Jeff Standridge:
Hey guys, welcome back. This is another episode of the Innovation Junkies Podcast. I’m Jeff Standridge.
Jeff Amerine:
And this is Jeff Amerine. I think we’ve got a great episode today, Jeff.
Jeff Standridge:
Man, we have our very first three-peat, right? So Tim Creasey is a three-peat episode for us. Many of our longtime listeners heard him in season one on episode 36, where he actually, I take that back. It’s episode 25, where he talked about the workplace of the future. And then season one episode 41, where we talked about innovation and change management.
And today we’re going to be talking about the use of AI. We’re probably going to break this into a couple of different episodes. So our first episode, we’re going to be talking about the use of AI kind of in general, kind of setting the stage and then talking about where Tim and his team is seeing AI being used in the areas of strategy, innovation, market research. And then we’ll, we’ll tell you a little more about the next episode a little later. Tim, thanks for joining us.
Tim Creasey:
It’s an honor to be back and it’s a real honor to be back for the third time. Thank you for having me once again.
Jeff Standridge:
Well, you are always a very popular podcast guest. And not only do we enjoy talking to you and learning from you, our listeners enjoy hearing from you and learning from you as well. So thank you for taking the time to be with us.
Tim Creasey:
My pleasure, my pleasure. Hopefully they’ll find something that they find useful on the other end of the podcast.
Jeff Standridge:
I am, I am confident they will. So, so tell us a little bit about you, just to remind our listeners a little bit about you and your organization and what you do for a living.
Tim Creasey:
Yeah, so Tim Creasy, Chief Innovation Officer at a company called ProSci. We focus on how organizations deliver more successful change by understanding what gets in the way and coming up with better strategies to navigate that. So the people side of change is one side of the change coin that’s often been ignored or underserved, underutilized. And so we’ve spent a couple of decades researching and then building training, advisory, and solutions around how do you drive adoption and usage of your solutions so you actually achieve change outcomes. Threading the needle of helping our people be more successful by preparing, equipping, and supporting them, which delivers better projects and better organizational outcomes.
Jeff Standridge:
And so how do you engage with your clients? What’s the most common engagement look like?
Tim Creasey:
Yeah, so ProSci at its core, right, there’s a big training wing where we’ll go in and help organizations build those internal skills and capabilities at Navigating Change, whether it’s change practitioners, senior leaders, middle managers, frontline employees. And then we have an advisory service that comes alongside project teams to help them build out and navigate that challenging people side of change with expertise and a tool set derived from the ProSci methodology and the ProSci research. I play a role in the innovation and research side of the house, so really trying to figure out what’s next to help organizations achieve successful change.
Jeff Standridge:
Very good, very good. Well, let’s hop into a little more about AI. So we have a pretty sophisticated listener base, but we don’t want to assume that they know everything there is to know about large language models and machine learning and AI. So why don’t we start with just a little bit of a primer and then hop into some of the ways that you’re seeing AI being used in strategy, innovation, market research, et cetera.
Tim Creasey:
Yeah, very good. So I still remember the email I got from my head of digital product beginning of December 22 that said I’ve never been more scared and more excited than I have been right now in terms of what my role is going to be as this new technology unfolds. Oh, it’s interesting. I rewatched that video here recently and in five minutes, you can feel the concern about what’s on the other side of this new technology.
Now when you talk to him, he’s just bubbling over with excitement in terms of what we can do once we unlocked the power of those zeros and ones. So, you know, I think in the end we’re doing word probability math, right? It’s really high-end probability math done on a bunch of zeros and ones that we’re finally able to do now that we have the capability to do so. And once we can see zeros and ones in a whole new way, we can make good on them and bring them to life in different ways. So I talk about, you know, it is that word probability math and then really how do we train the thing we’re engaging, the system we’re engaging to generate the kind of solutions we want on the other end. So I have to do this kind of fun, you know, Mary had a little, what’s next? Lamb, right? If you grew up in Western intelligence, lamb is the probability you come up with lamb is 100%.
Imagine we got this brand new AI bot though, and it’s standing outside the written word, the library of all the written word. It’s not read any of it though. Mary had a little what? Chance it comes up with lamb, one in as many words as there are, right? Now you give it one book, you give it 10 books, you give it a thousand books, you give it 10,000 books, you give it the whole library. Mary had a little, right? It comes back with lamb just like that. But what if instead of giving it the whole library, you give it just the medical wing?
So it’s only read the medical wing, just a bunch of medical books. Mary had a little, what would it come back with? Cough, appendix, procedure, complication, right? Because all we’re doing is word probability. Now, if you just give it the 16th-century European history section, this is a Mary Queen of Scots joke, all right? Mary had a little time to get ready for the throne. Mary had a little son named James. Mary had a little harp. We’re just engaging a new tool to access the digital universe in a different way. And so it’s fun to watch people start to engage it and to see what’s on the other end of using it in new ways.
Jeff Standridge:
So what are you seeing out there? That’s a good start. What are you seeing out there as you work with other organizations? And I know we’re gonna talk a little bit in maybe the next episode in some of your own research into AI, but right now, what are you seeing out there in terms of common use around, say, innovation?
Tim Creasey:
Yeah, great. And so I think just another kind of frame here, we talked in the audience that we engage with. These are change practitioners trying to drive successful change in their organization. We asked the question, you know, what percentage, how many of you are focused on how to use AI to do your job better? And I’d imagine, right, part of your audience is trying to figure out how to use AI to do their job better. How many of you are focused on helping your organization build its strategy for bringing AI to life? And then how many of you are doing the change management side of an AI deployment? So you’re rolling out some big AI solution and you’re helping your organization figure out adoption and usage. It’s about two thirds figuring out how to do their own work. About a fifth helping organizations build that strategy. And then only about 12% that are kind of doing the change management work on AI deployment. So I do think we’re pretty early in organizations trying to figure out, um, how to bring this technology to life. One of my frames I always bring to the table is AI is an intern, not an or, so we’ll play with that for a little bit before we get into the innovation. What can it do for us in innovation? But AI is your intern, not your Oracle. You go to an Oracle for answers. You engage an intern to help you do activities that were going to take you too long. That was going, it was going to, you weren’t quite sure how good you were going to do it. And so even when I’m talking about people about how do they engage AI in their day-to-day work, it’s think about this as an intern, not an Oracle and what work did you do yesterday that if you had a really skilled intern who has read everything in the library on the topic you want them to help you understand, they would be ready to help you execute in a whole new way. So that AI as your intern, not your Oracle is kind of the foundation. So you go into innovation, right?
Jeff Standridge:
Yeah, I really like that. I really like that comparison because as I think back on, and I’m in AI fairly regularly, and when I think back on how I’m using it, it is kind of my intern. I’ve thought of it in terms of a virtual assistant, right? A virtual research assistant. And so I’ve used it very, very much in that way. So I really liked the comparison.
Jeff Amerine:
Thank you. And Jeff, to follow up on that, as we think even about some of the companies that we’ve invested in through our fund, and Catherine Capital Partners, there’s one in particular based out of the valley called Omnikey, which I would describe in sort of layman’s terms, it’s a marketing, a junior marketing staff person in a can, effectively. So it can do really good content creation, it can do good imagery, it’s a time saver agency, but it’s going to make that agency exceptionally more productive. And if you think about it from our perspective, the reason why we thought it was an interesting investment target is with the labor shortage on high skill, and that’s true throughout the rest Western world, this is a force multiplier where it can allow you to do these things that are time-consuming a little more efficiently.
Tim Creasey:
Absolutely. That in turn, not your oracle that you can put to task on these kind of tasks that will help us get that force multiplier. One of my favorite takes on AI was Rick Rubin on the Tim Ferriss podcast. And so, I don’t know, you know, Rick Rubin, legendary American musical producer. He just produced one of my favorite bands, albums. It’s coming out May 15th of 2024. But, so there’s Ferris, he says, you know, you’re a creative, I come from a family of creatives, what do you think about AI and what’s it gonna do for creatives? And the analogy he gave was back in the day being a hip-hop music producer, and you would do these crate digs, right? You get a whole crate of music, and you’re just popping in music, tape after tape after tape after tape after tape. And he said, you’re not looking for the next big artist or the song of the summer, you’re looking for a moment, a unique take that you can pick up and helps you see the world in a different way. It’s that moment of inspiration that hip-hop producer was looking for. And you get that from sample size. And one thing that generative AI can give us on the innovation front is crazy sample size. For us to start to derive those moments that we can weave together to get that, you know, truly amazing outcome. So, it doesn’t surprise me that you have creatives leaning in to leveraging this unique power of leaving it together. I give you one other way that I’ve been playing with it personally, but I think there’s an innovation creativity flair on this. I bought myself an X tool. It’s one of these laser engraving tools. You can engrave on wood and leather and all this stuff. I started doing Black Slate Coaster because I liked it a lot. It kind of came out really crisp. So I began by going into generative AI and saying, give me, what kind of image composition burns best on slate coaster? Tells me exactly, I’m like, great, we’ll refine that. We made that our style guide for creating imagery. And then I started doing, I had some family here that lives on the ocean. They said, oh, harbor seals, everybody loves harbor seals and the faces they make. I said, I can make harbor seals making funny faces, gendered of A, I can do anything, right? So I’d say, make me a really happy-looking harbor seal. And I get a happy-looking harbor seal and I burn those on coasters and they were kind of neat.
And then I stopped doing it for a couple of weeks. I started teaching people AI. And then I came back to it about three weeks later and I was like, I want to get a really amazing look on this, you know, not just a happy animal, but something crazy. So I said, show me a spider monkey that looks like a Red Sox fan in 2004, right when they broke the curse.
Jeff Standridge:
Hehehe
Tim Creasey:
And what came back is this spider monkey. I’ll send you all a picture of it. It, it’s not just an excited spider monkey, right? It is the layers of 86 years of pain, generations of her, like being down 03 and then coming back, like all of that starts to show up in this expression. Um, in a fascinating way, because you talk about using generative AI and innovation, it can weave together stuff. That it doesn’t produce, it doesn’t bring you forward an answer, it weaves together things that it makes sense of given the richness of language we’re engaging it with. So we have folks that keep coming back to me on this thing we’re working on, they say where does the answer come from? And I was like there’s not an answer, there’s answer fodder, which is the stuff that’s on the shelves that we’ve been feeding it and teaching it about. All that answer fodder gets brought together in kind of a zeros-and-ones stew, and based on how good you instruct it, you get something really powerful back.
Jeff Standridge:
So you just, you just hit on how good you instruct it. So the prompt engineering, so to speak, I think is the term that I’m hearing out there now and, and having played around with it quite a lot, you know, I’ve, I’ve gotten pretty, pretty astute at if, if I’m wanting to do some research with, with AI to actually name some of the authors, some of the models, some of the, you know, some of the citations and what have you that I, that I want to be part of that, and it, that seems to have helped as well. So I’m, I’m learning a lot about how to structure my own prompts in order to get what I want on the back end.
Tim Creasey:
I’ll give you a couple of thoughts on kind of prompt engineering. One, it’s a guy, I think his name is Conor McGregor, he’s the UFC fighter, it’s something close to that. And he said, the word prompt engineering, they gave it the wrong label. That makes it feel really sophisticated and hard like engineering is. They should call it speaking human again. Because that’s the thing is you learn, you have to unlearn. And I have a slide that I use in the presentations when I’m teaching AI that talks about how Google broke us by teaching us to think in keywords. And so I use this example that’s like turkey, turkey steam oven, turkey steam oven, Thermador, turkey steam oven, 15 minutes, right?
We were taught to break down what we were looking for into keywords and sequence them in a way that hopefully gets us back a list of things for us to go look through. But if we can begin speaking human again, it takes us back to actually engaging with all those zeros and ones. It’s all the same zeros and ones that were on the webpages that were on that list that came back, but it comes back with specific instructions for how to cook my 15-pound bird and have it done at 3 p.m. And so, now take this one step further, right? Your intern, not your Oracle. I’m sure you’ve ever heard people say, I don’t know what to ask AI. I don’t know what to ask it. Well, what do you, you ask things to an Oracle. What do you do with an intern? You sit down and describe for them, here’s what I’m looking to get done. Here’s the steps I’d like you to explore. Like engaging with, go ahead, sorry, Jeff.
Jeff Amerine:
No, no, I was just, I’m agreeing with you. It’s, it’s forcing you to write more descriptive pros, you know, good, well constructed sentences that are, that specifically describe what you’re trying to achieve. I, I’ve, there’s such a remarkable difference as you iterate through and spend a little more time being descriptive in terms of the output you get, whether it’s images or written, uh, context for sure.
Tim Creasey:
Yeah, language is the richness, that’s the currency of engaging with generative AI. One other analogy I use when I start talking about AI as your intern, so I’ve got a kind of visual that I use, if here’s your intern, we imagine them like we’re commissioning some art from our intern, and they have a canvas and they have a palette. The palette is blank and the canvas is blank, and your job is to help define the canvas and put paint on the palette, so that what comes out is as close as to what you want as it’s going to get. So I walk people through this analogy, right? Write me a letter. That’s a pretty loose canvas. Write me a one-page letter. Write me a one-page letter to my HOA. Man, I’ve added a little bit of paint to the palette. Write me a snarky one-page letter to my HOA. Do some research about the height restrictions on holiday decorations in my city and then write a snarky one-page letter. Each time we’re defining the canvas or adding more paint to the palette, we’re getting a better quality output from the tool reasons.
Jeff Standridge:
What are, what are your, some of your favorite AI tools and are they desktop? Are they, you know, uh, iPad? But so what are you using?
Tim Creasey:
Yeah, good question. I personally primarily use ChatGPT. I leaned into it pretty hard in March of 2023, pretty early on. We were using it to do some interesting pattern extraction from de-identified data sets. And then I stayed away from it for a little bit. And then when DALI 3 came out, first time I used it, I pulled over to be safe into a parking lot of a gas station. I went into DALI 3 and I bought my five credits and I said, draw me a picture of a personified alarm clock walking out of his child’s bedroom, looking as if he just got beat down in a heavyweight fight. Right, because that’s pretty much what every morning felt like for me, an alarm clock that just got beat down by a high schooler. So that was my first experience of like tapping into image generation because I can’t even draw a stick figure. And then started to…
Jeff Amerine:
Now, DALI is amazing. DALI is amazing. I mean, I spent quite a bit of time in that one. And if you’re not an artist, you have difficulty painting, drawing, or whatever, and you want to do something that’s in a particular style as well, you can put the style description in there and make it look like something Salvador or DALI would do or surrealistic or whatever. And it’s remarkable. The kind of stuff it comes up with is like, you couldn’t pay somebody to come up with something that’s as good as some of what it will generate. It’s pretty spectacular.
Tim Creasey:
But I fell into a pretty deep hole. So the first thing I did, I was doing a webinar about the biggest changes on the horizon facing our organizations today. We did research, 700 data points, right? Technology and digital transformation, regulatory compliance, sustainability, talent, culture, customer expectations, right? Those are the big six. And I thought, wouldn’t it be fun to illustrate them visually? Because visuals are powerful, right? They create meaning and connection to the story. So I go open up GPT late one night and I said, can you do me a 1960s-era superhero comic that depicts digital transformation and technology? And I get this really cool image, right? And so I ended up building a whole set of these superhero comics. Then I did some flat icons for each of them as well. Then I did adult coloring books. So I got adult coloring book pages that kind of mapped each of them.
I’m driving my son to school one day, and he goes, well, if it can make an adult coloring book, can it make one for our niece? She’s six years old. She loves unicorns. And so I get home and I teach it. I actually created a MyGPT that does this. I don’t know if, have you leaned into the MyGPTs where you can program them? Let’s keep talking about that if you want to, because this one I built, it’s called a kid’s coloring book maker. And so you tell it the age of the kid and the recurring character. And it goes and says, well, based on the age of the kid, here’s the right complexity of a coloring book page. And based on the recurring character, it actually comes up with two of them. And you get to pick which one you like best. This actually was pretty hard to program. It does a character study on that one so it can recreate a character. And then you can take that character and put them in different situations. So here’s my six-year-old niece loves unicorns. There’s a unicorn doing judo. You know why? Cause she does judo. Is there a unicorn going down a ski hill?
There’s a unicorn on a stand-up paddleboard. There’s a unicorn scuba diving, right? You’re not gonna find, and then her little brother, he’s four, so his is a personified letter J, just for his name. So there’s J fishing, right? There’s a J going down on a sled. And then I made one with the letter J and the unicorn each separately in a boat with grandpa, my father-in-law.
And guess who had a tear run down his cheek when he saw the personalized coloring books, right? This is the power of generative AI when we start to engage it with the instructions around what we want to get out on the other side. Here’s the canvas. Here’s the paint on the palette. So yeah, it’s been really wild. So that was kind of my November kind of really got into that and got into emotive animals. But yeah, the richness of how we engage it, what we ask it for, what we tell it, we want it to take into account.
That’s kind of what we get on the other side.
Jeff Standridge:
Yeah, I’m using a, an iPad app called a Chatsmith. It started out with the name chat AI chatbot. I think is what it was originally called. Then it’s now kind of uses the term chat Smith and it’s got a chat GPT 3.5 and 4.0, but it also has tasks that you can plug into, you know, you can chat with it, just open-ended, you know, uh, prompt, or it has specific tasks, academic research, grammar, you know, multiple different tasks there that you could go in that are more specialized and it’s been very, very helpful for me.
Tim Creasey:
Yeah, and this is where I’d encourage you to lean into the MyGPTs, right? Cause I have the children’s coloring book one. One of the other ones I made at the beginning of this year is for nonprofits to write grants. So I did a one-hour session to help them because there’s money out there to be accessed and there’s people looking to do good with it. And there’s this thing called writing grants that sets between them. Uh, and so I wrote this whole kind of one hour webinar about your AI intern. We kind of laid that down as the foundation.
And then how do you use that in turn to write grants with you? So I built a myGBT where you put in, here’s the submitter. So the nonprofit that’s submitting the grant, here’s the funding provider who we’re going to submit the grant to. It does a bunch of research on both, writes a synopsis, writes a profile, weaves them together, identifies points of integration around where their missions overlap, and then writes really well-worded, well-researched first drafts in a flash.
And so you talk about where is it valuable for innovation, connecting those dots, grabbing all that content, weaving together, we’re the ones that get to tell it what to put in the stew pot, but it stirs that together and cooks a stew that we couldn’t even think about cooking in a flash. So first drafts in a flash, that’s one of my other go-to taglines you’ll hear in a bunch of the pro side content around AI because, and that’s time-freeing, right? It’s not time-saving. If you’re like, ah, it saves time, no, no. If it was gonna take you an hour to write a first draft of an email and it took you five minutes, you have 55 minutes to go do human work, to go connect with people, to go shake hands. So I’m definitely not one of these like AI can save the world, but if we figure out how to use AI, we are gonna free up some time to do some really, really cool things.
Jeff Standridge:
Very good stuff, very good stuff. So any other, as we kind of start thinking about ending this episode, we’ve been talking about what you’re seeing in terms of the use of AI in the areas of strategy and innovation, strategy development, innovation, any other things notably that come to mind before we land the plane and then start talking about it specifically to change management.
Tim Creasey:
Yeah, I’ll give you one kind of turn-a-phrase around strategy in particular, around how even internally we’re starting to really leverage it. Um, and I use this notion that AI helps us collect the dots. The team is who does the, is there to connect the dots. Right. And so a lot of strategy development and I think innovation, uh, research and development is around how do we collect all of those dots? All of the disparate data that sets maybe not where we think it is, where we haven’t normally looked, how do we collect all those dots and pull them together and start to say, which ones do we connect to create unique meaning for where we’re going to take this organization? And so I’m never one that thinks the human being is going to be replaced because it’s what each of us bring to the table that lets us connect those dots in a unique way, but… in about 15 minutes, I had 192 dots, right? 12 dots per 16 really key strategic areas, each of them with really nice crisp write-ups around the role they could play in the next five and 10 years strategic plan and the direction of the organization. So, and that gets back to sample size, like we talked about in Rick Rubin, right? Collecting of the dots within the effort of strategic planning. So, that’d be my kind of last turn of phrase area.
Jeff Standridge:
I had a couple of clients, a couple of different clients that are in relatively small healthcare organizations, rural health center type organizations. And, and, um, uh, one of them was using a relatively small IT-managed service provider, and there was really no service level agreement between them. And so I just wrote a prompt to say, I’d like a service level agreement, a simple service level agreement for IT managed services between this organization and this organization, named both organizations and, you know, it got me 80% of the way there, maybe 90. Right. Uh, another situation where a particular organization needed to hire their first CFO and wanted both a job description and a set of questions specific to the kind of organization they are that has a lot of federal requirements, federal reporting requirements. And again, 80, 90% of the way there. Um, but again, but again, it’s very dependent upon the quality of the prompt, right?
Tim Creasey:
For sure. And in those situations, did you tell it what kind of expert to be? That’s one of, uh, that’s one of my key go-to tips when you’re engaging GPTs. Act as a, an executive recruiting expert and write me the CFO job description. Act as a social media marketing expert and write me the script for this, that, the other. Um, so.
Jeff Standridge:
Tell me what you mean by that.
Tim Creasey:
So, out of the gate, I tell it what kind of expert and you can even blend them, right? My child coloring book one, that one is programmed to be a child coloring book expert with an adult or a child education, early childhood education degree as well. So as soon as you tell it what type of expertise, and we go back to our library analogy, you essentially say this wing of the library, make sure you know that wing really, really well, because that’s the kind of work we’re going to be doing together in turn. So that would be just kind of a fun tip for your listeners. Act as this kind of expert. You want one more prompt tip that I always use?
Jeff Standridge:
Yeah, we’ll do one more, and then we’ll land it.
Tim Creasey:
This goes at the end of the prompt. So the beginning of the prompt is act as this kind of expert. The end of the prompt is what other questions do you have for me before we begin? So I tell it what I think I want to have happen. I give it the details I think is going to help get us where we want to get. And then I say what other questions do you have for me before we get started? And sure enough, it’ll come back with two, three, five questions that said, give me a little bit more detail here, and here, and we are going to nail this. So that’s kind of my sandwiches, like how I kind of sandwich prompts. You put those two in play and it’s going to be crazy the way you’ll start to get some outputs.
Jeff Amerine:
I love that.
Jeff Standridge:
So play that, I’ll play that back right quick. Two kind of key tools or insights from using AI. Number one, sandwich your prompt between clearly defining the role or the expert that you want the AI to act as, describe your output, your prompt, and then end it by asking, what other questions do you have of me before you begin the work? Very good, very good. Jeff, any other questions, comments?
Tim Creasey:
There it is.
Jeff Amerine:
That’s awesome stuff. No, it’s fantastic. It was just, I’m spooling away here thinking about ways to apply some of that. And it’s, uh, I really think you’re only limited by the degree that you can be creative and how you can apply it. And really, I don’t, I don’t think we’re getting away from having to be a cognitive being that thinks things through. I think it’s a different kind of approach that we take to using tools we didn’t have before to come up with better outputs and outcomes.
Jeff Standridge:
Guys, listeners. We’re talking to Tim Creasy on the Innovation Junkies Podcast. As I said earlier, Tim is our first three-peat guest. He was on season one, episode 25, where he talked about the workplace of the future. You might wanna check that one out. Also check out season one, episode 41, where he talked about innovation and change management. And I’ll go ahead and put a plug in for our next episode where we’re going to be talking about, maybe doing a deep dive on some of the work and research that Tim’s involved in and looking at the use of AI specifically in the areas of organizational transformation and change management. That work, Tim?
Tim Creasey:
Sounds good. Thanks for having me, Jeff.
Jeff Standridge:
Very good. Good to have you. Thanks for being here. This has been another episode of the Innovation Junkies Podcast. We’ll see you next time.