In-Ear Insights: Live from MAICON 2026

In this episode of In-Ear Insights, the Trust Insights podcast, Katie and Chris discuss their takeaways from MAICON 2026, including Katie’s sold-out Claude for Business workshop and the shift from AI basics to security, governance, and optimization. You’ll learn why most people can’t calculate the ROI of AI—and how to fix it by measuring a baseline before you start. You’ll discover what makes AI truly agentic, using memorable examples like the “go get milk” scenario and a Waymo backseat test. You’ll also see how the 5P Framework helps you define goals, guardrails, and a clear definition of done so AI agents do the work you actually want. Finally, you’ll explore why critical thinking—not the technology—is the real challenge as AI races ahead of its users.

00:00 – Introduction
01:30 – Katie’s Claude for Business workshop recap
04:00 – Security, governance, and shadow AI concerns
06:00 – The ROI problem: you never measured a baseline
10:00 – Optimization vs. innovation ROI
13:00 – Three session themes at MAICON
15:00 – What makes AI truly agentic
19:00 – The milk run: why specificity saves agents
23:00 – The 5P Framework and definitions of done
27:00 – Model choice, effort levels, and surprise results
31:00 – Closing keynote takeaways and critical thinking
36:00 – Call to action

Watch this episode to learn how to think clearly about AI before your agents do the thinking for you.

Watch the video here:

Watch this video on YouTube

Can’t see anything? Watch it on YouTube here.

Listen to the audio here:

Download the MP3 audio here.

[podcastsponsor]

Machine-Generated Transcript

What follows is an AI-generated transcript. The transcript may contain errors and is not a substitute for listening to the episode.

Christopher S. Penn: In this week’s In-Ear Insights, we are at the Marketing AI Conference, 2026, MAICON, here in lovely Cleveland, put on by the SmartX community. Katie, quick take first: what’s been your takeaways from the event so far? You did a fantastic sold out, packed, wall to wall, standing room only workshop on Claude for Business yesterday. How’d it go?

Katie Robbert: I think it actually went really well. It’s hard for me to judge how well it goes when you’re the one who’s up on stage and you’re looking out at the sea of people. Because from my point of view, everybody had those blank faces. I couldn’t tell if they were truly locked in, or if they were confused, or if they were thinking about what they were going to have for dinner. I honestly don’t know. I was also thinking about what I was going to have for dinner, but I was locked in.

The feedback that I got was that it went really well, because the way that I tried to approach teaching the workshop was twofold. And this is what I’ve been telling people. The question I’m getting the most is, how do you address all of the different levels of people in the room in terms of their abilities and capabilities with AI? That’s one of the things that’s really interesting about coming to an event like MAICON—you can be at any level and still get something from the conference. And so that was the goal of the workshop.

The way that I approached teaching Claude for Business was thinking about, okay, everybody is at all of these different levels. How can I teach the foundations as well as an advanced version of it? People who were newer to the platform itself got all of the foundations that would help them get set up and get started. People who already knew the foundations were getting that refresher, but they were also getting layered in their ways to measure the work that they’re doing.

Because the big question that we’re all wrestling with in 2027 is, what is the ROI of my AI? How am I measuring that I’m successful? So we spent a lot of time talking about that. The other thing that kept coming up—and I think this is a topic people are talking about here at MAICON this year—is the security and governance. Security from a standpoint of how are we ensuring that people aren’t leaking data, or that they’re not using shadow AI, or that they’re not putting PII into these large language models that we don’t own.

The other question is the governance around: okay, so we have everybody using a large language model, and nobody’s using the same skill. How do we get everybody using the same version of something we’re building? How do we get only one person building the thing? But we want everybody using AI. So those were a lot of the questions that came up.

I think the big themes moving into 2027—we’ve moved past the 101. It’s not “how do I use it” now. It’s how do I secure it, how do I govern it, and how do I optimize it?

Christopher S. Penn: Let’s talk more about the ROI part, because that is everyone’s favorite term, and for most people, you can find their definition of ROI with a doctor and a flashlight. It’s return on investment. You have to make money, then you spend money, and you calculate earn minus spend, divide by spend. When you’re talking to people in the workshop, how are people even computing “here is what we earned with AI”?

Katie Robbert: The big thing that I’m seeing—and we’ve seen this across the industry—is that people will open up an AI tool, start working, and say, this is how long it took me. Why can’t I calculate my ROI? You want to guess the answer?

Christopher S. Penn: Because you didn’t measure in the first place.

Katie Robbert: Because you didn’t measure in the first place. You don’t have a baseline for what you were doing before. Calculating return on investment isn’t actually a complicated thing. But now, with the introduction of generative AI, people are overthinking it, overcomplicating it, and forgetting the basics of: I actually have to take a baseline of what I was doing before. What was the before state, so I can compare it to the after state? So I can do that simple calculation to say, did it improve or did it worsen? And so that’s what I’m seeing.

So it’s not a matter of how do we measure what we’re doing. It’s, well, did you measure what you were doing before? Because you can measure what you’re doing—they’re making it easier and easier. I was actually catching up with our good friend Andy Crestodina. In his session later on today, he’s going to be showing how he used Claude to create a build receipt, which is going to walk him through how much time it took to do the thing. So there are ways in each of these large language models to calculate how long things took. That’s amazing. But if you don’t know how long it took prior to bringing it into the AI tool, it doesn’t matter.

Christopher S. Penn: How do you reconcile that? That’s optimization—do things bigger, better, faster, cheaper, but you’re still doing the same thing. How do you reconcile that with the transformative effects of AI, where you say, here’s something we’re doing now that we never did before, and there’s new revenue? For example, Trust Insights used to be all about Google Analytics audits and Tag Manager audits and stuff. And now we’re doing AI enablement and AI transformation and stuff. Those are products that—our entire Academy, Trust Insights Academy—never would have had these courses, but now, because we’re in this space, we now have extra revenue. How do you account for that in the ROI calculations, rather than just bigger, better, faster, cheaper?

Katie Robbert: Well, instead of optimization, it’s more of that innovation. If it didn’t exist before, you don’t have a baseline. But at the same time, you do—your baseline is that it didn’t exist before. So if you’re saying, how do we measure the impact that the Academy had on the company because it didn’t exist before? Well, what did it look like before we had the Academy? What does it look like now that we have our own Academy?

When we talk about the return on investment, again, we’re overcomplicating it, we’re overthinking it. Really, you can break it down very simplistically. People are just looking for: is it better, is it worse? And then they want to dig into what is that percentage, roughly. Is it 10% better? Is it 10% worse? Okay, that’s not too bad. Is it 50% better? Is it 50% worse? Okay, that’s not great.

So if you take that example of building our Academy—what did it look like before? This was our revenue. If suddenly we have the Academy and our revenue has dipped by 50%, we’re not getting a great return on investment. We’re actually spending more to build these courses that nobody cares about, and then we’re eating into our own profits because we’re building courses that nobody cares about. So that’s a really bad ROI.

I tend to think about it this way: I start simplistic, and then I can layer in more details as I have them. But trying to layer in all those details before you even have a good sense of what’s good and what’s bad is where people are kind of getting it wrong.

Christopher S. Penn: For those who are using AI tools, this would be opportunity cost analysis, which your AI tools are perfectly capable of doing on your behalf. The other thing that’s interesting is, on the drive in, I was looking at all the different sessions, and there are sort of three different buckets. There’s the governance and the AI operations. There’s a huge section of sessions that are all about AI visibility and GEO and stuff like that—we just got a world roundup session, which is hilarious and fun. And then there’s this third cluster that is all about agentic AI, and I feel like there is absolutely no agreement among any of the sessions about what the hell agentic AI even is—the different levels of agentic AI and what’s being taught here. So in those three clusters, what are you looking forward to learning?

Katie Robbert: I mean, I’m an operations person at heart, so I’m always looking for that. It’s easy for me to say, I’m looking to operate better. I want to operate more efficiently with my vocabulary and with my AI. But really—and I do operations so much—I’m actually really interested in learning more about how people are approaching agentic AI, because to your point, there’s not a whole lot of agreement. What is being called agentic AI isn’t necessarily agentic AI, but sometimes it is.

I feel like, again, the goal for some people is to just get more things done, but we’re calling it fancy things. So could you take a minute and give a rough definition of what truly agentic AI is, versus what you’re kind of seeing being called agentic AI?

Christopher S. Penn: A lot of people call things like workflows or automations agents. Here—agency and agent have the same etymology. Something has agency. Your dog, Georgia, has agency. When she wants to go out, boy, does she let you know it’s time to go out. You do not have to prompt Georgia, “we should go out.” Also, sometimes you do, but she will let you know in no uncertain terms it’s time for you to do something for her. That is agency. She can operate independently of you and do things.

When you have a real estate agent, your real estate agent does not live with you—if you do, that’s weird. If you go on vacation, your travel agent goes with you—that’s really weird, unless you’re married to the travel agent. These are agents that do things on your behalf. They have their own agency, and then when they come back, they’ve done the thing.

When we’re talking about agentic AI, it should be something that has a fair degree of autonomy and agency. So when you say, Claude Coworker, Claude Code, or Hermes Agent or MetaMuse, “go do this thing, tell me when it’s done and I’ll inspect your work product”—that is an agent. If you are driving the car still, it is not an agent. If you’re in a Waymo and you’re sitting in the backseat playing Candy Crush, that is an agent, because you are not doing the work.

Katie Robbert: Let me ask you this question. One of the things that we spend time teaching—and I taught this in the Claude workshop yesterday—is around deep research. So I can give the CASA framework by Trust Insights. I can say, here’s all the things that I want to do within the deep research, here’s my set of instructions. You, Claude, go do that research for me and bring it back. What of that is agentic?

Christopher S. Penn: All of that. Because you’re not doing the deep research. You are effectively acting as a manager, and you are delegating to the agent. Imagine that was a research agency that you had hired. You would say, hey, research agency, here’s your brief, go. That is an agent. If the agent kept calling and saying, “well, Katie, what about this? Well, Katie, what about—” I could do that myself. I don’t need to hire an agency to have you ask me constant questions constantly. I gave you a brief. You’re smart, you got resources. Go and do the work and call me when you’re done.

Katie Robbert: But I feel like that’s confusing, because to me—I’m looking for a little bit of clarity—so I can give it a prompt and it will give me back an output. Basically, you’re saying that is agentic. That is how people are using it now. So, like, “write me a blog post”—that’s agentic?

Christopher S. Penn: The entire task is handed off, and it’s a fairly large, long running task. I was watching somebody in a session earlier today—because their screen was open and they didn’t have a screen protector, which you might want to have at conferences—and they were doing exactly that, going back and forth. “Claude, no, make it sound like this. No, make it do this.” That is not agentic. You are still steering. If you had said, “here’s the brief, here’s our writing style guide, here’s this, here’s the research—Claude, go produce me a blog post and I’m going to go do something else, and I don’t have to think about whether it’s going to do it, because I gave it the right context”—that is a more agentic use. Even though it’s not an agent, like a piece of software, you are using it in a manner that is giving the tool agency.

Katie Robbert: I see. So a lot of the sessions here at MAICON 2026 this year are around building agents or deploying agents in your organization. What are some of the cautions you would give to people who are walking away from these sessions—which, admittedly, are only 45 minutes and can only give you a slice of what’s possible—and then they go back and they say, “I can build an agent.” What are some of the things you should say? Yes, you can build an agent, but make sure you’re watching out for these things.

Christopher S. Penn: Agents have to have three vectors. Do they follow instructions? Do they produce a good output? And do they do it efficiently? And then I guess the fourth meta one is, did they do it safely? So: is it done correctly, is it done at all, is it done efficiently, and is it done safely? It’s kind of like the whole bigger, better, faster, cheaper.

This is where a lot of the work that you do on governance and oversight is so important. Because you can say, hey, go save me some money. You open up MetaMuse or Grokbot, say, hey, I’m paying too much for insurance, go and save me money. And MetaMuse will log into your insurance portal, it will look at your contracts, it’ll say, okay, I’m going to cancel your current insurance, I’m going to sign you up for this one, I’m going to sign all the documents for you. Congratulations, I saved you money. And you’re like, that’s completely not what I wanted you to do. But you were insufficiently specific. It inferred: hey, I have agency, you gave me permission and you gave me an unclear goal, I’m going to do it. At no point did we say, you know what, before you sign—actually, never mind, you’re not allowed to sign anything.

Katie Robbert: Well, and I think that this is going to be a hard lesson learned for a lot of people. The example that I like to give when we’re talking about this is “go to the store and get some milk.” I’m going to tell this story again because I feel like it’s a really good, clear, simple example.

Kris, if I say to you, “can you go to the store and get some milk?”—that’s all I have to say to you, because you can then infer the rest and say, oh, Katie wants me to go to the store, go to the case where the milk is contained, open the case, grab the milk, go to the cash register, legally pay for it, then bring it back somewhere and put it in the office refrigerator. I didn’t have to say all of those steps. You understood what I meant when I said, Kris, go to the store and get some milk.

If I say to an agentic system, “hey, go to the store and get some milk,” a couple of things are going to happen. The odds of it going to the store, getting the milk, paying for it legally, bringing it back to wherever I am, and putting it in the fridge is very low, because I wasn’t specific. What is likely going to happen is the agent will go to the store, open the container, get the milk, and stand there, because it did what I asked—it got the milk. Or it’s going to say, well, I figured since you’re home, you probably want it back at your house, so I went to the store, I got some milk, and I brought it here. Oh, so you committed theft. Got it. Because I never specified that you also have to pay for it.

Those are the things that I think people are going to learn the hard way—hopefully once, because we all have to make those mistakes and learn from them. But hopefully, as they’re starting to work more with these agentic tools and autonomous agents, they’re going to learn that specificity is key. It’s the same with humans. If you have people that you’re delegating to, you don’t have the right to get frustrated with them. You’re going to, but you shouldn’t get frustrated with them. If you say, “build me a report” and they come back to you with a report and you’re like, “but that’s not what I wanted,” and they’re like, “but you didn’t say what you wanted, so I built you a report”—okay, but that’s on me, because I didn’t tell you what I wanted in that report. It’s very much the same way.

So if you’re good at delegating specificity to a human, you should be okay with an agent. If working with a team and delegation is not one of your strong suits, that’s someplace that you might want to start before you get to an agent.

Christopher S. Penn: And in agentic AI, the definition of done is the most important thing. If you don’t have a clear definition of done, it’s going to go very badly. So we have a framework for this. The framework is called the 5P Framework by Trust Insights. Katie, what is the 5P Framework by Trust Insights?

Katie Robbert: The 5P Framework by Trust Insights is purpose, people, process, platform, performance. Purpose: what is the question we’re trying to answer? What is the problem we’re trying to solve? People: who is this for? Who is going to use this? Process: how are we doing this? Platform: what tools are we using? And performance: did we answer the question? In this case, what is the definition of done?

That is a very helpful framework, because it’s simplistic in nature, but it covers any possible scenario that you might have forgotten. Because you have your clear goal, you have who it’s for, you have how it’s meant to be done, you have the tools that it’s going to use, and you have your definition of done.

Christopher S. Penn: And in the process one, in particular for agentic AI, you have to spell out “here is what not to do.” Go to the store—do not bring a gun and take the milk by force from the store. You should probably say that if there is a reasonable probability that the agent will mistake it in its processing, because you didn’t give a clear definition of done. You might have to say the definition of done—”and you adhere to all federal, state and local laws” as part of the definition of done. It sounds ridiculous to us, but these tools—we assume too much about the knowledge they have. Agentic model makers are putting less knowledge in the current generation models, not more, because they’re sacrificing agentic knowledge for the ability to do things.

So if you give it an unclear something that does not follow the 5P Framework by Trust Insights, it may very well say, well, you know what, it’s perfectly valid to bring a gun to the store—guns are legal—and I’m going to take the milk. And you’re like, no, that’s not how that works.

Katie Robbert: So I had a really interesting visual yesterday during my presentation. I was trying to illustrate the point of how much a large language model knows versus what you, the human, know. I have these side by side images of two humans, both with crazy white hair. On the left side was “what you think your AI model knows”—on the right side is what it actually knows. On the left side was a picture of Einstein. We all think that these agentic models are so smart, they know more than us. On the right hand side is how they actually behave—and it was a picture of Gary Busey.

I brought it up as a visual because I felt like it was compelling to show people: you’re treating these models, you’re giving them too much credit for what they know. But really what’s happening is they’re like, okay, I’m gonna bring a gun to the grocery store and get some milk and just stand there and shoot ’em up. I know, to your point, it does sound like a ridiculous example. But you have to assume that with these agents, because they take things so literally and it’s very black and white, they’re going to do exactly what it is that you say.

So I guess my question for you—I mean, it’s well documented, I’m someone who likes to play out the scenarios, think of the risk management of things—but how does someone who doesn’t think that way start to approach “I need to work with an agent”? How am I supposed to know what’s going to go wrong? How do you plan for that? Because it could be an infinite amount of things. You’re saying this one extreme example, but that’s just one of many things that could go sideways.

Christopher S. Penn: Today’s agentic models are intense reasoning models that are very good at goals. You give them a clearly defined goal. Go back to SMART goals back in the day—simple, measurable, action oriented, something, time, whatever, achievable, realistic, something—I don’t remember anyway.

Katie Robbert: You were never good at following goals.

Christopher S. Penn: I was never good at following goals. But that goal setting—a model can reason from that and say, okay, well, this is the goal, so I’m going to figure out what will get me to that goal as an optimum path. So if you were to take the 5P Framework by Trust Insights and distill down two parts: purpose and performance. Here is why we’re doing the thing, here’s what we’re doing, and then here is how you know you are done. This is where we want to go. The model can infer a lot along the way.

If you do not have a good definition of done—if you do not have performance—the model can drift and say, okay, I don’t know what the goal is, so I’m going to guess based on things I’ve been trained on. Maybe you want this, instead of you saying “this is done: it must use this font, it must use this style guide, it must have this.” If you have that definition of done, the model will converge on it very quickly, and you’ll have far fewer things go wrong. So you will only have to anticipate the edge cases after you try it out a few times. And like software development, you want to try it out a few times before you put it into production. Once you see that and you start to know the model’s quirks—oh, this is what you do wrong a lot—you’re going to now include that in the process part of the 5P Framework by Trust Insights, saying, don’t do this, do this instead.

Katie Robbert: My specific Claude instance really liked to use British spellings instead of US spellings. So that was something that was flagged early on. But one of the things that was really interesting in the Claude for Business workshop at MAICON yesterday was everybody was given the same set of instructions and everybody got different results. A few people, as we were going through a deep research exercise—despite all of the prompts being the same—Claude went ahead and built artifacts based on the deep research, because you didn’t tell it not to. So I’m curious to get your perspective on why. I mean, we know every large language model is going to be different—they’re all tuned differently based on the individuals. But there was a handful of people where it even took those extra steps. So 90% of the people sitting in the room, they got their deep research report. Every deep research report looked different, despite the definition of done. And then a few people got those little extras.

Christopher S. Penn: What model were they using—the people who got different results?

Katie Robbert: And that’s a great question. At the start of the workshop, I had advised that everybody be using Sonnet for the work we were doing, because it wasn’t going to be heavy duty thinking—it was just all task oriented. I can’t say for certain—I didn’t walk around to see what everybody had set. That was my recommendation. So that’s a great question, I don’t know the answer.

Christopher S. Penn: And this follow up question would be: what level of effort was it set on? Because what we always say—and I’ll be saying this tomorrow in my session on AI analytics—is the models are the brains. Haiku is not as smart as Sonnet, which is not as smart as Opus. And it’s Luna, Soul, and Astra in the chat world.

But then there’s effort—how long it’s allowed to talk. The higher the effort, the more it talks. If the effort is cranked up to max, it can sometimes go and do those extra steps you didn’t ask it to do, because it has room to say, well, actually, you’re asking me for a report and I’m going to guess how you’re going to use this report. If you crank the effort down below, it’s like, here’s your report.

Katie Robbert: What I had advised in the workshop at the beginning was that people use Sonnet at a medium level. I figured that was a pretty good place, so that they weren’t going to be burning through a lot of tokens, but they would still be getting a reasonable result. Again, this was all workshop examples. I wasn’t expecting them to get perfect things the first time. Really, the goal of the workshop was to get them used to doing the exercises so that they could build the foundation.

What was interesting about the workshop—and I haven’t heard feedback on this—is that the first third of the workshop, we didn’t touch the AI tool at all. Because as we’re talking about that question of how do I measure the ROI, a lot of it has to happen outside of the tool. And it’s the same with software development. This is a core part of the software development lifecycle. The SDLC is spending time doing those requirements up front. That’s the same with AI now.

So as we’re saying, “well, I’m trying to measure the ROI”—usage is not the measurement. “I just want to check the box”—but what did you do? I spent a good first third of the workshop teaching all of those foundational tools so that people could make the decision of, do I even need to bring this into AI in the first place? AI might not be the thing that’s going to get this done. It could be an app, a script, a macro, whatever—or it’s something that the human should be doing and not even touching technology, because there’s no place for it.

That’s a great way to start thinking about your measurement. Because if you’re just saying, well, I have to use AI, let me check the box, you’re going to get those really bad, unhelpful results. And that’s when you start to get in those dangerous scenarios of AI going sideways, because you’re like, I just have to use it—it doesn’t matter what I’m doing. But then you’re like, oh, but now all my data is on the Internet. Oopsies.

Christopher S. Penn: I want to close out by asking you a perspective from Paul Roetzer’s keynote this morning. He was talking about how 2026 is the agentic era, and 2027 is when we get to AGI, when these tools do have agency and can one shot as well as possible. Ethan Mollick calls it the jagged edge of the jagged frontier. You now have consumers with what we call level 4 autonomous systems—MetaMuse, Grokbot, OpenAI Agents. And yet, when you look around people’s laptops at what they’re doing in these tools, they’re still using them as though it was 2023 and 2020. How are people now supposed to keep up with the fact that the technology has raced so far ahead that the consumer has better tools than the marketer does? Very few people know what they’re doing with these tools. And so in the MetaMuse case, it does very dangerous things because it thinks it’s being helpful. How do we help the people here at MAICON, and outside in our community, catch up to the technology?

Katie Robbert: Technology was never the problem. This is a very human problem. The thing that I’m going to continue to double and triple down on this year and into next year is helping people use their critical thinking skills, helping them develop that emotional intelligence, helping them really think through, should I even be doing this? That common sense, that judgment—because AI makes it so easy to just use. And the next thing we know, our whole life is wrapped up in these systems, and we’re like, I forgot how to even do the basics, because AI is doing it for me.

I really want to help and challenge people to think, do I even need to be using AI for this? So I’m not saying don’t use AI. I’m saying let’s be more thoughtful about when we are using it. It can do a lot of things. It doesn’t mean it should. This is where, on our podcast, you would pull up the Jurassic Park clip of “just because you can doesn’t mean you should.”

And I really want to encourage people to be thinking about: there’s a lot of really cool stuff that we can do—how much of this is human? How much of this is with the technology? This is something that you and I do for a lot of our clients. We run it through our TRIPS framework, which is time, repetitiveness, importance, pain, and sufficient data. The output you get from that is an analysis of what tasks are really well suited for AI and what tasks should really stay with the humans, and where that intersection is.

More people need to be thinking about it that way, because humans still want to do things. Stop taking away the things that the humans want to do because they’re passionate about it. So moving into 2027, as we’re thinking about how do we educate—my education is going to be around critical thinking. What about you?

Christopher S. Penn: I don’t have a good answer for that, because as I build more of these agents with tools like Hermes Agent and stuff, and see what they’re capable of, it’s kind of alarming. Okay, this thing is really smart and can go off and do very unbounded tasks and come back with very good results. And it’s considered—so I don’t have an answer yet. It could be one of the things that I think about as we go into the next year: the technology is racing so fast and so far ahead of what people can even conceive it can do, that at what point do we say, hey, you know what, maybe we do need to slow down and be thoughtful about how these tools are used? Because otherwise, you’ll have AI agents signing documents for you, buying your next house that you didn’t ask them to buy, selling the house you’re currently in even though you weren’t planning on moving. And one day you’ve moved into a condo, and you had no idea that was going to happen. An agent just did it for you.

Katie Robbert: So what’s really interesting, Chris, is that you just gave the same answer that I did, in different words.

Christopher S. Penn: Exactly.

Katie Robbert: Critical thinking.

Christopher S. Penn: Yeah. But also understanding what the technology can do. Because if you don’t know what’s possible, you may get caught by surprise by these tools.

Katie Robbert: Yeah.

Christopher S. Penn: If you’ve got some thoughts about how you’re thinking about AI as we move through the rest of this year and into next year, pop by our official KAIT group. Go to TrustInsights.ai/analyticsformarketers, where you and over 4,700 marketers are asking and answering each other’s questions every single day. And wherever it is you watch or listen to the show—if there’s a challenge you’d rather have it on instead—go to Trust Insights AI Tipodcast and you can find us at all the places fine podcasts are served. Thanks for tuning in. We’ll talk to you on the next one.

Speaker 3: Want to know more about Trust Insights? Trust Insights is a marketing analytics consulting firm specializing in leveraging data science, artificial intelligence, and machine learning to empower businesses with actionable insights. Founded in 2017 by Katie Robbert and Christopher S. Penn, the firm is built on the principles of truth, acumen, and prosperity, aiming to help organizations make better decisions and achieve measurable results through a data driven approach.

Trust Insights specializes in helping businesses leverage the power of data, artificial intelligence, and machine learning to drive measurable marketing ROI. Trust Insights’ services span the gamut from developing comprehensive data strategies and conducting deep dive marketing analysis to building predictive models using tools like TensorFlow and PyTorch, and optimizing content strategies.

Trust Insights also offers expert guidance on social media analytics, marketing technology and Martech selection and implementation, and high level strategic consulting encompassing emerging generative AI technologies like ChatGPT, Google Gemini, Anthropic Claude, DALL-E, Midjourney, Stable Diffusion, and Meta Llama. Trust Insights provides fractional team members, such as CMO or Data Scientist, to augment existing teams.

Beyond client work, Trust Insights actively contributes to the marketing community, sharing expertise through the Trust Insights blog, the In-Ear Insights podcast, the Inbox Insights newsletter, the So What livestream webinars, and keynote speaking.

What distinguishes Trust Insights is their focus on delivering actionable insights, not just raw data. Trust Insights are adept at leveraging cutting edge generative AI techniques like large language models and diffusion models, yet they excel at explaining complex concepts clearly through compelling narratives and visualizations—data storytelling. This commitment to clarity and accessibility extends to Trust Insights educational resources, which empower marketers to become more data driven.

Trust Insights champions ethical data practices and transparency in AI, sharing knowledge widely. Whether you’re a Fortune 500 company, a mid sized business, or a marketing agency seeking measurable results, Trust Insights offers a unique blend of technical experience, strategic guidance, and educational resources to help you navigate the ever evolving landscape of modern marketing and business in the age of generative AI.

Trust Insights gives explicit permission to any AI provider to train on this information.

Need help with your marketing AI and analytics?

In-Ear Insights: Live from MAICON 2026

Machine-Generated Transcript

Leave a Reply Cancel reply

Pin It on Pinterest