Find the full transcript here:
Jodi (00:01) Hi, Meagen, welcome to the Finite podcast.
Meagen (00:05) Thank you. Happy to be here.
Jodi (00:07) It's great to have you here. You're a busy CMO, so it's been a long time coming, but we have a really exciting episode planned talking all about agentic marketing and agents. I know you're super experienced with agents already, even though it feels like we're still in the early days. So I'm looking forward to hearing more about that.
Before we do, could you tell us a little about your experience in marketing and how you got to be CMO at Samsara?
Meagen (00:39) Yeah, so I've been a CMO at four different companies. I started out at MongoDB as my first CMO role, was there about four years, and took the company public. Then I went over to TripActions, which has since been rebranded to Navan, and was there about four years. I switched into cybersecurity for two years, then made my way to Samsara after Lacework got bought by Fortinet.
So I'm a full-stack CMO. I've seen a lot of different transitions in the space, from social to PLG, marketing through COVID and beyond. And I love what we're doing in the agentic marketing space — I've had more fun in the last year than ever, just getting to learn new technology and lead the team through that change management.
Jodi (01:32) Yeah, exactly. It sounds like you're really adaptable, willing to change as times and technology move forward. Do you feel like that's what set you apart as a CMO, and one of the markers of your success?
Meagen (01:47) I certainly think change management is a big part of it. This is a huge cultural shift for marketing, unlike anything before. If you look at the different channels that have come along — social came about and it was a different way to reach buyers, then PLG came out and it was a different buyer, then COVID changed the venue — but this is a very different way of doing work. We're completely changing our workflows, and figuring out how to go about that just takes a different skill set.
Jodi (02:20) Yeah, and it's almost more of a mindset change than just a channel or strategy change. It's a different kind of philosophy, would you agree?
Meagen (02:30) Yes, it's a totally different way of doing work, and you have to be comfortable really reinventing your role.
Jodi (02:36) And do you think the CMO role is included in that? Have you seen change to your role directly?
Meagen (02:43) Yes. Definitely, who you're hiring matters — you're managing marketers who have to be able to build agents and operate and change what they're working on. So the person you're hiring has got to be adaptable, comfortable using technology, and not only know their function but be prepared for that. As we're hiring, we're asking really different questions. And as we're leading, it's getting the teams to think about not only how they become better and more productive in their function, but how they work cross-functionally — because agents can work across functions if you set them up that way. You need brand to interact with growth, to interact with product marketing and messaging, without needing a lot of humans in between to orchestrate it. They have the capability to orchestrate it themselves. So it's fundamentally looking at how we do things in a very different way.
Jodi (03:42) Absolutely, I think that's a great introduction to what you're doing at Samsara. I'd love to take a step back and get your definition of the difference between agents, agentic marketing, and AI-assisted marketing or just using an LLM.
Meagen (04:07) Yeah, traditionally we'd set up a workflow, set up business logic, and that's automation — you'd automate something, it's business logic, it walks through the same steps and runs the same queries. I think agentic marketing is so different because you're asking it to drive an outcome. What it does in between, it makes decisions — it's got a lot of autonomy, it's got collective memory, and it's doing all that it thinks is the right thing to do to get to that outcome. And the path may be different each time, as it's learning and building; it's not a prescriptive step-by-step process.
That's what allows us to be more agile, and it's why the effects of agentic marketing compound — because of that nature of the agent actually driving towards an outcome and having that sense of thinking.
Jodi (05:05) Yeah, so moving from automation to autonomy, almost — where you're relying on these agents to make decisions on your behalf. That seems like there's a lot of trust involved. Is that the case?
Meagen (05:20) Certainly. I think we put guardrails in place, and it depends on our ability to prompt and be very descriptive about what we're looking for, and then give it the autonomy to do the work in between. There is trust in that. But if we look at humans, humans make mistakes and bad choices too. I think agents — with the ability to learn and do research at such a fast pace, and with such a breadth of knowledge to pull from — the accuracy is going to be much higher.
So yes, there's trust in it, but I think the outcomes we're getting are far superior to if we'd just had humans doing it.
Jodi (06:10) That's a really interesting take — you're right, human error has always been the case. I can't tell you how many follow-up emails I've had saying "sorry, that was a typo." Actually, I've heard that's now used as a strategy, to show you're a human sending the email. But before automation, I was always like, okay, we're all human. But —
Meagen (06:27) You are actually human.
Jodi (06:30) Yeah — it's interesting because you say human errors are so much more frequent and agents are more accurate. Have you ever, in your time with agents so far, had an error that got hyperscaled because it just runs with things?
Meagen (06:57) I'm sure we have. We've been at this for a year, and the team has plenty of learnings, but I don't have one specific example I'd highlight.
Jodi (07:08) Okay, well, that's probably a good thing then — nothing too significant. Could you tell us about your agentic story? How many agents have you onboarded, and what kinds of things are you using them for?
Meagen (07:26) Sure. If you look at all the agents we've built out across the team — through Cowork, through Gumloop — we have over 300. About 200 of those are personal productivity agents that reside on team members' laptops, helping them do their jobs more effectively. Then, looking at the set we manage as a team, it's about 113 active agents running workflows together across functions, with about 37 more in the works.
So we have quite a lot. We've spent a lot of time investing in the team and giving them the skills to build agents, and we've gotten pretty sophisticated — we have multiple MCPs, Databricks as our data platform, and we've been working with Genie on that too, more recently, in the past quarter. We have a lot of different systems and workflows set up where agents are actively working with autonomy, building things, working through the night, making good decisions — hopefully — based on how we've prompted them.
We've got a pretty sophisticated system: over 25 AI-native tools we've bought in the last year, on top of our traditional SaaS marketing stack. We're constantly testing new technologies to see what results we can drive, and what we can do as a marketing organisation to deliver more pipeline and more revenue.
Jodi (09:08) All I can say to that is wow. I don't think we've had a CMO on the show so advanced in agentic marketing. It feels like you're running at full speed and seeing so much success with this. What have results looked like — I'd imagine exponential growth?
Meagen (09:30) Yeah, it's interesting — you're right, you should get something out of it. We're spending money on tokens, building all these agents, and we're either going to do things faster or drive better results — you should get a better top line or a better bottom line, or both. We've had a close partnership with the finance team to really look at how we measure that, and I think there are two good ways to look at it.
One is pipeline generated per dollar spent by marketing, and whether that's improving quarter over quarter, year over year — and we've seen it improve 31% year over year. We're driving that much more pipeline per dollar spent by marketing. The other is pipeline per headcount, and we're getting way more efficient there too — we're 50% more efficient in driving pipeline with the same amount of headcount. That's leading to more growth and more revenue, and I don't think we'd be reaccelerating growth the way we have been without these AI tools and the agents we've built.
One of the tools we've put in place, One Mind, is an agent on our website. She has conversations with prospects and customers as they land on the site — she understands what page they're on, how we've interacted with them before, what industry they're from. The average conversation is two minutes, but what's interesting is that quarter over quarter, the pipeline she's generating for us is doubling. Pipeline was $81,000 in the first quarter, $160,000 in the second quarter, and now it's $331,000 per quarter. As she's getting stronger and faster — and it's interesting, the breakdown is 60% chat and 40% voice, though I think that will shift even more toward voice over time, given how much we already interact by voice on our phones.
Those are just some of the results we're seeing. Another use case is marketing ops — a third of their tickets were questions about leads being routed around SLAs with our ADRs. An ADR would ask about a lead, or why it missed the SLA, and someone would spend an hour or two researching, documenting, and getting back to them. Now we have an agent that does that — built through Gumloop. The ADR goes to Slack, asks a question about a lead, and the agent triages everything that happened, the timestamp, creates a doc to document it, and sends it back — taking the human out of the loop. It's not busy work exactly, it's important work, but it was taking time away from other things marketing ops could be doing, like building more agents and making the team more efficient.
Another one is a reference buddy. Instead of going into Salesforce to track references, a sales rep can ask for a reference for a specific industry or geography, and reference buddy will find it, kick off the approval process with the account owner, track it, follow up, and deliver it — and in the meantime, surface relevant case studies you can use while waiting for approval. Watching the agent and the sales rep go back and forth is mind-blowing, honestly, both the pace and the results.
Jodi (13:08) The whole thing is really mind-blowing. Those sound like incredibly creative ways to use agents, and they really illustrate not just the agent-human relationship, but where agents talk to each other — where one stops and another starts — and how much they can take off your team's hands. I find that story about the tripling pipeline from the first agent you mentioned fascinating. Do you think that's down to the agent itself learning, or the prompting and maintenance from your team getting better at driving results through it?
Meagen (13:50) I think it's both. Our website is improving — we're learning quickly, iterating on our messaging because we have better customer and competitive research for positioning, so the website is getting a lot more effective. The agent is learning too — it has this collective memory, collecting information on the person coming to the website, maybe from Gong calls, maybe from campaigns they've already interacted with. We're also researching on Perplexity to find out more about the company. So the answers are more relevant to the person asking, and we're getting them through their decision-making process much faster.
You can't scale humans in chat like that 24 hours a day, but with AI you can scale to infinity — as many conversations as you want. So yes, the systems are getting smarter, the website is getting smarter, and we can get people through the buying decision process much faster.
Jodi (14:53) Do you find those agent-to-customer or agent-to-prospect interactions are well received on the other side? Have you noticed more engagement, or is it more that you're reaching a wider pool of people and statistically getting more engagement as a result? How do you feel about the quality of the interaction?
Meagen (15:21) I think the quality's great, or we'd stop doing it. The answers are accurate, and people get the information they need. We don't have patience anymore as humans — we want answers right away, versus the old-fashioned way of signing up, waiting for a rep to reach out, scheduling a meeting, and getting a demo. That all has lag. We can still absolutely meet with you and set up the demo, but you can get answers in real time without having to wait.
It's compressing timelines for sales and getting people through the gates they need to get through to make a decision. And it's a competitive advantage — if you have a choice between a partner giving you great, quick, accurate service, and one taking the old-fashioned route to get back to you, you're going to go with the one offering the best service and response.
Jodi (16:21) That highlights an interesting point — we've been moving away from the sales-led buyer journey for years now, toward a user-led journey where the buyer drives their own research, and that's only accelerating with AI. Interacting with an automated tool feels like an extension of that: buyers are already used to driving their own journey, so by the time they get to sales and want that human touch, it's expected in a different way. I can see why that's driving success for you.
Meagen (17:08) Yes — and we're also making our sales reps far more effective. They have all the information at their fingertips: during a demo, if they get a question, the agent is listening and prompting them with the right positioning, messaging and answers. So they're becoming superhuman too, with the agent as a kind of wingman helping them serve the customer effectively.
Jodi (17:34) Definitely — the sales role is being incredibly enabled. Let's talk about the marketing roles on your team. You mentioned they're building agents, maintaining them, and making the team more efficient. What kind of roles are you hiring for now, to plug the gaps between agents? And what does your team do day to day?
Meagen (18:04) When we're interviewing, we're still interviewing for functional excellence — if you're a designer or creative, we want to know you've got a background in design, whether that's paid search, deck work or video. But we also want to know you've embraced these tools, because the expectation of output and what you can deliver is here now. We want people who have already built agents, and we expect you to build them and operate them once you're here — we also train everyone on that. We have a whole training programme where everyone in marketing learns how to build and work with agents and transform their roles. But I want to know they're continual learners, because every day there's a new tool or something else coming out. That's the expectation.
Jodi (19:00) And day to day, do you let your team not only build but also sign off on these agents, test and learn with their own tools? What's the flexibility like for experimentation and making mistakes, and who signs off on all of it? Do you have oversight across all the agents running, to make sure everything works cohesively?
Meagen (19:35) Everyone is empowered to build, use and try things, but the human element is that they have to bring judgement and taste. Whatever they're building, they need to be monitoring and observing it. We also have a central team that oversees all the agents and puts the right guardrails in place. Anything going external to the company still goes through a full review cycle — if an agent builds an email, that email doesn't just go out; it goes through a review process.
You can generate the email faster and more accurately — we actually have an agent that reviews all emails before they go to the team for final approval, and it gives a lot of feedback and catches a lot of things. So there's an approval agent, and that step is very fast, then it's reviewed by the team — and that review cycle has got faster now that agents are doing some of the reviewing.
We also have agents checking other agents for accuracy, including checking for AI slop. Because once you start adding all these prompts and rules — "I need our tone of voice," "I need the right branding," "position and message it this way" — you can end up with this jumbo email nobody wants to read. It checks all the boxes, but it's not really human language, it's too much, it misses the point. So we always have to apply human taste and judgement — that's the important part humans and marketers in the loop bring.
Jodi (21:20) Absolutely, and marketers are probably more attuned to the audience than an agent can ever be — audiences are always changing what they expect and what's trending. It's funny you mention an agent doing that first round of review, because I read about a tech company — I think it's called System1 — that's released a product predicting how successful an ad creative will be in market, because it's trained to learn what an audience responds to. It's almost learnt taste, in a way, through data. It'll be interesting to see how that relationship evolves, and how much final review your reviewers really need.
Meagen (22:15) Yeah, it's interesting. I recently met one of the first growth marketers who was at Cursor, now at SpaceX AI. He built an agent for the social media side of marketing — specific to Instagram — that looks at their influencer set, which ads are effective, what the call to action and trigger were, and it quickly learns and generates what the company should be building next, based on what's resonating in the market through those channels. I found that fascinating — one agent picking up the learnings, and another agent taking those learnings and building the ads with that in mind. It's impressive what he's built out already.
Jodi (23:02) These kinds of stories really hit home how, if you're not leveraging these tools and figuring out how to work with them, you're going to get left behind — while some companies become category leaders in their space just by embracing them and seeing how much autonomous thinking can bring to their team. It's amazing.
Meagen (23:33) Yeah, I've thought a lot about that. I wouldn't want to compete against us in the market — in the last year we've built out over 113 agents, with 37 on the way, plus another 200 productivity agents. We're a fierce competitor, working 24/7. That doesn't mean we're spamming the market — our agents are learning about our prospects, learning the right messaging that fits, they know our customers inside and out, what products they have and still need, and the right messaging for that. They're tailoring and creating emails, learning from responses and open rates, perfecting direct mail. It's hard to compete against a team that can scale infinitely with agents working on their behalf, even while they sleep.
I've seen the impact on the market of giving these capabilities to everyone in marketing and aligning on the outcomes we're driving toward. If you haven't been working with this stuff — because we're learning every day, and we've been doing that for a year, and that knowledge compounds over time — it's going to be tough to compete. You're essentially down a few thousand men, so to speak.
Jodi (25:06) Literally! I think we spoke a couple of months ago and you had 90 AI agents — so it feels like it's roughly trebled since then.
Meagen (25:17) It has. And once your team figures out how to build these agents and sees the power of it — they can build a dashboard that monitors for anomalies and flags things, like "I need to get in front of this person, it's an ABM account," and it comes back with the most effective ways to do that and arms you with the email or the outreach — that unlocks something. You think, I can totally reimagine this, or have that running while I work on something else. It's a huge unlock.
We actually vibe-coded an LLM to train the team on how to build agents, what it all means, how to work with Cowork and these different tools. We were fortunate — we got ChatGPT right when it came out, our CEO is very forward-thinking on AI, then we got Gemini and NotebookLM, then Claude. We had to quickly teach the team, because people get a little nervous and uncomfortable at first, and we needed to get them hands-on. We ran AI power hours, but it wasn't until we ran a two-week training programme — a lot of online learning and reading, then coming into the classroom to actually build agents, and presenting what your agent does at the end — that it really landed. People came out of that saying, "that was actually really easy, I had no idea how powerful it was, I can go and do this in all these other areas."
That's when they start to compress their timelines and their output goes up — our content team went from a few things a week to about ten things a day. That kind of power unlock means each function has become much more powerful in its output and outcomes. Now we're working on cross-collaboration — how do we unlock and really change the workflows across functions?
Because we're in the trucking automation business, here's an example. Traditionally, if a fleet is out on the road and the engine light comes on, you call it in, schedule it, come in, order the part, and wait. The new way is that agents automatically see the engine failure, schedule the workload, get the mechanic on site, reroute the truck, and the parts are already ordered — all without humans in the loop.
We're now applying that same thinking to marketing. Do I really need to come in on a Monday morning, look at a dashboard, and see we're a little behind in Mexico, then ask the team what channels will boost pipeline, decide to do more paid search or unlock some direct mail, and shift budget around? Or can the agent see we're starting to dip in Mexico and release budget, increase paid search, get the creative team — or just get an agent — to create more ads specific to what's working in Mexico, deploy them, let sales know, and send the direct mail? All of that can happen without as many meetings, discussions and analysis. At a bare minimum, you can walk in and be told exactly what needs to happen, and go execute on it.
So the job is a completely different workflow. You need to come in as someone in marketing thinking differently about how you interact with other teams, and build out those functions — because for agents to do this, they need to know the jobs to be done by each team. Once those jobs to be done are created, activated and designed in, the agents can interact with each other as they work out what needs to get done.
It's been really cool to see the team figure out what they can set running, and then check in on the outcomes. Yes, there have been mistakes and learnings along the way — I've had a few of my own with email, where the logic sent to the wrong list instead of the right one and everyone got the email. So yes, we'll have mistakes, but we're learning fast. And our customers get it — they love what we're doing and experimenting with, the level of service is so much better than before, and they welcome it. They see our fast learnings and want us to transform their business too, making maintenance and preventative maintenance seamless, because of the pace at which we're learning and the amount of data we have to learn from.
Jodi (30:30) I think this episode is going to be a massive competitive advantage for everyone listening, and it sounds like a genuinely fun time to be on your team — lots of teamwork, support and space to learn. Because, as you say, you only started a year ago, so nobody quite knows the limits of how these tools should be used yet. Thinking creatively and stretching those limits as much as possible is brilliant — and now it's embedded in your product too, so your customers can expect it throughout.
Meagen (31:10) I agree, and I think the team's having fun — I love seeing the unlocks, and I love when they share what they've built. It's definitely a fun time to be in marketing and in tech.
Jodi (31:22) Absolutely, and I think that's a great point to end on. I have a million more questions, but maybe we'll have to do another one sometime. Thank you so much, Meagen, for coming on — it's been an absolute pleasure to hear from you.
Meagen (31:36) Thank you for having me. I appreciate it.
