, you know, 100% of them, and they’re running through the agents, and they’re analyzing and breaking them down and scoring them on spiced efficacy, or Scoville level.

1636
04:45:02.390 –> 04:45:10.080
Kevin Marasco: And then taking the data, automatically filling the CRM for the sales rep, so it saves them time, providing coaching feedback to the rep and the manager, and then

1637
04:45:10.080 –> 04:45:35.049
Kevin Marasco: what I love is getting, like, these market insight briefs that summarize data from all those calls to understand trends in the market and what’s happening and what’s not. Hey, we’re seeing more of these buyer specialty in our world, you know, and here’s some of the emerging pain points, competitive and product insights, and things like that. I didn’t… there’s no way I could have scaled and done that in the past, but having agents do that without fail for 100% of calls is pretty… is pretty

1638
04:45:35.050 –> 04:45:35.870
Kevin Marasco: pretty valuable.

1639
04:45:36.640 –> 04:45:38.110
Carilu Dietrich: Yeah, that’s a huge impact.

1640
04:45:38.410 –> 04:45:40.169
Omer Gotlieb: I’m actually, you know.

1641
04:45:40.340 –> 04:45:50.709
Omer Gotlieb: completely different than my previous company, I’m always thinking how not to hire the next person. Literally, how not to hire the next person. I think in go-to-market, there are many…

1642
04:45:50.840 –> 04:46:05.530
Omer Gotlieb: tools and technologies that you can actually use in order to, you know, do a lot of things that are doing great. And, you know, we’ve developed an internal signal-based system. We identify signals from all places, whether LinkedIn or website or, you know, you know the drill.

1643
04:46:05.530 –> 04:46:12.009
Omer Gotlieb: and eventually come up to a central intelligence system that tells us what to do. Some of the actions are being taken

1644
04:46:12.010 –> 04:46:30.700
Omer Gotlieb: automatically. I’m very careful with that, because eventually, it’s my brand, it’s the company brand, but at least, hey, homie, you need to pay attention to this. Hey, this person just posted, there’s a great article that you actually need to take a look at. Here’s the data that we found out from everything. So, I would say it’s a combination of a set of mini-agents. Some of them

1645
04:46:30.700 –> 04:46:35.110
Omer Gotlieb: Our inside of products, some of them things that we’ve designed internally.

1646
04:46:35.110 –> 04:46:36.780
Omer Gotlieb: But literally help us

1647
04:46:36.880 –> 04:46:42.009
Omer Gotlieb: do a lot of things and go to market that, initially, I would probably need to hire a lot of people to do that.

1648
04:46:43.870 –> 04:46:59.689
Lisa Sharapata: Yeah, I just wanted to add on to what Kevin was saying. I think… I don’t know if it’s called agency, we had something like that called agency. So it’s sitting over the top of the calls, and it’s analyzing all the information, giving the insights to the sellers.

1649
04:46:59.730 –> 04:47:15.419
Lisa Sharapata: like you described, Kevin, but then I’m actually taking that on the marketing side. I’m pulling that into Manus, which I’ve found just to be better at analyzing that kind of information, and then I’ve given it,

1650
04:47:15.510 –> 04:47:21.960
Lisa Sharapata: a brand kit and criteria, tone of voice, everything to help turn that into LinkedIn posts.

1651
04:47:22.010 –> 04:47:35.709
Lisa Sharapata: So, what are the sentiments we’re hearing? What are some themes that are coming up? And then I can actually have, basically, set up the agents from Manus to schedule posts for me.

1652
04:47:35.870 –> 04:47:55.150
Lisa Sharapata: So, it’s… it’s really powerful, and to your point, like, I used to listen to those calls, and oh my gosh, I go down into this pit of despair, I’m like, I can’t listen to any… you know, but now it’s just like, I get these insights, it’s very helpful, and I can get that spun up really quickly into actionable, usable information.

1653
04:47:55.930 –> 04:48:00.089
Kevin Marasco: Yeah, I love that, Lisa. We’re using a tool called Momentum, it’s great, but I love, like.

1654
04:48:00.340 –> 04:48:04.969
Kevin Marasco: you know, being able to get those insights, but then how do we take that full circle? So now we’re able to compress

1655
04:48:04.970 –> 04:48:25.420
Kevin Marasco: our innovation cycles, right, and how we’re able to take that into action on the go-to-market side, and I love your… how you’re immediately topping that up into, content that obviously buyers care about. So I think that’s, that’s… that’s… that’s really neat. There’s so much we could do with that, you know, it closes the loop in a way we weren’t able to do before, or it’d take a lot of… lot of time.

1656
04:48:26.570 –> 04:48:44.359
Carilu Dietrich: And I’m gonna share a slide here. I have a blog at carilou.com called Hypergrowth Leadership, and this was a state of AI and B2B marketing scorecard, which was showing some of the processes that I saw CMOs attacking most, and so we’ve talked about a lot of those today. Content creation.

1657
04:48:44.360 –> 04:48:55.030
Carilu Dietrich: SEO and answer engine optimization, like, certainly this, SDR and BDR, like, these are some of the places where we’ve seen, like, it, it easiest to address some of the challenges.

1658
04:48:55.030 –> 04:49:18.040
Carilu Dietrich: And then… and then some of the others coming farther down the list, the email marketing, the analytics, the ad campaigns, which Lisa was talking about, really, in the experimentation phase, moving into the operational phase. And so, I think, you know, last year was really an experimental year for marketers. This year’s really an operational year for marketers, and even as we’re changing from kind of, like.

1659
04:49:18.150 –> 04:49:42.069
Carilu Dietrich: more simple AI processes to full agents, it seems like, you know, there’s still this prioritization of where it makes the biggest impact, where we can see the biggest ROI immediately, and this has been a helpful chart for some folks. I’m actually in the middle of doing some research to get this more statistically significant, so I’m gonna put in the…

1660
04:49:42.070 –> 04:49:54.559
Carilu Dietrich: a link in the chat here, oops, but I don’t have the link right this second, I will grab it. A link in the chat for the survey that I’m doing to try to determine,

1661
04:49:54.690 –> 04:50:00.079
Carilu Dietrich: like, you know, where are we now, and how are agents affecting this? So, I will put that in in one second.

1662
04:50:00.350 –> 04:50:14.059
Carilu Dietrich: How about, kind of, closing comments? Like, predictions for the future? Are you, bullish that B2B buyers will be buying, in agents shortly, or do you think,

1663
04:50:14.640 –> 04:50:16.280
Carilu Dietrich: That’s still a couple years out.

1664
04:50:17.570 –> 04:50:34.969
Omer Gotlieb: I’ll start… I don’t think that people are going to allow an agent to buy for them, or make the decision for them right now, but they’re definitely starting to see, you know, people are using some kind of, you know, buying agents to take them 90% to the journey. So eventually, you know.

1665
04:50:35.910 –> 04:50:46.200
Omer Gotlieb: imagine you’re an executive and you need to decide on a technology. Basically, what you do is you go to your director, and you say, run a project, analyze everything, come back with your recommendation, and…

1666
04:50:46.400 –> 04:50:49.370
Omer Gotlieb: Let me know who are the top companies that I need to be speaking with.

1667
04:50:49.500 –> 04:51:08.499
Omer Gotlieb: That’s a clear job for an AI agent to actually do those things for you. So I definitely see the future, which, again, the future could be right now, it doesn’t have to be the long-term future, but people are going to rely on their journey and their decision based on AI agents, and I think we as marketers, we have to design

1668
04:51:08.510 –> 04:51:13.400
Omer Gotlieb: for those guys. But I don’t think people will… Let AI buy for them.

1669
04:51:16.760 –> 04:51:35.290
Carilu Dietrich: other predictions for the future, Kevin, you know, your, buying audience is kind of, like, SMB, and so they’re, like, getting pretty deeply engaged in their off time, trying to solve their problems very quickly. It seems like some of your agent responsiveness is, like, creating more velocity in your sales cycle.

1670
04:51:35.290 –> 04:51:37.180
Carilu Dietrich: Like, what’s the next level?

1671
04:51:38.480 –> 04:51:40.209
Kevin Marasco: I think, like.

1672
04:51:40.360 –> 04:51:58.900
Kevin Marasco: I guess my bold prediction… first of all, this is, like, changing everything, right? Like, as us leading go-to- every aspect of marketing and sales, you know, content marketing’s gone. Now it’s, like, model priming. And so, like, all those things change. I think, stepping back more broadly, I guess my bold, prediction is that

1673
04:51:59.030 –> 04:52:11.989
Kevin Marasco: you know, we shift from this, like, world, you know, era of automation and programmatic marketing and predictive AI to generative go-to-market. And how do we think about

1674
04:52:12.030 –> 04:52:30.630
Kevin Marasco: generative agents and systems that are automating and improving all aspects of the go-to-market, starting to break down silos and close loops, and it might even make some of these legacy conversations, things like attribution and stuff, maybe some of those things go out the window and blow up.

1675
04:52:30.710 –> 04:52:36.309
Kevin Marasco: So I think, like, I guess that’d be my bold prediction, is this, like, new world of generative, go-to-market.

1676
04:52:36.500 –> 04:52:41.000
Carilu Dietrich: What’s the difference between generative go-to-market and automated, AI automated?

1677
04:52:41.310 –> 04:53:01.030
Kevin Marasco: I think of, like, automated is, is, is like the, like, like a calculator, right? And you have to be very, like, okay, 2 plus 2 equals 4, 2 plus 2 equals 4, like, that was, like, the marketing automation system, send this email, then this email, then fill out this form, and not, like, deterministic, and so now we’ve gone from the calculator to we have an analyst.

1678
04:53:01.030 –> 04:53:08.429
Kevin Marasco: that can make decisions and do things, and you can say, hey, I want this outcome. I don’t want… just… don’t go get me, you know.

1679
04:53:08.430 –> 04:53:26.950
Kevin Marasco: get this form fill, go find me one of the best buyers, go find me 10 more of those, build the sequences, the personalized landing pages, and, you know, take it full circle. And so it’s like building a more comprehensive system that all works together, versus just one, like, discrete set of tasks.

1680
04:53:27.190 –> 04:53:30.309
Carilu Dietrich: Like, more predictive and responsive, yeah.

1681
04:53:30.950 –> 04:53:32.510
Carilu Dietrich: Anyone else? Lisa?

1682
04:53:34.280 –> 04:53:36.190
Tooba Durraze: I would say, can I… can I jump in really quickly?

1683
04:53:36.190 –> 04:53:36.680
Carilu Dietrich: Go ahead, too.

1684
04:53:36.680 –> 04:53:49.439
Tooba Durraze: and predictive. I think we’re going to move from predictive into more prescriptive, so not so much reliance on a ton of historical data. Instead, the systems that tell you something is going to go wrong before it goes wrong.

1685
04:53:49.440 –> 04:53:56.710
Tooba Durraze: And I think that, in turn, is going to build enough trust that people just, like, let agents loose. I know we feel like we’re further away from that, but I think

1686
04:53:56.730 –> 04:54:06.120
Tooba Durraze: if I was, like, a marketer, a CEO of a company, I would pay close attention to the fact that you can’t… you won’t be able to afford to be left behind in this sort of stuff.

1687
04:54:06.140 –> 04:54:19.230
Tooba Durraze: Right? So the more you kind of shy away from it, or the more guardrails you kind of throw around it within reason, I would say, you will hinder your own development in that sense. I think companies

1688
04:54:19.270 –> 04:54:30.489
Tooba Durraze: will go from, like, everyone needs to be, like, AI-native or AI-powered, all of that, into really understanding what are the systems that are getting me the outcomes that I want, and, like, then applying those systems at scale.

1689
04:54:32.500 –> 04:54:40.309
Carilu Dietrich: It’s pretty exciting. It’s a… it’s an exciting time to be alive. One more answer, and then let’s tell people how to find us, or follow up with us.

1690
04:54:40.680 –> 04:54:41.600
Carilu Dietrich: Lisa, were you?

1691
04:54:41.600 –> 04:55:00.800
Lisa Sharapata: I think that there’ll be a learning curve and a scale. So, for example, like, would I be okay letting an agent rent a car for me if I give it parameters on how much to spend on, you know, here’s the dates I need, this is the size car I want, I don’t want to go over this price. Yeah. But would I want to buy B2B tech?

1692
04:55:01.270 –> 04:55:20.800
Lisa Sharapata: use, you know, with an agent today? No. So, I think that there’s gonna have to be kind of this curve and cycle that’s gonna take time for… and marketing, even in B2B, to figure out how do you serve up an experience that would make it, like, safe to go agent to agent. So, yeah.

1693
04:55:22.230 –> 04:55:46.059
Carilu Dietrich: Well, it’s been such a pleasure talking to you guys. I was nervous about this topic because I didn’t know even what agents were as the distribution channel before I started researching it, and then realized I’m deep in agents, both on the interface and operator side, but now I know that they’re the channel as well. So, thank you for making it so practical and approachable, and let’s just go through, or maybe we can share in our

1694
04:55:46.060 –> 04:56:01.460
Carilu Dietrich: Oh, I see now there were some audience questions that we didn’t answer. Sorry for that. But maybe each of us can share in the chat how to get in touch with our companies and learn more about what we do, because I’m sure the audience

1695
04:56:01.460 –> 04:56:08.599
Carilu Dietrich: would be interested in that for their favorite panelists. Let’s see, let me just look really quickly at this last question.

1696
04:56:12.480 –> 04:56:31.059
Carilu Dietrich: I think the last question in the last 3 minutes, is around, how should marketers start using tools to track AI mentions and citations for queries, and how should they think about creating content, for AI? Marcel, this is, like, right in your crosshairs. Do you want to talk about that one a little bit?

1697
04:56:31.260 –> 04:56:35.259
Marcel Santilli: what we do for customers quite a bit, and I would say, first of all.

1698
04:56:35.540 –> 04:56:59.919
Marcel Santilli: gain a deeper understanding of your customers, and so what that means is, like, understanding, like, their jobs to be done, what questions they have, and then from there, you can map out all the things they care about, both informationally and commercial-related, and then you can start to figure out, are you showing up for those? Do you have the best answer to those questions? And then figure out a way to systematically become the best answer to those questions, and track along the way if that’s having an impact.

1699
04:56:59.920 –> 04:57:21.820
Marcel Santilli: or not on those answers. And so, a lot of the things that matter to humans in becoming the best answer still apply to ranking well and showing up well in those answers. It’s just you have to do more, because the long tail is longer, and a lot of the questions that people might have during your sales process are now becoming, you know, upfront in their research as well.

1700
04:57:23.410 –> 04:57:39.120
Carilu Dietrich: Well, agents being created by some of us on the phone, agents operationalizing our work, and agents talking to agents very quickly here in our future. I just want to thank all of you, and I don’t know if Julia’s here. Let’s see. Hi, Julia. Hi.

1701
04:57:39.570 –> 04:57:46.250
Julia Nimchinski: Incredible session. Thank you so much, Carolyn. And, we have a question from the audience for you.

1702
04:57:46.250 –> 04:57:47.030
Carilu Dietrich: Questions?

1703
04:57:47.030 –> 04:57:57.050
Julia Nimchinski: Yep, people are asking about the CML tenure, and your thoughts on it shrinking, and what’s the implication with agents.

1704
04:57:58.350 –> 04:58:10.260
Carilu Dietrich: It’s a really strange time in AI-native companies. You know, if you look at OpenAI, Cursor, and Anthropic, companies that have grown in AI astronomically quickly with

1705
04:58:10.260 –> 04:58:27.810
Carilu Dietrich: It’s very small marketing teams, and without CMOs for part of that time, certainly in their early days, it seems like the role of CMOs is becoming even more hands-on and technical, and Kevin is one of the CMOs that I admire most, who’s, like, been really deep in building all these orchestrations.

1706
04:58:27.810 –> 04:58:48.999
Carilu Dietrich: So, I think it’s a challenging time for CMOs as heads of growth and people within the org can do more and make a bigger impact, much faster, and it could lead to shorter tenures in some roles, but hopefully strategic CMOs that stay in front of it and, like, hire the right team, and set a vision, will continue to extend their tenures as well.

1707
04:58:49.000 –> 04:58:51.829
Carilu Dietrich: Kevin, do you want to say anything else about that? Oh, yeah, go ahead, Martin.

1708
04:58:51.830 –> 04:58:55.270
Marcel Santilli: One quick thing, just because we talk to so many CEOs.

1709
04:58:55.270 –> 04:59:15.049
Marcel Santilli: And I would say the number one thing is, I talked to a CMO at Zapier and several others, they’re like, I can’t find a CMO that’s hands-on and can be a top-notch practitioner and be in the weeds as well. And so, if you’re a CMO and you’ve been disconnected from the work for years, go change that, because otherwise, I think you’re going to be out of a job. Sorry, to Kevin.

1710
04:59:15.050 –> 04:59:30.199
Kevin Marasco: No, I agree, Marcel. I radically agree. I’ll tell myself the same thing I share with my team. I always tell them, hey, you don’t have to worry about AI taking your job, but you do have to worry about someone who knows AI taking your job. And I think that’s true for CMOs, and I talk to myself, and

1711
04:59:30.200 –> 04:59:39.750
Kevin Marasco: what does that… I mean, I literally block off time for myself to go dark and just play and tinker, and that’s the only way. You gotta, like, roll up the sleeves and do stuff, and so I think…

1712
04:59:39.750 –> 05:00:01.510
Kevin Marasco: I agree, like, we have to get more hands-on and get our hands dirty, and push our whole, not just our teams, but our entire organization, because it’s changing everything, you know, risk tolerance, and speed of execution and, you know, all that, because we’re talking about go-to-market and the shifts there, but it’s also changing all aspects of SaaS, and so I think we gotta be at the tip of the spear on that.

1713
05:00:02.070 –> 05:00:16.429
Omer Gotlieb: I just want to share something final, you know, in my previous company, I went on stages and explained to everybody why I think the chief customer officer job is the most difficult one in the company. And I can tell you, after spending two years in marketing, especially where the market is right now.

1714
05:00:16.540 –> 05:00:28.170
Omer Gotlieb: CMO job, man, that stuff, it is difficult, and you guys are right, you have to do all those things. The future is unclear, but I think that’s the opportunity. The future is exciting, and I think everybody that will

1715
05:00:28.240 –> 05:00:39.150
Omer Gotlieb: align themselves into, let’s be innovative, let’s be hands-on, let’s evaluate, let’s experiment. That’s what gets me excited, and I think this is what will help CMOs in their future careers.

1716
05:00:40.330 –> 05:00:46.919
Julia Nimchinski: Incredible session. Thank you so much, Caroline and everyone, and we are transitioning to our demos. Lisa?