Podcast: Play in new window | Download
Katie Robbert
Moving Beyond Generative AI
With the explosion of ChatGPT onto the scene, generative AI has become the focus of most people’s—and most marketers’—attention. But what other roles can AI fill in our marketing and business workflows, and how do we get started? Katie Robbert of Trust Insights gives a look into the data side of generative AI, but also what other tools are out there that we can put into use today.
Rich: My guest today is an authority on compliance, governance, change management, agile methodologies, and dealing with high stakes, no mistakes data. As CEO of Trust Insights, she oversees the growth of the company, manages operations and product commercialization, and sets the overall strategy. Her expertise includes strategic planning, marketing operations management, organizational behavior, and market research and analysis.
Now prior to co-founding Trust Insights, she built and grew multi-million dollar lines of business in the marketing technology, pharmaceutical, and health care industries. She led teams of Microsoft partner software engineers to build industry leading research software to address and mitigate pharmaceutical abuse.
She is a Google Analytics certified professional, a Google AdWords certified professional, a Google Digital Sales certified professional, and holds a Master of Science degree in Marketing and Technology Innovation. She is a published researcher in the Pharmacoepidemiology and Drug Safety Journal. That was a mouthful.
And today we’re going to be talking about use cases beyond generative AI, and AI in general, and whether it’s right for your business, with Katie Robbert. Katie, welcome to the podcast.
Katie: Thank you. I appreciate it. It’s always a little uncomfortable to hear your own bio because you’re like, who is that person? Did I actually do all those things? I think you did it a really good justice. So thank you for the enthusiasm with which you read it.
Rich: There’s a lot of good stuff in there. I do agree, sometimes people just fill their bio with these things, but this was every single thing in there was worth talking about, and very admirable.
I had somebody the other day say, “Wow, I was looking at your bio. How did you accomplish so much?” I’m like, I’m old. I’ve been doing this for a long time. Of course, my CV is a little bit long at this point. But, anyway.
Katie: We didn’t all do it in just one week. This is 20 years. It pains me to say that, but yes.
Rich: Now as we talked about, you’re the CEO of Trust Insights. Tell us a little bit about Trust Insights. What do you do and who do you tend to work with?
Katie: This is such a “fun” question. And I put “fun” in quotes because, depending on the context, the answer might change. But at the end of the day, we are a management marketing consulting firm, and we tend to work with midsize B2B companies that are looking to get more sophisticated with their data.
So we do a lot of data infrastructure setting up. We do a lot of work with governance, and training, and process development. And then at the end of the day, we do a lot of data analysis.
And so I was actually talking with someone the other day about this exact question. I’m like, the problem with my company is I don’t sell instant gratification. And that’s why a lot of people don’t want to buy it. We sell the longer term foundational bulletproof, like you’re not going to have to stick your MarTech stack together with chewing gum and sticks that you found outside. It’s going to be solid, but that takes time. And that’s what people need to decide if that’s important to them or not.
And then the good news is, eventually they come back around when they’re looking at this big pile of data and going, I don’t know what to do with this and I don’t know if any of it is correct.
Rich: It’s definitely a challenge selling something like, that when really it’s long-term strategic outcomes that you’re helping companies find. And most of us are just looking for, how do I use this week’s shiny object. So that’s got to be a little bit of a challenge.
Speaking of shiny objects, we’re talking about AI today, and talking about different use cases as well. But, just thinking about is as you approach businesses as they approach you talking about the strategy side of things. A lot of companies these days when they’re thinking about AI or they’re reading about it, they’re thinking of what’s generally considered to be generative AI. And we want to go beyond just that today, but let’s start there.
What in your mind or how do you define what generative AI is, and how are we using it today? And is this right for every business? I know those were a lot of questions, so take them at your own pace.
Katie: All right, so let me start with the first one of what is generative AI. Generative AI is one of three different types of artificial intelligence. And because I personally can never remember all of the proper terms, you went through my bio so there’s a lot of useless stuff up there now.
We came up with this acronym called FOG, also to represent the fogginess that’s in my brain. But the three different branches of artificial intelligence are basically artificial intelligence that finds the data, artificial intelligence that organizes and categorizes the data, and artificial intelligence that generates something based on the data that it’s found and organized. And that’s FOG, F O G.
And so what we’re talking about is that generative AI. Back in November, everyone started talking about OpenAI and ChatGPT. And so this has now become a household word. Everybody who is and isn’t in the marketing and tech spaces are talking about it. People’s parents are talking about it. My parents are talking about it. Restaurants are talking about it in terms of generating recipes.
I saw the other day that America’s Test Kitchen, who is very widely known and respected, was doing an A/B test to see if generative AI could create a better cookie recipe than them. And they’ve been doing it for a very long time.
And so generative AI, which is what we’re talking about, what everybody’s talking about, is using this system – any system, there’s a lot of tools out there – to create something that to the end user is net new.
Rich: And so a lot of content creators, copywriters, and so forth, marketers are starting to rely on these tools, ChatGPT probably being the Kleenex of them as far as the public is concerned, to create this content.
As you’re out there in the marketplace, what are some opportunities when it comes to something like generative AI? And what are some of the concerns that you have for companies that may be wanting to implement generative AI specifically into their business practices?
Katie: So some of the opportunities are definitely process efficiencies. And so generative AI, ChatGPT, Otter, another great tool, they’re really good at summarization. And we’ve talked with other teams at other companies and a lot of their time spent is on someone having to attend a meeting, taking notes, summarizing the notes, sending out the action items, following up on the action items, and so on and so forth. We all know how that goes. Some of the big opportunities with artificial intelligence are just to automate that and not to remove that person, but to remove that task.
Otter Chat, which just launched, has been really a game changer for my very small company because we can record and transcribe every meeting that we have with the integration. And then Otter Chat, you can literally ask questions the way that you would to something like a ChatGPT and say, “What are the action items?” And it will, based on who attended the meeting and what it’s recognizing, it can assign or give you the summary of action items per person with really good accuracy. And so that’s one of the big opportunities and time savers that we’ve seen.
Now that feels very basic but it’s so important in terms of factoring. Where does AI fit into my company? What are the opportunities?
And my general advice is always to start small with these kinds of things, because it can be very expensive, and it can be very disruptive. And so that’s when you start to get into the risks of introducing artificial intelligence, when you have a very short-term strategy versus a longer-term strategy.
Rich: The Otter example is really a very interesting one. Because on one level, you’re right, it’s fairly basic. Basically, we have this tool that’s going to record our meetings, take notes, summarize them, and give us to do list. On the other hand, I’m like, oh, my God, as somebody who takes terrible notes during meetings, I can rarely read my own handwriting afterwards. I haven’t used Otter for that, but I have used Fathom, which is built into Zoom.
So just to go down this particular Otter rabbit hole, this otter hole, are you just integrating that with Zoom or the equivalent, and letting it run in the background and then getting the notes afterwards? I’m doing something like that with Fathom as well. It’s nice to know that there’s other options and other tools out there.
And then how about a lot of companies, I’ve heard stories of copywriters who have lost their jobs and marketing companies who have had their budget slashed because suddenly the owner of the company or the accountant is like, why don’t we just use ChatGPT to do this? Any thoughts on whether or not we should be going all in on letting AI create our content, or is this a “penny-wise, pound-foolish” type of approach, in your opinion?
Katie: It’s a terrible idea. This is my opinion.
Rich: As a content creator, I really appreciate that. So yes, please go on.
Katie: But I can also back it up with really solid information. So one of the things that may or may not be well known and/or ignored by these decision makers who are deciding that ChatGPT is the right way to go and then let go of all their humans, is that there are limitations. to the learning models that ChatGPT is using to create content.
First and foremost, ChatGPT – or GPT3, which is the learning model that powers ChatGPT – only goes back to 2021. And if you are looking for information beyond that point in time, you’re not going to find it in GPT3. GPT4 goes back a little bit farther, but again, it doesn’t go back far enough to really pull from references that you may find are more valuable than something that just happened maybe two years ago.
The other side of that is that what we’ve seen, the whole Google algorithm and SEO and ranking, conversations aside, is that the content that comes out is mediocre at best. And sure, you can generate higher volumes of content using these tools, but the quality of the content is going to start to suffer over time. It’s going to be very generic.
Rich, if you and I put in very similar prompts, even though we work in different companies, different industries, different customers, we’re going to get the same generic content. And so it’s just really going to be the same message over and over again with no personalization, with no voice of customer data, with no real brand voice.
The way for companies to do that is to build their own large learning models that are trained on their own content. And that, in and of itself, is a huge undertaking. Because not only do you have to build it, you have to maintain it, you have to keep retraining it, you have to keep tuning it. And it just it’s a big task and not for everybody.
Rich: And obviously we’re talking about this, it’s July 2023, who knows what’s going to be happening by July 2024 or even three months from now. Who knows? Somebody could come up with something that would work really well.
But in this day and age, what I’m hearing is without spending extra time on this, we’re going to get to basically everything goes to the middle, everything becomes more generic, and there’s less brand differentiation, less opportunity to be remarkable or to stand out in the marketplace. Which is one of the problems that I think a lot of brands already face. And it’s only going to get worse if you rely on this tool that is, by design, almost vanilla.
Katie: And that’s exactly it. And that’s not to say that the tool itself isn’t really interesting and does a lot of great things. But one of the challenges is that we, as end users, have really decided that net new content creation, that’s what it does. And that is probably the use case that is used most, and what it is the least good at.
And so better use cases are the Otter and the Fathom where it’s summarizing. It’s creating those action items. You are giving it information to create code. The code creator came out recently and people were losing their minds about it because it’s saving a lot of time in terms of QA and checking things and helping you get unstuck.
We do a lot of custom code for the machine learning that we do at Trust Insights. And Chris, who handles all of that, is not a coder by trade. A lot of it is self-taught. And I think a lot of people who are in the industry are self-taught, versus have that traditional schooling background, and tools like this can help you get unstuck. They’re almost like training tools. They’re that supplemental piece.
And so generative AI is better suited for those kinds of use cases versus what everybody’s using it for to just write me a blog post, write me a press release, write me a summary, or write me a blurb for this social media post. It’s generic.
Rich: And so in places where generic copy or code may not be problematic, that might be a best place to use it in this day and age. Especially if something’s not customer facing. I could even see employee handbooks. There’s only so much… or maybe not legal documents, because we want to stay out of hot water. But just that kind of thing that’s less problems where we don’t have to show off our unique attributes, that might be a great opportunity. Interesting.
So we talked, you mentioned FOG – find data, organize data, generate data. So we talked a little bit so far about generating that. How about finding data? What are some of the tools out there and how can businesses take advantage of that aspect of AI?
Katie: So this is where companies get into building their own large learning models. Which is again, not a small undertaking. But one of the initial risks with tools like ChatGPT was the lack of understanding of the type of information that end users were putting into it, specifically personally identifiable information.
Because what happens is, let’s say I want to do a search on everybody born on the same day as me. I first have to put that information into this tool, and then the tool then stores that in its large learning model. And it’s learning personally identifiable information about me and then finding other personally identifiable data points about other people. And so that’s part of the risk of these tools is the end users lack of understanding of what information is appropriate. So that’s where companies need to start and say, this is what we will and won’t do with these tools.
In terms of your question of where companies find the data. It really depends on what their end goal is and what they’re using it for. And so if they’re building their own large learning model and their goal is to understand how cancer is diagnosed, for example. They would want to then train the learning model on all of the cancer research that has ever existed. So that way the large learning model number one is focused, and number two has all of the information that it can possibly get.
These tools are hungry. They want to get more data, more information. They want to get more sophisticated. I’m talking about them like they’re sentient. They’re not. Let’s just be clear on that. They are not really, “they”, it’s “it.” But they’re not. And so in terms of ChatGPT, it’s not great at citing its sources. But if you use ChatGPT through Bing, for example, it will cite sources. And then I believe Google just rolled out its version of generative AI, the chat. And so it does start to cite sources.
And so you can start to see where the data’s coming from. And that’s the other risk, is when you’re using these generative AI tools, unless you solely own them and you are the one inputting the data to train it, you don’t necessarily know what the source is. And so it’s the same thing as doing a Google or an internet search, not to pick on Google too much. But you have to really do your own due diligence and do your own fact finding of, if I look up “what’s the weather today”, I can just look outside and see whether or not my search result is correct. And 10 times out of 10, it’s wrong. It says it’s supposed to be raining and it’s sunny outside, or it’s pouring and it says it’s great beach day.
And so you have to treat these tools the same way in terms of verifying the information, but also trying to figure out where it’s getting its data from.
Rich: Okay. So let’s talk a little bit about the organizational aspects. You talked about that. What are some of the tools, what are some of the ways that businesses might be approaching this aspect of AI?
Katie: I don’t know that they are. More sophisticated tech companies are. And I also don’t know that they are finding the data. Most businesses at this point are in their infancy in terms of their journey with artificial intelligence, and so they’re mostly using the generative AI. They’re letting companies like OpenAI and GoCharlie do the finding and organizing. But essentially what it comes down to is the categorization of data.
So we used to do a very basic version of that called “topic modeling”, where it’s essentially saying, okay, find all of the common terms and phrases and put those in a bucket. So this is all of the content that is about Rich Brooks. And this is all of the content that’s about Katie Robbert.
Another way to think about that categorization is sentiment analysis. Is this positive? Is this negative? Is this neutral? And a lot of tools already have that kind of technology built in, especially the social listening tools. And those are the kinds of tools, if companies are looking to categorize their data, I would start with those.
Rich: I think a lot of businesses don’t realize that they’ve been using AI and machine learning tools for a while. Because it wasn’t really until ChatGPT came into the consciousness that we saw a lot of these tools. Like Google and Meta’s ad platforms are very much based on AI and machine learning. We don’t think about it because we’re not the ones actually working on them, but we’re integrating them.
For businesses that are listening to this podcast, to this conversation we’re having, is there an action item for them as far as you see, or should they be going out of their way to learn new tools, or should they be waiting for some other company to finish the tools? Because we’re not in the business of AI, we just want to take advantage of them. What might you say they could do after this podcast that might help them along their journey?
Katie: Surprisingly enough, or maybe it’s not surprising, I wouldn’t start with the tools themselves. Because I can look at the roster of artificial intelligence tools, but 99% of them aren’t going to make sense for my business.
So the action item that businesses should take away is understanding what is the problem I am trying to solve. And looking at your processes, looking at your organizational structure, looking at the quality of the data that you’re analyzing. I would start with analyzing and auditing those pieces first, because then you can start to figure out what tools, what pieces of software, what platforms are actually going to solve my problem.
And it might not be an artificial intelligence tool, and that’s okay. I think the challenge right now is that companies. want to be relevant in the conversation. They want to say they’re AI driven. It just doesn’t make sense for everybody. It doesn’t make sense for every single task.
And so the takeaway is start with really being concrete about what is the problem I am trying to solve, period. And not solve with AI. It’s just what am I trying to solve, period.
Rich: I love that approach because it is so easy to just be interested in the shiny objects. And I know I’m worse than anybody about it. I think AI is fascinating. I’m learning everything I can. I’m talking to everybody smarter than me, more advanced than me about what they’re doing it for. But I think that’s not necessarily right for everyone. And if you’re sitting there trying to understand how can AI solve my problems, first, you have to identify what are those problems.
And like you said, is AI the best tool for this particular situation? And I think that there are going to be a lot of opportunities. I think that the tools are going to become better. I don’t know about your life, but I would say that because of who I am, 50% to 75% of the business-related emails that are coming in these days have the word “AI” either in the subject line or in the body copy. And I just almost feel like AI has become, “Now with AI”, whatever the business is, it could be a mac and cheese box and they’re advertising their use of AI.
Do you think that consumers are excited about seeing AI implemented into things they are used to using every day? Do you think that’s a turnoff? I remember when every business added “.com “to the end of their name before they all took it off, because they wanted to show that they were on the internet. Are we going through a similar phase right now with AI?
Katie: I think we are. I think it’s definitely the crypto/ Bitcoin/NFT. There was some other.. oh, Clubhouse, whatever trend. I think we’re going through that again.
So what’s interesting is when we launched Trust Insights five and a half years ago, we’ve been working within the AI space for about a decade or so. And we initially launched a company leading with that artificial intelligence foot. And what we found was that people weren’t ready for it. They were like, “Oh no, that’s big and scary and expensive”, when really it was just how we powered our very small business. And now we’re having to rethink our messaging to make it more AI forward.
But the challenge there is do people who don’t know us think we’re just jumping on the bandwagon with your AI powered mac and cheese? Consumers don’t care. They just want to know that they’re getting a really good product at the end of the day. You can definitely start to branch off and segment the end users into the conspiracy theorists who think that the bots and the man are always watching, so don’t give it any of your data. And then the people who welcome their robot overlords. And then there’s everyone else who sits in the middle.
And so everyone who sits in the middle, but that’s the majority of the end users. They don’t really care. They just want to know that their problem is getting solved, and it’s getting solved in a way that they can afford and that is repeatable.
Rich: Awesome. Katie, this has been a really fun conversation and I loved your takes on things. If people want to learn more about you or more about Trust Insights, where can we send them?
Katie: They can find our website at trustinsights.ai. Yes, with the AI at the end. Not a dot com. They can find me at Katie Robbert on most social platforms, and they can find Trust Insights also on most social platforms.
Rich: Awesome. And we’ll have all of those links in the show notes. Katie, thank you so much for coming on and having this conversation with us.
Katie: Thank you for having me, Rich.
Show Notes:
Katie Robbert is an AI and machine learning expert who helps marketers better understand how to collect and measure data so they can make smarter and more impactful marketing decisions. Check out the work she and her team at Trust Insights is doing in the AI/machine learning space.
Rich Brooks is the President of flyte new media, a web design & digital marketing agency in Portland, Maine, and founder of the Agents of Change. He’s passionate about helping small businesses grow online and has put his 25+ years of experience into the book, The Lead Machine: The Small Business Guide to Digital Marketing.