How Marketers Should Really Work With AI Agents, with Eric Zenz, SVP of Product Management at Salesforce | Ep. 437

0:00We are back with another episode of the Market Melanized podcast and I'm live at Salesforce Connections. I'm Daniel Murray and today I'm talking with Eric Zenz, SVP of Product Management for Agent Force Marketing at Salesforce. Eric breaks down how AI agents are tackling the coordination chaos that's plagued marketers forever. Silo data, endless approvals, waiting days to get an answer, using that shared context on Salesforce Customer 360. We get into how the agent force goals and content agents actually build and optimize campaigns for you. Why 75 % of marketers have adopted AI but still aren't seeing the output change.

0:41And the one AI hill Eric will die on. Let's get into it. Welcome to the Marketing Millennials. The No BS Marketing Podcast. I'm Daniel Murray. and join me for unfiltered conversations with the brains behind marketing's coolest companies. The one request I tell our guests, stories or it didn't happen. Get ready to turn the f*** up. Okay, cool. We are back with another episode of the Marketing Millennials podcast. Today, I am here with Eric Zenz, Vice President of Product Management at Salesforce.

1:24Eric Zenz:is that correct yeah uh i lead product for agent force marketing amazing and i i want to just get into the agent force stuff because um i saw a quote today at the keynote that kind of stood out to me and i'll read the quote it said marketers are creative why is coordination crushing us i'm like we have siloed data it dependencies not enough agency and that's not like a real new complain from marketers that it's been there forever why is now the moment it's actually getting solved yeah so uh of course the the challenges as you said aren't new uh the coordination costs uh of of being a marketer and like getting your audience data together getting your content assembled getting your content approved like activating like in a large organization that's like tens of individuals that are involved right there and what we're doing now with AI agents is we're bringing AI agents to assist the marketer in those tasks and even automate some of those tasks to free them to be creative to free them to free their their minds and their time to be able to return to that creative work the sort of essence of marketing and elevate what they're putting out.

2:48Yeah. And I think one of the things, because I used to be a marketing ops guy, and one of the things that really stood out to me was

2:57pulling data really fast. Because I used to get requests from CMO and they would say, why is this channel down? And then I would have to go to talk to the data analysts and like, let's go talk to the data and I'll take it and then I put a ticket in and it'll take three or four days and then I would get it and then we're already halfway to the month behind the month and then we haven't even solved the problem yet and then there's a new month and a new goal. Can you explain what's happening right now if they would use AgentForce to do that and pull data?

3:31Eric Zenz:Right. Yeah. So any LLM is going to be able to sort of leverage the context that it has available to generate a response for your prompt. And what we have in Agent Force Marketing and just Agent Force in general is the shared context of what we call a customer 360 or all of the customer interactions across sales, service, marketing, commerce, etc. So when you're a marketer and you're trying to ask about the performance of a campaign, or if you're trying to determine the segment of individuals to address, you ask that question of age and force in the same way that you would ask of your data specialist.

4:19Eric Zenz:request. We think we have the ability to respond to that request better than anybody else in industry because we have that shared context of your customers and their entire life cycle, as well as all of your campaigns and your campaign performance data over time. So yeah, it's a big shift in being able to just ask those questions of the system, ask those questions of ages for marketing instead of just turning to the human and asking the same question and hoping that that answer would come in a short amount of time. Yeah, that makes total sense because one thing that if I was running marketing ops or marketing would be excited about is I can get that answer and now I can actually go and spin up a campaign to solve that.

5:09And one thing that I also saw today earlier is there is a newer were featured too that can spin up a campaign for you as well, which is pretty cool. Could you explain what's actually happening? Because I know that a lot of marketers have to get into that nitty gritty of tweaking workflows, tweaking audience, tweaking campaigns, and it takes forever to do that in one fell swoop. So the optimization of campaigns.

5:38Eric Zenz:Yeah. So we've in Asian force marketing, we've had several assistive agents that help with parts of the journey of a campaign for a while. We've had a campaign agent for creating the essence of the campaign, the brief, the base content, the base orchestration. And what we announced today were a pair of agents that go beyond just like creating the brief and creating the campaign, but helping you design a goal for your business or helping you define a goal for that campaign or a set of campaigns. then define the campaign that's going to address that goal and then actually orchestrate and optimize that campaign over time.

6:32Eric Zenz:So the key difference between what we had previously in being able to generate those assets and what we have now is the Agent Force Marketing Goals agent that does all of that creation and then the orchestration and optimization. And of course, there's a human in the loop. So the agent recommends changes to the campaign, like either a change in the distribution of channels and how you're spending across SMS or email. It can recommend changes to the content. If it sees content is going stale, it can recommend generating new content. and then not just recommend it, but to take the action to do that for you.

7:18Eric Zenz:The other side of that is we announced a pilot for our AgentForce content agent. I'll throw the word agent in there enough. And the AgentForce content agent is, you can think of it as like a canvas to be able to create multi-channel content around a concept that you've defined in your brief, as well as the variants and even translated variants of that same content. So with the content agent and the goals agent working together, you can use the goals agent to set up the campaign, the content agent to fill it absolutely to the brim with content variants for experimentation. and then the goals agent will carry on to optimize that campaign moving forward.

8:09Eric Zenz:So just really freeing the marketer from having to investigate whether the campaign's performing and then figure out what to do. It suggests an alternative and then just hits the easy button to do it for you. One thing I also found interesting is I saw in the latest Salesforce marketing report that 75 % of markers have adopted AI, but they're still sending off one-off campaigns. They've adopted basically AI, but the output hasn't really changed. What's your gut feeling of why that output hasn't really changed? Yeah, I think it's because there's not a strong connection between a lot of those tools.

8:59Eric Zenz:So you can take a collection of point solutions to generate content or to define an audience segment or even to reason over your marketing data that you've pulled out of the system. But those are optimizations made to moments in the marketing journey. what we're trying to do here is trying to have an agency of AI agents. They're actually all kind of working for the same company in a way off of that shared context so that they can work together to not just make you faster at doing a certain step in the journey, but to make the entire journey faster through either assist or by fully automating the step.

9:52Eric Zenz:So I think in summary, I'd say it's the shared context, which really helps accelerate a marketer's time to getting a campaign out the door and time to moving on to the next. I know I should have probably started up here, but can you give the high-level view of what actually is an agent, like what an agent's doing? Because I feel like when someone says I'm working with an agent, Some people could say, is that agent going to replace my job? Some people are going to say, I'm nervous to use it because my data is put, I don't want to let my data, is it going to go do something crazy? Is there no guardrails?

10:35So what is actually an agent at the highest point?

10:39Eric Zenz:Yeah, so an agent inside of Salesforce and inside the context of AgentForce, it's a set of instructions on top of a large language model governed by what we call trust layer, which ensures things like third-party LLMs aren't getting any smarter off of your data and your customer's data as a marketer, but it's a set of instructions on top of the LLM with that trust layer as well ensuring that the agent performs in a way that maintains the security, the privacy of your data, but also that it acts in a way that is responsible and supportive of your brand. So it's not just the agreements that we have in place with our LLMs.

11:36Eric Zenz:There's a whole set of instructions and code under there that make sure the agent is operating in a trustworthy manner. And so all of the agents that we build operate in this ecosystem. They also have access to the customer data that you're as a marketer, your customer data that's mapped into data 360. So it's when I was talking about shared context earlier, it's that entire context of the customer, your brand, the brand definitions inside of Salesforce, those instructions all together and that data all together supports the agent in making decisions. So every agent's got a set of skills or actions that it's capable of, but it's all on top of that platform, on top of the trust layer, on top of the shared data that's available in Data360.

12:35One thing I want to ask you too, because it sounds like your agent is only as good as the inputs you put inside of the platforms you're using, so Slack. How do you recommend people either clean the existing data to make the agent perform better or to make sure that the inputs are clean to make sure the agent is performing at the highest level? I think sometimes like an agent or ai is only good as right how clean your data is and that you're not doing it in there some people input things in different systems and a weird way that might give you wrong

13:20Eric Zenz:contact or stuff like that yeah i think there's there's two parts of it one is making sure the agent has access to uh the data that's necessary for it to perform its job and for us um as a platform feature, we've got data 360, you're able to connect to a variety of sources, pull that data in, unify it, harmonize it around a notion of a unified individual. That's part of the solution. The other part is that the agent itself and its definition needs to be something that is robust, customizable, and also well-tested. And so we've got a studio inside of Salesforce for being able to customize and edit those agent instructions and a harness to be able to test it as well.

14:09Eric Zenz:The brand name has been paged out of my memory. Maybe I need Slack bot here for an assist, but basically a set of tools to be able to see the agent's instructions, to be able to edit those skills, to be able to test those skills against like real, like with your real data, with new instructions, to iterate on that loop, and then to ultimately like activate that agent for your use. And this goes for any agent, whether it's like an agent that's helping you with segmentation, whether it's an agent that's helping you create content, the agents that would help you with understanding your marketing performance data.

14:57Eric Zenz:The same arc fits all of those agents on the Agent-Enforced platform. I also want to ask you, what do you see as the working relationship between marker and agent? Also in the future, what does a marketing team look like with agents and marketers? What is the human's role versus the agent's role in that picture? Yeah. We're like, I don't know that any of us know how this is going to end. And I don't, it's not something that's going to, I don't think there's an end to it, just a continuum. But the way that we're positioning it is that the agents are, the agents assist your team. The agents assist your team to free you to have more time to be a maker, to be more creative.

15:55Eric Zenz:You think of it as an agency of AI agents that you can hire for those assistive tasks on your team. We want to have a suite of agents for every task that a marketer or every job that a marketer needs to do to be able to assist there. and yeah, like, so I think of it as more an augmentation of your team, not a replacement of any role. And this is also like how we build the software too. Like we're not, we look at AI to assist our developers and our product managers to make better product faster. The last question I have for you is, and I ask everybody in my podcast, but I'm going to switch a little bit because it's about AI.

16:40What's like an AI hill you would die on?

16:43Eric Zenz:Ooh, an AI would die on. Crickets, an AI would die on. Relevant to this discussion, it's that the AI is the response that you're going to get out of any agent is limited by the amount of context that you're able to give it. and you need to have a trustworthy way to get that agent as much possible context as it needs to do its job. So I think people look at the consumer expectations on AI or that it's like magic and just go into chat GPT and bang out a question and it's always going to be right. And like, yeah, it's got broad context, but it doesn't have the depth. It doesn't always have the depth in every subject that you'd expect.

17:33Eric Zenz:and so I guess the hill I would die on is if you want great results out of agents and AI you've got to give it great data I love that because the way I've been seeing at least like Salesforce ecosystem like the conversations you're having in Slack as data the inputs you have in your Salesforce either your data 360 or your CRM or whatever you're putting into, it goes into that context. And I love that it can pull from conversations you had with your marketing team and pull from conversations in your CRM and then give you an answer and see like, oh, you talked about this like a week ago and here's some context that your marketing leader said.

18:22Eric Zenz:And I liked you so much. Even outside of a marketing context, we see this with Slackbot inside of Salesforce where it has our conversations, it has your Google Docs, it has your calendars if you've opted into that and yeah, the richer context really makes it perform as an outstanding member of a team. Well, thank you so much for your time and great chatting with you. Yeah, happy to do it again sometime soon. Thanks so much for listening. Keep tuning in to hear more great insights from the coolest marketers from around the world. If you haven't already, make sure to subscribe and follow the Marketing Millennials podcast on Apple Podcasts, Spotify, YouTube, or wherever you get your podcasts.

19:11And if you like what you hear, I would greatly appreciate you giving us a five-star rating. It helps bring more marketers into our community.

19:27Thank you.