frankly… Episode 62: Get Your Trusted AI Insights Here

In this episode of frankly…

The transcript below is AI-generated and may contain minor inaccuracies. Tune in to the episode audio to hear the full conversation! 

Transcript 

Dan 

Hello, welcome back to frankly. 

Rachel 

Welcome. 

Dan 

So today we are talking with Katie Robert who is cofounder and CEO of Trust Insights marketing analytics consulting firm. It’s so much more than that. And Katie really dives deep with us on the topic of AI among a few other things. You know how how companies can. Find the right AI tools, the right processes, even to try to automate with things like AI and some of the things that you should consider if you’re starting to build out a strategy for your own organization. 

Rachel 

Yeah, maybe it’s AI. It’s not. She talks a little bit about. That AI isn’t gonna solve the end all end all. Be all for for any problem that you have and you know she talks a little bit about it, but. She founded Trust Insights with her Co founder in 2017. You know, the fore thought of. 

Speaker 

Yeah. 

Rachel 

Using AI you know for marketing analytics is very impressive back. So she spent a lot of time in this space and has a lot of incredible knowledge. 

Dan 

Yeah. So we’ll we’ll get into it and turn it over to her. But but I think this is this is a great resource for anyone who’s who’s looking to get a little bit smarter to get a little bit more familiar or just starting to explore the vast world of. AI. With that, we will turn it over to Katie. Hi. Welcome to frankly, thanks for coming on. 

Katie 

Thanks for having. I’m excited to talk with you today. 

Dan 

Start out with with kind of a kind of a basic question, but tell us a little bit about your background, your career path that ultimately led you to your current role with trust insights. 

Katie 

You know, it’s a, it’s a bit of a journey, which I’m sure is true of anyone that you ask the question to. I I rarely meet someone who goes in sort of in a straight line in their career path. The one thing I can tell you is that I have been bossy since birth, so that was always. Is the end goal was to be in charge of something. But I actually have a film degree for undergrad and so so I’m in the Boston area and at the time when I graduated from college, the film industry hadn’t quite made it to New England. Were still based in New York and California, and at 22 years old. I couldn’t afford to move to either 1, so I had to rethink that. 

Dan 

Fair. 

Katie 

So I started working. Full time as a project coordinator. Actually doing clinical trials through NIH, so substance abuse and pain management of opiates and stimulants. But that’s where I learned a lot about data governance and process and project management and operations. So I did that for about a decade. You know, building consumer products based on the research and then after that I transitioned into managing a marketing technology team at APR Agency of All Places. And that’s where I met my Co founder of trust insights. And after about 2 1/2 years being at the agency. We we wanted to get more into deeper tech such as AI, which is something that we have been working in and decided that our best bet was to spin off. 

Speaker 

And. 

Katie 

Start our own agency, which is where we are today, almost eight years later with trust insights. 

Rachel 

It was just going to ask that what year was that because to have the four. 

Speaker 

Yeah. 

Rachel 

Like the foresight into starting an AI based company 8 years ago is incredible #1. #2, I love the fact that you said that you’ve been bossy since birth. I feel that so. I feel like that’s a I’m gonna need to take that. So what has changed in AI in the last eight years of working in this space? Is different. You know, how is this evolved? 

Katie 

The short answer is everything. 

Rachel 

Right. 

Katie 

Everything has changed. When I was first, so I’ve always been aware of artificial intelligence, but in a, you know, using it through marketing was something that was newer to me. And so what’s changed? A lot is the accessibility of using AI, so there’s traditional AI, and then there’s generative AI, which is one of three components of AI as the whole umbrella. So. Everybody’s talking about today is generative AI. And so that’s just one of three components. So the way that I remember it is I use the acronym Fog, so find, organize, generate so. One component of AI is finding the data. One component of AI is organizing the data and the component that we’re all familiar with now is generation. And so I’m. I was introduced to AI as a whole for all three components, and so we were using. You know, offline models building our own pieces, doing a lot of within the marketing space, doing a lot of topic modeling and. You know, synthesizing data for market research and now with generative AIA, lot of that is automated. So it’s it’s interesting to see how it’s evolved from, you know, only very technical. Resources could use it to pretty much anyone can use, and it’s now built into all of our devices. 

Dan 

Yeah, that barrier to entry has really come down in recent years. 

Katie 

It really. Yeah. 

Dan 

And and what’s something you know as you talk with marketing professionals or as you kind of do your work in this in this? Like consulting space. What are some of the? What are some of the maybe misconceptions that marketers have about AI? Or maybe some of the reasons why they’re hesitant to look at it as an option? 

Katie 

Well, a lot of the misconceptions is that number one is to set it and forget it, that you can just, you know, plug it in out-of-the-box and boom, magic happens. A lot of upfront work that goes into it and that I think is also. You know, to your other question where a lot of people are hesitant to start because like any good piece of technology you have to do those requirements scattering up front. Have to do that? Government governance documentation. You know, goal setting, KPIs, all of that stuff in order to get it to work correctly. And you know in a sustainable way long term. Sure, I can open up any generative AI platform and start typing, but that. That doesn’t give me something that I can use long term, so that’s sort of the misconception is you can just use it right away. 

Dan 

So, so to kind of like maybe put some context around that, are you talking about kind of taking those models and and building something that’s more of a purpose built application for you or is it is it something else that would be that perhaps? 

Katie 

No, I think that that’s really so really getting a better understanding of why. You’re using it so a really good. I work with a local Humane Society, and so I do the dog volunteering. But I also work with the executive director because as a non profit their resources are stretched really thin. So she came to me and she’s like, I know that you’re in the AI space. Is there something we can we can do with that and? Was like, yeah. But once we started talking through. A lot of the challenges that she’s facing. AI is not necessarily the solution, and that might be another misconception. To answer your question is AI is going to fix it and that’s not necessarily true. A lot of where she. Is actually people and process and so once those pieces are cleaned up then you can start to introduce different text such as AI. A big misconception is if I just put AI on it, things are going to get better. 

Dan 

Yeah, I. 

Speaker 

Yeah. 

Rachel 

Also think that that brings up a point of people thinking that AI might be the end all be all solution when there might be some other pieces that need to be in there to make it happen. Right. So one thing and maybe background for some of the listeners, Lexi and our team went through the Trust Insights program and is now helping our agency roll out an AI process, right, so. I am lucky enough to sit on that team and task force to help roll it out to the agency and I think one of the best. Things that we could have done is instead of jumping in and saying like, let’s just start using AI we started with put AI to the side and ask yourself what you want to do, right? It doesn’t. Maybe AI is a solution like you were just saying, Katie, right? Maybe that is a solution, but maybe it’s not. And so building those situations of what you’re trying to solve or make easier is the first step is where you realize like, yes, AI can help some of those. Or do I need an? Automation tool that links with AI. Or do I need a third party program outside of just these generative AI models? Do some of my current tools have some AI feature that might help solve that problem? And so I think. That might be another misconception is just that AI, like you said, is going to solve everything. You might actually have a fully other solution available, so I think that. Your your process of helping people with AI. I’ve seen it and I’ve been able to use it at our agency has been really insightful because that’s not where many people start when they look at their issues. 

Katie 

Yeah, it’s, I mean, I get it. You know it. AI for a lot of people is a shiny. It’s, you know, the solution in search of a problem. The way that I approach that with clients and you know with our community is I give them. What’s called the trust Insights 5P framework. And so because marketers love alliteration, the five PS are purpose people, process platform and performance. It’s, you know, a lot of people like, oh, that’s just digital transformation. It’s built on the back of digital transformation. Challenge I have that I’ve experienced with digital transformation. Excuse me, digital transformation is that it puts the technology first. So you’re choosing a tool and then trying to retrofit your people and process into whatever you whatever you’ve selected so. So if it’s a cream so great example, we chose acrm when we first started the company based on cost not based on need. Now, 8 years later. We have the CRM with eight years worth of data that doesn’t do what we need it to do because we chose based on you know, the wrong things. We didn’t do the evaluation. We didn’t start with our. What is it we needed to? Who are the people who are going to use it? What is the process for getting data in and out of it and then choose a platform? 

Dan 

Yeah. And it’s at a certain level, you almost feel stuck in that. Because you have so much of that background there that it’s overwhelming, I’m sure to think about moving all of that to something else. Might be a better fit. 

Katie 

Everyday. 

Dan 

Yeah. So so this is maybe a question that goes that goes back a little bit but but when you were looking at when you were looking at founding or Co founding trust insights? Some of the gaps that you saw are what was like the why that you that you came across, that you felt like this is something that the industry really really needs in this kind of analytics AI kind of space. 

Katie 

Transparency was the Big 1 and I say that because if you’ve ever worked with a consulting firm. You know, you may have had the experience where they send a bunch of people and you all sit in a big conference room. They collect a bunch of. Black box it and then give you recommendations and then say good luck. 

Speaker 

Yeah. 

Katie 

Here’s your bill, and I wanted to change the approach because. Having been on that side of it, I knew that there was better education that could be. There was better execution that could be done and one of the things that we bring to the table that we’re really. Really, you know, proud of is that we can not only give you guidance, but we can also do the thing. And so we’re not a consulting firm that just sort of says here’s all your recommendations. Good luck. We can also take it that next step and say. Here’s the five things of the 10 that we can do, those other things, we have partners that can help you with those that we’ve vetted, that we already trust. So we’re not just going to leave you with a pile of. Things on your to do. We can actually get you to that next step and we’re going to be transparent about exactly what we’re doing every step of the way, so that if you have to justify the cost or the value or even if you just have questions about how we do it. We’re happy to share that because I’m not someone who worries oh, they’re going to steal my, you know, secret sauce, my IP. What my approach? Go ahead. I welcome you to. 

Dan 

Yeah, yeah, yeah. If you’ve got the tools and you can make it work then. But but that that I think is a is a really interesting point about not being kind of like this black box provider where you can actually show your work. Can explain? The the Why behind this and and kind of show show why the recommendations make. I think that’s so important in in both what? What you do, what we do, what kind of this whole outside agency. Approach comes down to is the why. 

Katie 

And the Y is no. 

Rachel 

You have to prove your your why they hire you. And it’s by giving you that giving them the why. If you can’t tell them that, then no one has a reason to hire you, right? I mean, that’s how we all kind of work in this agency setting. 

Katie 

No, it’s. And you see a lot of the more successful agencies. It’s less about their, you know, marketing tactics and more about that trusted word of mouth referral because you’re going to, you know, if Dan comes to me and says, you know, I need someone who’s going to help me with my e-mail marketing. I’m going to give you a list of five people that I trust and you’re going to, you know, look at those resources first before you’re just going to do like a. Search on the web. 

Rachel 

Yeah, network matters. In so beyond just personal network, but company network and trusted partners, right? That I would love to hear from. What are some things or processes that companies can put in? So maybe they’re they don’t have an AI team yet, or they don’t have an AI, you know, policy. But people are out there using those tools. We know that. So what can companies do to help with AI adoption to ensure their team members aren’t left behind or using it in? Ethical. Or are you know just? Armed with the right tools to use these tools. 

Katie 

It’s going to sound very silly and basic, but the first thing any company can do is better documentation of their processes. That is the Super secret UN secret key to success with adoption is having clearly defined process. That you can then say if we insert AI at this stage of the process then we can see is it working. Are people stumbling over it? It creating you know, bigger roadblocks or is it actually making things more efficient because otherwise? You know, this was voice of customer feedback that I got the other day was, you know, our executives just invested in copilot, and now they just want us to use it. Like, that’s not really helpful because you’re not being given any direction and you don’t even know where to start. And so one of the things that I like to do is, you know, go through the exercise of what are the things you’re doing. More than. So what are all the repeatable things that you’re doing, even if it’s not a well defined process? And then you know, getting some of at a high level some of those processes so that you can see, OK, this is a great opportunity to use AI versus this, right? There’s a lot more work that needs to be done and it really kind of helps with the conversation of. Where do we start? And so if you’re doing, you know, at the end of the month, you’re buried under 203040 reports that are all structured the same way. Nobody ever looks at them in a bunch of PowerPoints. That’s a great opportunity to use generative AI because you can automate a lot of that because the process is the same over and over and over again, and it takes a lot of that time off of your plate to focus more on actionable insight. So that maybe some. Will look at it. Versus that one thing you did that one time. That’s probably not the best opportunity to start with generative AI. 

Dan 

Right as that time investment, you know to build a process to find a tool really worth it to automate one thing that you do once a. Yeah, probably not. 

Speaker 

Mm. 

Katie 

But a lot of people start with those things because it feels like, oh, this is the outlier. This is where generative AI can fit in versus looking at what’s right in front of you. With things that you know, like like the back of your hand. And like, here’s how I make my To Do List. How I set my calendar? How I outline content? Here’s how I structure a. Those things you’re going to see very quickly whether or not generative AI is going to be beneficial to you or not. 

Rachel 

Yeah, it’s like those repeatable things, repeatable things that can take like administrative tasks off of a team member. 

Katie 

Mm. 

Dan 

Yeah, and and it, it seems to me that that’s something that will naturally become. Well. Kind of understand how they can use it to make you know shopping lists to organize their own calendars like in their personal time. I think a lot of it is just, maybe just most people have such a basic or low level or no familiarity with it at this point that as you as it becomes more pervasive these you start to use it you you’ll start to naturally have that mindset. Of I could use AI for this to automate something that’s so frequent, so repeatable like you said. 

Katie 

I think there’s, you know, one of the questions you had asked was, what’s the hesitation or why? Isn’t there more adoption and? A lot of it is around. What does this mean for me as the individual? What does this mean for my, you know, daily workload? My weekly, you know, reviews or my end of year performance review? Does this mean that I’m being asked to do more being asked to do less being asked to do something different? That I’m not comfortable with generative. Using it proficiently can be a different skill set. So that’s. Of the barrier to adoption. The other is just lack of expectation of what it is you’re supposed to do with it, but also the other side of that is, what do you do with all that time you have back? You know, one of the things that we. Advise people on is great. Now you’re no longer buried under 30 hours a week of reporting. You have that 30 hours. Do you do with? And that’s where that critical thinking comes in. Not human to human connection. If you’re in a client services type agency. Deep in your relationship with your clients, get better. You know, understanding of what their pain points are ’cause you may not have time to do that. Think about all the stuff you don’t have time to do as it’s going to. Enhance the work that you do on a day-to-day basis. 

Dan 

Yeah. And it’s, and even beyond the like client relationships or things like that, it’s also more time to think strategically and to kind of connect better understand the industries that you work with, better understand what’s being talked about out there, better understand your clients, customers, your customers and put. This together, so you can. So that you can spend more time on the things that. You need a human mind for right and kind of leave behind the rest. 

Katie 

Exactly. Yeah, and that’s exactly. There’s a lot that AI can do. There’s a lot that it can’t do, and you, as a human have a really great opportunity to let AI do the admin stuff, and then you can do the strategic thinking. The critical problem solving, you know, the creativity that AI is just not going to. Because you, as the human, understand nuance and have that, you know, insider information that AI doesn’t have. 

Rachel 

And that is a good segue into this question about tying maybe using it. Talked about reporting, right. So marketing communications, traditionally it’s very difficult to tie to like real world results. Like media relations, we used to report on ad value or something else along those lines. Really doesn’t tell you what you’re getting out of these placements. Uh. All the analytics can be a little bit. Just broad. They’re not telling us. So what are some ways that marketing communications professionals can help clients or their own company connect that data to objectives? Is there a way that AI might work together with it to analyze it, to try to get some of those insights and results? 

Katie 

It. It depends on the quality of the data that you’re putting in, which you know is always going to be different from person to person, company to company, you know. For example, if you are. Tracking. If you’re doing a lot of placements and you’re trying to track the back the traffic back to a Google Analytics system, for instance, you can take that Google Analytics data like just a screenshot of it. Give it to generative AI in addition to some context around, and here’s the other things that we’re doing. What do you see? What are you seeing that I’m not seeing or what should my next steps be? You might not get anything good, but you could get insights that you hadn’t thought about because you’re kind of blind to the data ’cause you look at it all the time. We did this as part of a measurement workshop where we were, you know, the things that feel like they’re harder to measure. If you have, you know the level of effort data or the amount of content that you’ve put out data or the engagement data, and then also your web analytics data generative AI. Can do a better job of tying those pieces together to find the trends that you, as the human might just be staring at four different spreadsheets going. What am I looking at? 

Dan 

1. Yeah, being able to compare and see like where spikes in one in one area may be connect with spikes in another and and and tie those two together is is something that takes a lot of time, even if it’s even if it’s something that you can do it. That takes quite a long time for for a person to take a look at. Those two things against each other.

Katie 

The other thing that you know, a marketing communications professional could do is so let’s say they have a really good grasp of the data you can give it to generative AI and say ask me questions about this questions that I might not have thought of so that you. Really start to dig into that critical thinking and problem solving. Because again, you’re looking at it through the lens of maybe your client or your team. But let a you know third party who knows nothing about it. Start asking questions. 

Dan 

Yeah. What would somebody looking at this want to know what would be important to them? Oh, that’s that’s such an interesting one. I hadn’t thought about it in. Way. 

Rachel 

Before it reminds me of asking. AI to check its work. You know, like we’re asking it to check our work, but I always love telling AI. Why don’t you go ahead? Check that. See if you see anything weird about. 

Katie 

It that I mean, that’s exactly. You know, it’s a really great resource for that and it’s not going to give you. Attitude about it. If you ask it to check its work, which is also very nice. 

Rachel 

They do tell it. Often, just in case anything weird ever happens, I find myself saying please do this. 

Dan 

Say on the good side. 

Rachel 

Yeah. This I’m just curious. I have found in using all the different tools that some are stronger in some places than others and you know that’s going to be, I think true for any tool that we use, not just AI. But have you found any strengths in certain tools versus others that really stand out to you that might be helpful for people? 

Katie 

Yes. So about once 1/4. We do on our live stream, we do what we call the generative AI Bake Off. And so we look at the, you know, 5 or 6 major models and we put it through I think a dozen different tasks. So there’s 6. And I. I’m going to forget what they are off the top my head, but there’s six major categories of use cases for generative AI, and then within those we give 2 tasks. So you know you have classification. Generation summarization. And three others that I’m forgetting off the top of my head. And So what we? Is we pick them all against each. We have all 5 or 6 models running simultaneously with the exact same task. And so there are models that are better at certain tasks than other ones. And I can certainly dig out that information and send it to you to include. With this. But yeah, we absolutely do find that. 

Dan 

Yeah, happy to link to that. 

Katie 

What? You know, ChatGPT is good. It’s not necessarily what Gemini is going to be good at, and so you have to know going into using those models what your goal is so that you can choose the correct one. 

Dan 

Yeah. 

Rachel 

I think that brings up a great point. Like I found that Claude specifically I find does better at like writing or generation or actually putting together some words to get me a start for something versus. It’s not as good if I’m asking it to search for something, right? Tends you better with that because it’s Google based. So, like understanding the strengths and even really who’s managing that model, I think can be very helpful to get you in the right direction. Like copilot is is the best one we found for transcription of. Summarization will tell by Microsoft, which has all of your documents in it, right? If you look and think about each one and. Who it’s. By it also starts to kind of like put you in a right direction, just as a you know, I haven’t run them side by side like you guys have, but that can help get you a good start too. 

Katie 

So just looking at our results, so the six categories of use cases are generate, extract, summarize, rewrite, classify and question and answer. Any. Use case that you come up with for generative AI is likely going to fall under one of those six categories, and so the last Test that we did was copilot Gemini GP Open a IGPT, Claude Sonnet, Mistral and Meta’s Lama. And of overall, the best overall model as of Q 4/20/24 was open A is GPT. That’s the best overall model. If you had to just invest in one. But then as you get into specific tasks, there were tasks that, you know, copilot couldn’t do that. It looks like GPT could do and vice. So it really comes down to how you’re going to be using it. The majority of the time. 

Dan 

Yeah, and and that goes all the way back to maybe your first response. Of know. Know what you’re looking? Know what processes you need before looking into tools and where to go with it. 

Katie 

And that’s where I always I sound like a broken record around my. But I always point people back to the five P framework, and that’s how you’re going to help define those pieces. 

Dan 

So. So just one kind of closing fun question that we like to ask. What? What is one thing or one piece of advice that you would have given your 21 year old self? Know looking back. As throughout your career, anything that’s helped you along the way, you know, develop as a leader, things along those lines. 

Katie 

Don’t wait for someone else to Pat you on the back. You know, have confidence in the work that you’re doing that you’re bringing value to the. But if you’re waiting for accolades, don’t just keep moving forward, because depending on the situation, they may or may not come. That doesn’t mean the work that you’re doing isn’t valuable. 

Dan 

Yeah, yeah. Trust in the value of your own work. Love that. 

Rachel 

I love that. I saw something the other day that said if your standard is overachieving, then you’re never going to feel like you are achieving right because you’re that’s just like your only option for you. You think that is the normal baseline. And so I think it. Of goes. With what you said of just. You don’t have to overachieve to achieve pat yourself on the back for everything that you can do and cross off. Is bigger than you think it is. I think that’s fabulous advice. 

Dan 

Well, Katie, thank you so much for coming on today. It was great talking with you. Learned a lot. And. 

Rachel 

I could ask you 700 more questions. 

Speaker 

Yeah, always. 

Dan 

But yeah, we hope to catch up soon. 

Katie 

Yeah. Thank you so much for having me. 

Dan 

All right, bye. 

Rachel 

Thanks again to Katie. Like I mentioned there at the end, what incredible knowledge she has. 

Speaker 

Yeah. 

Rachel 

Could go forever. 

Dan 

Yeah. Yeah. And I love the idea of coming back around to this someday. You know, AI will will probably be having a different conversation entirely in six months in a year and two years. So I think there’s a lot of. 

Speaker 

Got it. 

Dan 

A lot of potential here to come back around and just learn something new. 

Rachel 

And. I even think about. We could revisit our AI episodes from a year ago and have a completely different conversation. 

Dan 

Oh. Clustering. No clusters, yeah. Yep. Yeah. Hey, six weeks. We’ll probably be saying something else. So we’re seeing something new on that out there. 

Rachel 

Absolutely. 

Dan 

But yeah, but yeah, again, thank you to Katie. Great time with this discussion and looking forward to the next. 

Rachel 

Thanks so much for tuning in this week and we will see you in a couple weeks. 

Dan 

Yeah.