So What? Marketing Analytics and Insights Live
airs every Thursday at 1 pm EST.
You can watch on YouTube Live. Be sure to subscribe and follow so you never miss an episode!
In this week’s episode of So What? we focus on Identifying Generative AI Use Cases using a 2×2 Matrix. We walk through how to set up a 2×2 matrix for decision making, what AI tasks are optimization versus innovation and how to use the 2×2 matrix to communicate with your team. Catch the replay here:
In this episode you’ll learn:
- How to set up a 2×2 matrix for decision making
- What AI tasks are optimization versus innovation
- Using the 2×2 matrix to communicate with your team
Have a question or topic you’d like to see us cover? Reach out here: https://www.trustinsights.ai/resources/so-what-the-marketing-analytics-and-insights-show/
Katie Robbert 0:47
Well, hey everyone, Happy Thursday. Welcome to so what the marketing analytics and insights live show I’m Katie joined today by only Chris.
Christopher Penn 0:57
John is at inbound. He is
Katie Robbert 0:59
he is representing Trust Insights and marketing over coffee at inbound in downtown Boston this week. And so we’re very excited to hear back from him everything that’s going on what’s new with Hubspot, and then just general catching up with people. So if you are in the area, go say hi to John. He’s wandering around the Boston Convention Center today. On this week’s episode, we are covering identifying generative AI use cases, using a two by two matrix. We’ll get to the two by two matrix in just a moment. But you know, we’ve covered AI use cases before we’ve talked about them, you know, we’ll cover again what they are. But we wanted to take it up a notch by really helping you think about okay, that’s great. Now, I know the use case, but what does that mean for my company? And so really finding a thoughtful way to assess whether or not the use case was going to be something you should prioritize. So Chris, where would you like to start today?
Christopher Penn 1:59
Oh, well, that’s a good question. I mean, one of the things I was delivering a lecture yesterday at Wheaton College, and one of the students asked the very salient question, what jobs is generative AI gonna affect? I said, Only the ones that use language, which by the way, is all of them, there is I can’t think of a single career. Even someone you know, like beveling stuff, you know, outside road, you still use language to do to reports to your time sheets and stuff, you are still communicating, because communication is such an integral part of who we are as human beings. And as a result, every job will have been impacted in some way, by large language models and generative AI. So the challenge is, for executives, for decision makers, for leaders, to figure out a, what are the use cases with the use case categories even? And then look at your organization and say, How do I even map this up, because again, it’s gonna affect everything, your finance team, your HR team, your recruiting, your sales, your marketing, your operations, if humans are involved, it’s going to be impacted.
Katie Robbert 3:11
I think you hit on something that I think is a really important distinction. There is use case categories. And then there’s the use cases within those categories. The use cases themselves are numerous. But they all have to eventually fall under one of the categories of generative AI. So Chris, what are those six categories for generative AI, where as we start to think about the use cases,
Christopher Penn 3:37
the categories our generation, aka making stuff, write blog posts, emails, limericks, you name it, extraction, we have a pile of data in some text format that you need to get data out of. So if you’ve ever written Are you ever read a white paper, and there’s tabular data, like buried in the PDF, like 22 pages down, you’re like, I don’t want to copy and paste that up by hand. A language model can go and fish that table out and make it machine readable for summarization. So if you’ve watched any of our podcast episodes, or previous live streams, you know that their transcripts, maybe even that’s too much to consume. So you can take the transcript, put it into a language bottled say, just give me the high points, like, give me the CliffsNotes version of it. There’s rewriting where you can take text in any form or convert it to any other form. And this can be silly things like hey, rewrite the hobbit but in emoji, right. It can also be explaining concepts re explaining your thoughts with one my favorites is explain this in terms of pizza. And you’d be surprised at the number of very complicated things like quantum superposition that can be explained with pizza pizza, sort of the universal explainer. The fifth category is classification where You can take bodies and chunks of text and classify them categorize them. And the last was question answering where you can take either the model itself or specific pieces of text and ask questions. So again, one of my favorite things to do with this. I was looking at an NDA the other day, like, Oh, my God, this NDA is like 27 pages long, I loaded it into a language model and said, Okay, tell me how fair this NDA is. Is it bilateral? Is that equal? Or is it skewed towards one party? The other should I sign it? And those that’s an example of question answering. And, again, across departments, if you think about it, if you have, if you’re in accounting, and you have an accounts receivable ledger, you might to a private server, you might load that ledger and say, I want now I want to ask questions of this of this ledger and say, Well, you know, what are the categories of expenses that are that are look weird. So those categories, help us understand how these tools work, because they cover so much of just the ways we use language within within business.
Katie Robbert 6:02
And, you know, I would imagine that some of the viewers might be wondering, well, why do I need to care about the categories? If I just have a specific use case? Well, if you want to catch up on older episodes, we actually think it was last week, we might have done this, we actually did a head to head to head, no, it was a few weeks ago, a few tests of the different tools in terms of the different categories of uses. So you can catch that live stream episode at trust insights.ai/youtube. Because the generative AI tools that are on the market, as of today, have different strengths and weaknesses. And so if your use case falls under generation versus extraction, there, you’re going to want to consider different tools. And that’s why understanding which category the use case falls under first, is going to help you decide which tool you’re going to need to use. So Chris, how do we get from use case categories into decision making? How do we bring it to a two by two matrix, what is a two by two matrix,
Christopher Penn 7:09
it’s a consultants best friend. No, a two by two matrix is you have two dimensions of some kind that you want to simplify things. So cost versus quality, speed versus features, you know, pepperoni versus pineapple versus pepperoni on pizza. Whatever the case is, you have to sometimes they mostly two dimensions that you want to try and how to get arise, different things on and those two dimensions ideally are scalar. Like, it’s not just an either or. But it’s there can be variations. So again, if you’re talking about things like cost versus quality, it’s not just expensive or not expensive. It’s like, well, this is $10. And the quality is here, and this is $100. And the quality is over here, and so is paying the extra $100 worth it, you know, depends on how much more quality you get for that. And so that’s, that’s a two by two matrix in its essence. And there are so so so many of these, again, this is a consultants best friend, because even though it runs the risk of oversimplifying, sometimes it can take a very difficult or complex environment, and just narrow people’s focus just to get them to pay attention to a couple things that really matter.
Katie Robbert 8:30
Well, we’ve done this exercise, we’ve actually used the two by two method to help categorize keywords for SEO. So difficulty versus now I’m forgetting I’m blanking on all your volume. And so you would want to go after, for example, high volume, low difficulty, or if you are feeling really, you know, ambitious, you could do high difficulty, low volume and try to create your own space. And so it’s a way to categorize a lot of information so that you can prioritize it. And so today, what we’re going to do is we’re going to look at the some of the use cases for generative AI in terms of optimization versus innovation. And so these are things that we’re looking internally, we’re looking at the things that we do in a business, not necessarily the things that we’re providing to our customers. But that will come into play in the conversation we’re focusing on, will this use case optimize what we do, which is to make it faster to make it more efficient to make the processes even more automated and repeatable versus innovation, which tends to be a new spin on something we’re already doing something net new, something we haven’t done before, something we haven’t had the ability to do before. And so we want to start to break down the use cases into that two by two matrix. So Chris, if I said to you, hey, I have a bunch of things that I want to do with generative AI, I’m guessing that you would say, Great, let’s put it on a two by two and figure out where it all nets out,
Christopher Penn 10:10
I would exactly do that. And I was, I would say, you know, maybe the easiest way to do that would be with something like, you know, even just putting on a on post it notes on a wall can be one of those things. So let’s bring up your standard two by two matrix, and then start talking about the processes that you have. Is it internal? I mean, it’s not going to, you know, it’s something that the company does is external means something that the company does with customers or clients, or, or vendors or lobbyists or whoever, inside or outside the company. And then to your point, is making something existing better, which is optimization? Or is it bringing something new to the table? So a real simple example, suppose you’re an insurance company, and you have elaborate insurance policies. And your your call center is constantly swamped with people saying, I don’t understand what’s going on in my policy, I don’t know if my policy does this. You recognize that that is consuming an inordinate amount of your time. But one of the things you could do with the language model is have a chat bot, right? And you’d have that chat bot for policies. Now, this is not something that’s existing, right? This is something that would be net new, and it’s something that’s customer facing. So you may put that right there on that external innovation category. Because yeah, that’s, that’s new, that’s going to be different. But and but it could be a pretty big deal.
Katie Robbert 11:34
Well, so let’s think about some of the classic use cases that we’ve been hearing, writing blog posts, writing anything, I want to replace all of my writers with a chatbot. With ChatGPT. With llama, I think that is something that comes up a lot in conversation as people, as companies are examining, should we bring in generative AI? So writing content, I, you know, I would argue that it’s an internal optimization, unless your company doesn’t currently write any content at all. And then it sort of skirts the line of optimization and innovation,
Christopher Penn 12:13
I would say, it also skirts line between internal and external, because you don’t just write content for yourself. I mean, probably not. You publish on a blog for external use for resumes, SEO benefits, maybe it’s just, you know, customer retention, or customer care. But either way, that contents gotta be out there doing something, you can’t just sit on your harddrive. Right.
Katie Robbert 12:36
And so as we’re putting the use cases here, would, I’m guessing we would have first decided which of these boxes, we would want to focus on in terms of where things land. So for example, external and innovation, I’m guessing, because it’s customer facing, and it’s something new is going to have a higher cost associated versus internal and optimization. Because it’s something that’s pre existing that we’re just making a little bit better. But these are the kinds of decisions or these are the kinds of context data points that you would want to have. Because I would imagine that this is a great exercise. But then if you don’t know which box to focus on, then you just have things in boxes.
Christopher Penn 13:24
You do. And so I would almost like in this phase of sort of your AI strategies. This is almost like brainstorming, like what are all of our processes, right? And you just start slapping stuff on the wall, and things and then you can start saying, Okay, well, now, let’s start thinking which these things are. So let’s go back to the old fashioned Eisenhower matrix, right? urgent, important. Easy, how urgent is it? How important is it? And then how easy it is to do it? So a chatbot for policies, not super urgent, although your Customer Care Center would probably disagree? Is it important for the customer? Yeah, to get right answers. And is easy, it’s not really easy to do. But going back to those three factors, it would be important to to reduce call volume, right, so that you could save some money on your call center. And also, as long as it’s giving the right answers to customers, it will make customers happier. So with any of these tasks, and again, this is something that is not just marketing, this is IT department by department, your HR department should be looking at all of its processes, right? Your sales department, your accounting department, your compliance department, your legal department, everybody should be looking at their internal extra processes and saying, Well, what do we do all day?
Katie Robbert 14:44
I can imagine a version of this a two by two where it is, you know, urgency and cost are the x and the y. And so you can start to say Is it urgent but it’s high cost, it’s urgent, but it’s low cost, and then you can start to get pick off those small wins. versus it’s not urgent and it’s high cost. Those things go on the backburner. But I think you would need to go through this kind of categorization exercise first to even know what the things are.
Christopher Penn 15:10
Exactly. So let’s think back to your, your pharma pharmaceutical days, when you were doing that kind of work, what was one of the tasks that maybe was just a pain in the butt?
Katie Robbert 15:27
We had, and I don’t so I don’t know the full way to sort of like summarize this. But basically, we had over, we had almost 10,000 data points every single month, that had to be cleaned by an SPSS syntax that was printed out and then manually processed. And so I guess data cleaning is the best way to put it. Because it was high volumes of data, we were constantly running out of server space because of how manual the process was. And so that to me, would have been an internal optimization, which would have allowed for a higher volume of reporting as it was because the process was so manual, we could only offer we like, basically hit the wall in terms of resource time, every single month.
Christopher Penn 16:21
When you said the words SPSS, SPSS is a programming language is a statistical programming language to me, that immediately screams AI, because it’s which, right and a language model can generate language, code, llama or GPT-4 can write SPSS syntax review easily. And so yeah, I mean, that would be if that’s a big pain, meaning it’s important. And you’re running out of disk space. So there’s, there’s some urgency there. But that’s easy.
Katie Robbert 16:53
Well, and you could we couldn’t scale it because we couldn’t add on more reports. You know, it’s, I don’t want to get too tangential, but I do recall, are these big binders of printouts of the SPSS syntax and someone going through with a pencil, checking it for errors every single month, because every single month when we ran it, we will get hundreds of errors in the SPSS syntax when it would pause. And then we had to figure out where it was, and then update the master syntax. It was it. When I look back, It’s bonkers. years, so yeah, data cleaning. Writing code for synth, I guess it all falls under the same category. The others. The other thing was the actual write up of the reports themselves, the reports themselves were about 50 pages long of Word docs. And so it was analyzing the data and summarizing it.
Christopher Penn 17:55
It was that word that were that tells us that’s one of the six core use case categories. Again, that goes squarely in the internal optimization, although the if the reports are going externally, it can go on that line that because it’s going to the customers are going to see it too. Oh, yeah,
Katie Robbert 18:11
they absolutely. They were external reports, not internal. You know, there was one of the things that I was charged with was helping to identify early warning signals of substance abuse. So basically, I worked within opiates and stimulants. And so looking for conversations on chat in chat rooms, that people were it was early indicators that the new opiate that just got released, even though it was meant to be non abusable is now being abused, so that we could try to get ahead of it. So I guess it’s internet scraping? I don’t know.
Christopher Penn 19:00
No, that that would be that would be I would call it web scraping. I would I would just say it’s called data acquisition.
Katie Robbert 19:07
Yeah. Again, I remember sitting there with printouts of papers of like, chat conversations and hand coding them, because we also did, we built our own sentiment analysis tool, but a lot of it was manual.
Christopher Penn 19:23
Yep. What I find interesting is we’ve got a lot of stuff in three of the four boxes. internal innovation doesn’t have anything in it yet.
Katie Robbert 19:32
Right? And so what is what’s an example?
Christopher Penn 19:37
So an example would be, maybe you have maybe your team internally, just need something that doesn’t exist. So I’ll show you real quick example here. I was working on someone asked, Hey, what are you possible to make a a mobile app can can ChatGPT Make a mobile app? So I said here’s the requirements. Tell me what the thing is the best options to framework like flutter. And here’s all the requirements for as I said, okay, cool. Flutter sounds like a good choice. Then it says, Okay, well, great. So here’s here’s the process set up this thing said, when Google Sheet, start the back end, and then it starts giving me instructions, okay, here’s how you set up the flutter project, go download it, install it, and stuff like that. Here’s how you do the Google Sheets integration. This is the code that you use. And I was doing this for about 20 ish minutes the other night, and it just just for fun, because as one does, and in that time, with its help, I built a bare bones mobile app that can work on Android and iOS. Now, again, if we think back to our chart here, this app is not supposed to be for external use, right? It’s an internal use for an organization. So something like that would be super powerful. Right? So let’s go put this year internal app development.
Katie Robbert 20:56
Well, so let me understand though, the, the ChatGPT, or llama tool didn’t build the app. It gave you instructions to build the app.
Christopher Penn 21:07
It gave me and gave me some of the code.
Katie Robbert 21:10
Instructions. Yeah, yes. Okay. But I think that that I just want to make that distinction is that these tools will not build you an app, but it will give you the blueprint to build the app at this time.
Christopher Penn 21:24
It depends on the language. So flutter know, if I had specified Python, the answer would be yes, because a system called GPT. Engineer, we give it extensive requirements, and it will actually write build and compile the code. It is a little buggy.
Katie Robbert 21:40
Well, I can’t imagine that there’s a UI for it. Like, oh, for the ability F and that, yeah.
Christopher Penn 21:47
Oh, yeah, then the system itself runs at the command line. But the the final app will have a UI, you will specify what you want in the UI and the requirements gathering.
Katie Robbert 21:56
And I think that this is another really good example as to why you would want to categorize the use cases under one of the six main categories, because not all of the generative AI tools are going to be able to do that they’re not all going to be proficient in Python, or, you know, CSS or whatever the other colbalt, all the other programming languages are, some will handle it better than others. And if that becomes a real need for your company, you want to make sure you’re choosing the right tool.
Christopher Penn 22:31
Exactly. So that’s an example of an internal innovation, something that doesn’t exist, but could dramatically streamline the process. One of the things that has held companies back and we’re in this pile of companies is you have an idea, like, it’s going to take six months to write the software and QA and deploy and stuff like that. And we don’t have that kind of time. Because we’re all busy doing, you know, paying work. With these tools and their ability to generate code, you can cut that time down by two thirds, three quarters, maybe even 10x, maybe, you know, 90 or 90% Less time, you still have to spend time on it, but you can spend less time on it, and do both innovations and optimizations. So we have for example, we’ve talked about this a lot recently in Slack, our content curation software. So we built software years ago, to do content curation. And it was always a little buggy. Actually, it was more than a little buggy. It was always a lot buggy. And then about a month ago, something crashed, I remember going and going, Ah, I gotta fix this thing. And then wait, wait, no, I don’t, I’m gonna feed it to ChatGPT to the GPT-4 model, and like fit this. I said, here’s the prompt, here’s, here’s my PHP code, fix my code. And it’s like, Hey, here’s the 18 things you did wrong. And this code I had great, just fix them all. And now, it runs so much better. It runs faster, it handles errors better, because it has error handling now, and it didn’t used to. And so even that, that’s internal optimization, we can put that in here and just sort of code maintenance. But that is an example of where even something that is so mundane, is super, super valuable.
Katie Robbert 24:26
And, you know, having run software development teams, the QA team, especially it so where I worked, we had about 20 software developers, we had three separate product lines, plus all the internal things for the company in the website. And we had two QA engineers, just too and they were responsible for testing everything that came you know, they got landed on their desk and they were always backed up because you know, despite In the product managers best efforts to stagger the different product lines, it was myself and two others. And we always work together to try to stagger so that we could share resources, things happen, things get backed up, someone goes on vacation, code breaks. And so we were always behind schedule, especially when we came to QA because our QA team was top notch, they were very thorough, which meant it took more time, and you had to get in line.
Christopher Penn 25:31
Exactly, exactly. So the other thing I’m noticing here is that the external optimization book doesn’t have anything he’s really squarely in it yet at giving that some thought about something that’s customer facing that could use some optimization, what would what comes to mind for you? I was thinking about
Katie Robbert 25:47
that, and it depends on your kind of business. So for us, I would think that an external optimization might be email deliveries, response times to, you know, increase. So you know, for example, if we have a client who’s high touch, who sends a lot of, you know, different email threads, then is there a way to use one of these tools to, you know, basically help with that comes to customer response, have some sort of acknowledgement, we got your email, the information goes into our task management system, so that we can see it and get it faster. To me, that’s an external optimization, because the customer benefits it, there’s some internal but the customer benefits mostly from, you know, that relationship building, it doesn’t replace the human, the account manager, but it helps make sure that their information doesn’t get lost.
Christopher Penn 26:52
I would agree with that. I think it’s a great use case. Another one, I mean, optimizations literally, in his name, SEO, search engine optimization, there are so many ways to do SEO with these tools. One of my favorites for things like content creation, is take the persona of the person that you think is your ideal customer, and have the tools compare that persona and what they’re likely to be interested in with content you’re creating and say, you know, what, there’s some words and phrases and, and concepts and topics that are just not in your content that need to be if you want to appeal to CEO.
Katie Robbert 27:34
I would agree with that. And so when we start to look at all of this, you know, we can see that we’re very heavy on optimization and very light on innovation. And I think that that’s one of the misunderstandings of innovation in general. I wrote about this a few weeks ago for the newsletter, if you want to subscribe, it’s trust insights.ai/newsletter. And the thing that is misunderstood about innovation is that innovation, a lot of people approach it of like, I’m creating something brand new that the world has never seen. And that’s not necessarily true. Innovation is always built on the back of something that’s pre existing, it’s finding a new solution to an existing problem. So you know, maybe, you know, electric cars that drive themselves is an innovation. But having to get from point A to point B is not a new problem. So an electric car that drives itself is a new solution to an existing problem. And that’s the difference. And that’s the distinction with innovation. So when we look at innovation versus optimization, yeah, an internal app development, that’s a new innovation, because it doesn’t exist to us. But developing an app is not a new problem, we’re just finding a new solution to an existing problem. Our customers needing more efficient responses to their policies, it’s not a new problem, it’s a new solution to an existing problem.
Christopher Penn 29:01
For those of you who are not word nerds, innovation literally comes from Latin from nfra, which is broken out in pieces means to renew, to make something into something new. So it’s not making a new thing. It’s making something as old into something new.
Katie Robbert 29:18
And so with that distinction, it does not surprise me that the majority of use cases would be optimization versus innovation, because what we are all collectively trying to do is squeeze more things into the same 24 hours. And so using systems like ChatGPT or Allama, or Stable Diffusion or whatever, you know, helps us get there faster, more efficiently. And that’s the optimization piece. You’re muted.
Christopher Penn 29:52
So if you think about it, if you think about those, those six use case categories, those six use case categories really are about out working with data you already have, right? So generation, not necessarily generation is, hey, write a blog post. And you still have to provide a lot of information about that in your prompt. question answering is actually a question answering even that you have to provide some data, you have to ask questions of something, you can ask them a model generically, but you’re not going to get as good results as asking a specific knowledge base. The other task extraction, summarization, rewriting classification are inherently taking stuff you already have, and doing something with it. So by definition, like you said, that innovation is going to be coming from transforming something old into something new.
Katie Robbert 30:38
And so when we think about, so we’ve been plotting these use cases on the two by two, you know, some, some of these are going to fall in every, you know, the internal optimization bucket. And so you could look, you could look at this and go, okay, but that’s everything falls into the same bucket. Now, what do I do? So the next step that you would take is you would take those pieces that fall into that quadrant, and then start to, you know, dig deeper, like, okay, so is data acquisition a higher priority than code maintenance? Or is, you know, data cleansing, more costly than writing content and finding additional ways. So you’d start, as Chris is doing, you’d start at second matrix, and you would start to figure out the categories. So we’re going to do cost. Versus I would assume, value, importance? Yeah, high importance, low importance. And that makes sense. So things that are low cost, high importance, are, what we would want to start with those are, those tend to be the quick wins. So they’re not going to cost you a lot of money, but you’re going to get a lot out of it. And so So let’s put all these everything that isn’t in the internal optimization.
Christopher Penn 32:00
Well, let’s take all these things off the grid here. Now. So let’s take some of the data cleansing the the idea of of saving that time, it’s important, right? Because it was cost you an enormous amount of time. And because SPSS is language based that goes, right, they are high importance, very low cost, because that’s an easy win, right? Code maintenance, also an easy win. You know, it depends on on where your code fits, like how important code is, but like that stuff of for our content curation software, again, an easy win, the machines can do that.
Katie Robbert 32:33
I would also imagine that it also depends on how big your code base is. And the same is the same as how big your database is that your data cleaning. So, you know, when I think back to that example, you know, we had, you know, 10s of 1000s of rows of data, versus we’re pulling this month out of Google Analytics, those are two very different datasets. And so we may have to start to break it down further of like, the time invested. So it may feel like it’s going to be low cost. But if you’re going back through months and years of code or data, you’re going to start to increase the cost.
Christopher Penn 33:10
Exactly. chatbot for policies, I would say, you know, that’s of medium importance, but the cost is pretty high. Because you have to, you know, do the embeddings, build a vector database, all that stuff? There’s a lot of infrastructure cost that goes with that. So that’s, that’s fairly high cost.
Katie Robbert 33:29
And I would say internal app development is low importance, high cost.
Christopher Penn 33:33
Yep. I would agree with that. That’s that that is, our employees not quitting because of that. writing content? Is I’d say, probably, you know, it’s it, depending on how you’re doing it, it can be somewhere along here.
Katie Robbert 33:51
And I yeah, I think it depends on what content you’re writing. So if you’re using the tool to write, you know, a 300 word blog post, you know, it’s not going to cost you a lot. But if you’re using it to write an academic paper that’s thoroughly researched and say cited, then you’re going to have higher costs associated.
Christopher Penn 34:09
Exactly. customer response, for example, you were talking about to do the way you were talking about require you to build, basically build an AI app with length chain, and start chaining models together, because you’re gonna need a couple, a few different models to accomplish few different tasks. So I will put that on the higher cost side. So but how important is that to you?
Katie Robbert 34:31
Given that we are a service agency, I would say it’s a higher importance, because without customers we don’t exist.
Christopher Penn 34:37
Right. Okay. So we’ll put it there. SEO right now, where would you put that?
Katie Robbert 34:46
I would put that in the upper half of the low importance. And I would keep it I think it Yeah, I think it’s lower cost, because it’s a lot of topic generation keyword research. Things that historically, you know, aren’t big cognitive overload, it’s just a matter of finding the time to do it.
Christopher Penn 35:09
Hmm. What about data acquisition that scraping you’re talking about?
Katie Robbert 35:14
I would say that that is, I’m gonna guess that that’s high importance, but also high cost.
Christopher Penn 35:22
It? It depends. I would put it here, because it depends on the kind of data. But yeah, it could be there. And then report generation just streamlining those reports,
Katie Robbert 35:32
I would say that’s high importance, low cost,
Christopher Penn 35:35
I would agree. So what I find interesting is, in the internal external innovation versus optimization, you’ll definitely have things weighted towards one half of the board, when we put it in terms of cost and importance, we’re actually pretty concentrated in low cost high importance quadrants. So A, that tells me we did a good job of identifying tasks that are actually meaningful. But B, that tells me, there’s a substantial opportunity for AI at our company.
Katie Robbert 36:03
And when you start to look at the tasks themselves, you can see there’s a lot of overlap of the kind of thing that we’re needing to do. And so that’s where I, you know, sitting in the CEO seat, I would say, Hmm, what’s going on with our data team, that these are all the things that we need to be focusing on first? You know, are they, you know, overwhelmed with the amount of things? Or is this truly just an opportunity to bring in some automation? You know, and I’m not saying like you specifically, Chris, I see you sort of smiling, I don’t I don’t mean, you specifically. I mean, like, if I were objectively looking at this, as someone who isn’t mean for not Trust Insights, I would say, hmm, I’m seeing all of our developer work, all of our analysts work, those are the things that I should be paying more attention to? Do they have repeatable process? How much volume of things are they generate? Am I getting what I’m paying for? And is this a really good opportunity to bring in a lot more artificial intelligence?
Christopher Penn 37:03
Exactly. Katie, what would you say it is that you do?
Katie Robbert 37:09
I, I just bark orders and boss people around and get cranky. That’s all I do.
Christopher Penn 37:17
But no, I agree. There’s a lot of things here where and again, this is, this is an exercise that every department in every company should do. It’s not just marketing, it’s not just engineering, it’s not just it, everybody, HR should be doing this, because this is where you’re going to see like this, this sector here, the low cost, high important sector, if you can knock out these five tasks here, your business is gonna grow, your business is gonna grow, because you’ll be much more efficient that things that right now consume, a lot of time, they consume a lot of resources. And they don’t have to, you know, AI may not be able to do it all. But AI will certainly help the workers who are in those departments do their jobs better and faster.
Katie Robbert 38:00
Well, especially if you sort of if you take like data cleansing, for example, there’s a lot that could go into that. So if you start to pick apart the actual steps that go into the data cleansing, then maybe it’s just, you know, having a I do the SPSS syntax, versus, you know, formatting, everything, like it really depends on what the task itself is. So I think that it should be looked at not as, Oh, crap. Now this task is going to be taken by AI. But how can AI enhance this task so that I can focus on internal app development, so that I can focus on Chatbots for policies because I’m no longer in undated with report generation?
Christopher Penn 38:45
Exactly, that’s a great way of looking at it. If you can knock down this stuff that you know, is consuming time and money and people, then all the stuff that’s on the board will get taken care of, it will actually get the attention it needs.
Katie Robbert 39:00
I would agree with that. And so Chris, what? So it sounds like the first step for people who are interested in this kind of an exercise is to understand the six types of the six categories of generative AI uses. And so that’s number one. Number two is to then go through the exercise. You know, don’t worry about getting it right or wrong, but just start to list out the different things that you think AI could be doing. It doesn’t matter if it’s a realistic thing right now, it could that could go into innovation. Or if it’s something you’re like, if I could just get this machine to write these reports for me. I would be 10% less cranky. I would say that’s high importance.
Christopher Penn 39:46
No, I think that’s exactly this is to look at his processes and say the question we asked when we do consultations like this is what processes are consuming the most time that are repetitive like every month I have have to do this. Okay, that that’s anytime I hear the word repetitive. That’s like an alarm bell saying you need to be looking at a technology platform coupled with people and processes to reduce the burden that that repetitive process imposes on your organization. That’s like the red flags. Okay, the the one that you say, every month or every week or every hour, okay. That’s something that that we should be digging into. If you also that comes along once a year, like okay, you know, maybe maybe not depends, again, depends on the size of the task. But those monthly reports, those cube ers, those things that we everyone is saddled with those daily roundups, exactly all that stuff. That is that is an immediate sign that there is some kind of process and platform that you can bring in there that will reduce the burden on the people.
Katie Robbert 40:56
And if you aren’t even sure where to start, then we can help. You can reach out to us at trust insights.ai/contact where John Wall who is not on this call, today, our chief statistician slash all around swell guy will be the one to talk you through what’s possible. And if you just want to sort of hang with people who have similar questions to you, then you can join our free Slack community analytics for marketers at trust insights.ai/analytics for marketers where you can ask and help answer other people’s questions. I think, Chris, we have what 3500 People in there now about that?
Christopher Penn 41:34
Yep, exactly. No, I think those are, those are good places to start. But when we’ve done this consulting for others, the other thing that doing this sort of work does is really highlights opportunities for collaboration. And if you can use machines and technology to streamline that, to accelerate the speed at which you collaborate, you will get more done and be much more productive, right? It’s it sounds so cliche, as as consultants are want to do. But it’s true. Knowing that this department and that department and this department all have exactly the same task and slightly different disciplines means that all three of those functions should be able to learn from each other, and should be able to share best practices internally,
Katie Robbert 42:20
is the reason that the phrase many hands make light work has stuck around for so long.
Christopher Penn 42:27
That’s true. That’s true. All right, Katie, it looks like folks have their homework cut out for them. And again, as mentioned, if this is something that you want to have done at your organization, we do this in half day and full day workshops and stuff. Let us know any final parting words.
Katie Robbert 42:45
Start planning, start experimenting, start brainstorming. You know, this is a great exercise to do ahead of selecting a tool because you don’t want to pick the wrong tool and find out oops, I just spent a lot of money on something I can’t use.
Christopher Penn 42:59
Exactly. And we’re recording this in September. This is the time of year when folks start thinking about annual planning and budgeting and forecasting and what the year ahead is going to look like. So this is the this is a great time to be doing this kind of planning and analysis as well. So thanks for tuning in, folks, and we will see you next time. Thanks for watching today. Be sure to subscribe to our show wherever you’re watching it. For more resources. And to learn more. Check out the Trust Insights podcast at trust insights.ai/t AI podcast and a weekly email newsletter at trust insights.ai/newsletter Got questions about what you saw in today’s episode. Join our free analytics for marketers slack group at trust insights.ai/analytics for marketers See you next time.
Transcribed by https://otter.ai
Need help with your marketing data and analytics?
You might also enjoy:
Get unique data, analysis, and perspectives on analytics, insights, machine learning, marketing, and AI in the weekly Trust Insights newsletter, INBOX INSIGHTS. Subscribe now for free; new issues every Wednesday!
Want to learn more about data, analytics, and insights? Subscribe to In-Ear Insights, the Trust Insights podcast, with new 10-minute or less episodes every week.