But what I've seen across all organizations is like the number one thing is helping educate people on what it is, what it isn't, and what it can and can't do. And you've got to get it into people's hands. Because when people start playing with this stuff and play is a really key word here that I'm finding out, just having fun. Like we don't need to go automate this process end to end. Like let's just go have a conversation with it and see what it can do. Lightbulb start going off. That's that like little extra space that you now have with people to insert new possibilities that maybe they couldn't have seen before. Today we're joined by Adam Hoffman, who leads Elixir's generative AI practice. Adam's journey from entrepreneur to AI strategist gives him a unique perspective on how businesses can move beyond experimentation to capture real value from generative AI. From his early days of building chatbots, now architecting AI strategies for Fortune 500 companies, Adam brings practical wisdom for how to get started on the journey. Adam, welcome to Expert Intelligence. Thanks so much for having me. Wow, what a generous introduction. You've had an interesting journey. You started off with doing chatbots for travel and trying to figure out how to use technology to help people get to outcomes or solutions faster. What was your aha moment with generative AI? Oh, that's a good question. So going back to the chatbot thing, I think my career, I've really spent it at the beginning of kind of the emergence of new technologies, whether it was like edtech or peer to peer marketplaces, things like that. And with the chatbot thing for travel, we were actually trying to build a marketplace that connected people with other people while they were traveling to get local travel advice. It just turned out that getting people to connect in real time was like a little bit challenging. So we were trying to figure out if there were some ways to hack it. And while my first big moment wasn't with anything that was close to generative AI, it was just a bunch of rules that were hacked together inside a slack. But I opened up slack. I typed in a couple of words about a pizza that I wanted to order. Domino's had just made their API available and I just sent it in and waited and a pizza showed up. And I was like, oh, there's something big here. But of course, this is like 2013 and the tech is still rule based garbage for all intents and purposes. But my real moment with generative AI with response, we were trying to build an AI for salespeople that would help them manage their pipeline better. And do everything proactively and connect up to their CRM and do all that kind of cool stuff. And the beginning, we still had to do a bunch of rules. But then I remember the first time that we were experimenting with one of the early GPT models from open AI and it wrote the query into Salesforce on its own. And I was like, oh, this is going to be different. And then of course, I think like everybody else using chat GPT for the first time was like, okay, no, we're actually going somewhere this time. So that was it for me. You talk a lot about present bias. Tell me a little bit about how you think of present bias and how people can start to imagine what is possible in five years or even three years. It's funny. You get a bunch of hype headlines that make you think that this AI stuff that we have now can do pretty much anything. It'll take your job, make you breakfast, help you find the next best place to get a haircut. And then you go and you use it and you're like, oh, is this thing really that good? Are you sure? And then if you stick at it, the next experience you have is, oh, this thing could literally do anything. It actually might take my job. And I'm just talking about people who use chat GPT for the first time. I mean, I kid you not a couple of weeks ago, I got an email from a CEO that I was working with. I started doing these like executive coaching sessions with CEOs just to help them navigate this AI stuff. Just as part of our projects. And they just got access to chat GPT internally and he sends me a message and it is just all, oh my God, it did this. It took five seconds. This would have taken me 30 minutes, blah, blah, blah. Two days later, I get an email. This is absolute garbage. I can't believe we're about to roll this out to the company. I just lost all of my work. And what it came down to was there's still some training that has to be done. There's still some literacy that actually has to happen to help people kind of up level. But to answer the question about present bias is that those experiences then start to color our feel it because we over generalize all this stuff. We see what's happening today and we forget that these technologies, this one in particular is moving at an exponential pace, which means every week it's getting significantly better than it was before. And we have a really hard time as humans seeing that. And so one of the things that I've done that has helped the most is I help contextualize that by showing people. This is where we were at just six months ago. And I show them the progression and you could do this with a lot of stuff. Now, whether it's image models or text generation models or what have you, you could show people how much better things have gotten in a very short period of time. Then they start to be like, oh, OK, so I shouldn't overindex on what I'm experiencing today because I can start to see where we might be heading. I experienced that when I started using the reasoning models. Everybody gets caught up in this model is coming out, this model is coming out. There's structural shifts in the approach that these models use. And now when I go in and I hit reasoning and it takes a minute, minute and a half to do its thinking, I sit down and watch how it thinks, air quotes, things. And that, to me, is the latest. Wow, in the future, this thing is going to be in the near future. This is going to be pretty powerful. I was working with a client and we were doing a strategy doc and say, hey, we want to go look at all of these firms and their approaches to AI and kind of what they're doing. And what it came out with was pretty amazing and very different than the one shot models. I tried the one shot and I tried the reasoning and the one shot, I would have said, hey, that's actually not helpful. It was actually more confusing. It lied in some places like it struggled. The reasoning model came out with something that back in the day, I might have paid Accenture a bunch of money to produce for me. And it was close, a lot closer than previously. You talked about coaching CEOs. One of the things I think people are struggling with is, hey, there's a lot of hype out there to your point. I tried it once. It wasn't really great. But I know this is something I need to learn and practice that. How are you approaching organizations that are AI curious? AI curious. Those are the best ones. That's what I would say. The other version. The either AI skeptics or the anti AI or the somebody said to me, we were talking about their strategy and where they were going. And they were like, we need to do stuff. Is this like sink the boat stuff or like miss the boat stuff? Tell me more. Well, I thought it was interesting and thinking about sink the boat is if you make this decision, you go down this path. Is there more risk in sinking the boat or is there more risk in missing the boat completely? Meaning you're stuck where you are today. You can't get to this future state if you don't do something. And they were trying to navigate whether or not now is one of those sort of pivotal moments. So from an organizational perspective, I think traditional tech adoption curves, right? You get people who fall across all an industry plays a big part. If you're in banking or insurance or health care, obviously the stakes are a lot higher. And I think that dictates your company's culture around innovation, experimentation, embracing new things. Risk. I mean risk in general. Exactly. And so you're going to have always some amount. And I think what we try to do is help organizations find the right amount. Like what feels right for them? Because you need both this like blend of a big vision of possibility to get people to get excited about to galvanize an organization, basically to create enough space for people to have permission to try things or do things in a different way. Then you have to be able to pull that back to something that's tangible that you can get started with today to start showing people that there's some momentum. But what I've seen across all organizations is like the number one thing is helping educate people on what it is, what it isn't and what it can and can't do. And you've got to get it into people's hands. Because when people start playing with this stuff and plays a really key word here that I'm finding out, just having fun. Like we don't need to go automate this process end to end. Like let's just go have a conversation with it and see what it can do. Lightbulb start going off. That's that like little extra space that you now have with people to insert new possibilities that maybe they couldn't have seen before. We were just talking before the show about going to do a hackathon because there's an education trust that is looking to your point. How do I create fun and just get people to experiment so that they're curious and they learn when you talk about literacy or you talk about that experimentation. Where are you seeing success in engaging organizations to help them not, hey, we don't need to plan the future or have a strategy or miss the boat or sink the boat. Let's just kind of dig in and all learn together. Where are you seeing like tactically programs, education, video, whatever it is like what is resonating with people? All of the above. You know, it's funny what's baked inside of this is the whole sort of personalized learning conundrum that I think is now more possible than ever because of this topic that we're also talking about with AI. Well, I'll tell you what seems to work across the board and this is going to sound real obvious. It's making stuff relevant to people. So if you go in and you do a presentation, they say, oh, we want to know about AI. What I hear all the time is, can you come and just like help us understand these AI terms? Well, if I come in and do a song and dance on what machine learning is and what computer vision is and neural networks like, OK, cool, you've just defined some terms. But like, what do you know at this point in time? You don't versus saying, OK, I'm on with a marketing team. Pick five things you guys had to do today. OK, cool. Now we're going to go do all of those with AI together right now and use walk through with people and show them what's possible. Show them the dos and the don'ts, just like some basic stuff with off the shelf tools that you can get approved inside your company to do this stuff with. You could start making a ton of progress. Then you can layer an actual programs about like, what's responsible AI usage look like? What data can I put in? How do I verify the output? Like you start to get into some tactical skills, but in the beginning you just have to get people to open to trying stuff. You talk about every company needing a proof of concept. So let's say you went into a client or working with someone and you did that. Let's say a marketing team because it's kind of an easier place to get started with content creation or sales funnels and stuff like that. What are the challenges with getting the first proof of concept inside a company up and running? The parts that take the longest would be tech and security, plain and simple, which is nothing unique to AI. It's classic like, okay, you're going to bring a new technology into a large organization. What checks and balances do you have to go through to make sure that it's going to check all the boxes? And it's why I try to work with companies as much as we can to create, here, call the bunch of different things. Sandbox, playground, like just give us some space to try some things. Because like you and I both know, this tech moves so fast, it's super new. Some of the stuff might not actually be possible today. It's almost like, why go and invest all that time in that stuff? If really you're just trying to test and learn in prototype early on. In the absence of that, I think where things really start to slow down is not picking a use case that's big enough. And I'll give you an example. So I was talking to a team inside of an insurance company. So, okay, what are the processes that take up the bulk of your time? And they start walking me through some of their processes. And they're all like, oh my gosh, yeah, this one, it's a huge pain point, huge pain point, huge pain point. So we map it all out and then we get to the end and I'm like, and how much time does this take? And how frequent is it? We're looking at the numbers and we're like, there's not a big enough ROI here to go spend any time on this whatsoever. And so there's something to be said too for if everything's going to be judged based on ROI, which in a lot of cases it should make sure you're starting off with something that's big enough, but then break down that big thing. If you're like, oh, if we could overhaul underwriting or underwrite 50% faster with less people and less data, that's like a 10 billion dollar challenge for us. Okay, great. Now let's start chunking that thing down into parts and get started. Who inside the companies is sponsoring these proof of concepts? Is it the technical team? Is it a business owner? Is it the CEO? Like, where are you seeing in your experience? People willing to sponsor sort of the, I call the first step. Yes. Is it like finding a needle in the haystack? Does it come down to the person? Like, hey, if this is a person that's really open to experimenting or kind of wanting to learn or maybe even have FOMO, or is it a department? The finance department is probably not the place to start, right? It depends. It depends. We've actually had the CFO end up being the champion, the buyer, the driver of the whole thing, because they see the opportunity to hit their spreadsheets. Was on with a client this week where we had talked to them seven months ago and they were like, yeah, we're trying to figure out use cases. We think we're good. Well, guess what happened? There was a board meeting last week and the board said, what are you doing? Why are you sitting on your hands? Let's go. And they called us back and said, I guess we need to go now. And when you say let's go, what services like, what are you helping them with? When the board comes down and says, hey, you guys aren't leaning in enough. I read the news. Where are they struggling? So I'd say it comes from a couple places and we really help them try to navigate what they're ultimately trying to solve for first. So in some cases, it is top level, like, let's align on what our strategy is. How is AI going to drive our corporate strategy, accelerate it, enable it, blow it up, whatever. What's our vision for it? Some people will come in and say, like, we need to start building. And so you kind of end up with people at like one end of those spectrums every now and again, we'll get somebody that comes to us with like, we need to solve for governance first. And it's like, governance is important. Yes. But like, maybe don't start with governance if you don't know what the heck you're going to be doing. Yeah, it's a hard place to start when you don't know what you're trying to achieve. So no matter where people come to us, whether it's we need to start building this stuff and we don't have the expertise or we need a strategy. We usually try to get them up to that level to say, well, do you know what your North Star is here? Like, do you know how AI is and is not going to be used? And do you have the right internal processes in place to actually make your life easier? Because most of the time what's happening is somebody gets put in charge of AI, sometimes multiple people compete and try to be like the AI person. A lot of times it ends up getting centralized in IT or tech or the CIO office. And you just end up with this internal cluster of people who aren't clear on who owns what and decision making and processes. So there's like some basic consulting stuff you got to figure out first before you even do the AI stuff. One of your best posts, because I follow you on the LinkedIn, was about a client who wanted to build 100 agents in 100 days. It was just an interesting hook line. But I think they were using Copilot, which is something that a lot of companies are starting to deploy, whether you agree it's the best model or not. Different conversation. It's a place to get started. Tell me about that approach and why it was so exciting and what happened. Yeah, sure. Microsoft, if you're listening, maybe simplify your product naming conventions, by the way, because people are really confused by what is Copilot and what's Azure AI foundry. No, but in all seriousness, yeah, so they came to us and they said, we just want to get started. We got directive from up above. We want to get moving. We can't bring in new tech, but we've got Microsoft Copilot licenses. Can you show us how to use Copilot to build low code, no code agents that empower our team? Which for me, as the rapid experimentation prototyping guy, I'm like, yeah, so that's why I posted about it because I felt super inspired to go after it. And that's really what it was. They were like, we want to learn by doing. We think that there's some small wins that we can do without having to go spend a bunch of money on development. Let's use tools that are already approved and just help us get started. And what tools besides just Copilot? So if somebody was listening, they're like, yeah, I want to do that. Like what other tools were they using inside the company? Well, I mean, for this one in particular, it was it was definitely Copilot Copilot Studio, to be clear. So that's like Microsoft's, you can go in and sort of drag and drop or simple interface to just describe what you want this agent to go do. And then assume you've got some permissions, you might be able to connect it up to say a one drive folder or give it some specific instructions. Or if you're a little bit more advanced, have it connect up to a database or another tool or go run automation. So it's still all within the Microsoft suite. But the same thing could be said for a lot of these other major platforms that are out there. Like this point in time, I think you probably do some version of this in service. Now sales force is kind of trying to set up their own one. So you're starting to see these major enterprise platforms that are adding this kind of build your own agentic layer to them. And so we got one the other day for DataCue. And I was like, I personally haven't haven't worked with that yet. But I know a lot of our team has on big data ML projects. And then turns out they've got an agent building platform now. So I think we're going to start to see more of it, which is why it's an interesting thing to think about using tools like that to empower maybe less technical people to build some simple stuff to start. Once again, to build that capability internally before integrating it with proprietary systems and PII data and stuff like that. If someone who helps people kind of navigate the future, this is change. We all read the the Internet's and watch the news. There's a lot of change going on in the world. And in some places, it feels like people are growing tired of getting change fatigue. They're wanting some sort of stability to do their jobs and still want to improve, get ROI, make the business better and stuff. From a change management perspective, as you talk about the future, what are you finding resonates with people that makes them more open, given that they may be a little tired of change? Oh, I agree with you. I'll start off by saying that the pace of change continues to accelerate and it can feel at times really hard to keep up. And just kind of reflecting on myself as an individual, you've got to figure out that grounding forces to you to come back to what really matters. And I know before we hopped on this call, we were talking a little bit about doing meaningful work and focusing on the right things. And at the end of the day, that's really what it comes down to is shut off all that stuff, turn off your phone, turn off your email, just come back to what's the most important thing? What matters right now? Get locked into that and go. And so I think from an organizational perspective, it's nothing new. It comes down to leadership. I think as a leader, we've all been part of meetings where you come into a, I've run these meetings, we're on the leader. And I'm like, okay, guys, there's one priority. And next thing you know, I've listed off 15 things, right? But really as leaders, we need to come in and say, this is the most important thing. What are we doing to move this forward? We have to help our teams feel grounded in all this change that's constantly happening. And I think the other aspect of it too is let them build that too. Co-create that stuff with them wherever you're going so that it's not just you dictating to them. Then they feel bought in, they feel ownership, they feel grounded. Nothing can get back and go do all the work and hopefully feel a little bit less crazy along the way. I'd be remiss in my responsibilities as a curious individual if I didn't talk about your burgeoning music career. And I say it tongue in cheek, but you know, one of the things that I think is important as it relates to this technology is having fun. Doing things that are not always finding the ROI or solving some automated workflow thing, but actually just kind of having fun with it. So I'm going to let the audience know Adam dropped his first single and I'll of course leave the link to it, buzzword bingo. And I've done a couple of these on sooner as well. Tell me about what got you excited about doing it because I actually listened to the single and tell me about that experiment. Yeah. My favorite part about this, by the way, is like how many people have unexpectedly seen that one post. So a full album dropping next week, by the way. What platform did you use? Just take people through. Yeah. If somebody wants to go out there because Adam, I don't believe you have a musical background. No, no. So this is the story I have always wanted to make music. Actually, I have a bucket list of performing on stage in front of a large group of people. And I like routinely sing songs to myself. I think it's a pretty normal thing to do. And I saw a couple of weeks ago, maybe months ago at this point in time on X, a new platform for music production called Refusion, R-I-F-F-U-S-I-O-N. And I was like, oh, let's see what this is like. And I listened to a couple of the tracks on it that just popped up on the homepage. I was like, all right, let's see. And it was kind of late in the evening. My wife and I were doing our wind down thing. I had my headphones in. And so I went over to chat to you BT and I was like, all right, first I need to like come up with my sound. I really love the Lonely Island. Like that's sort of a style that I think is just like fun. It's a parody type thing. It blends a lot of fun stuff. And so I took that as a style plus a couple other things. And so I had it kind of draft a paragraph for what my sound is, gave that to Refusion. And then I was like, I've always wanted to make an album about silly corporate life. Office space. And all the all the memes. So I just got with chat, GPT and started brainstorming topics and ideas. And then I picked the first one, which is buzzword bingo. I'm a consultant now. So I say synergy as much as I possibly can. And then I just started writing lyrics with it and iterating it, you know, so I could write some. I'd give it some feedback. I'd make some edits myself. And then it's really as simple as copying the lyrics and the direction for the sound over into Refusion. It generates you two versions. You listen to them, you can make some tweaks and changes or you're like, hey, this is the one I'm going to go with. And I generated one and I played it. And I was laughing so hard that I was crying because I was just so surprised at how one, how much fun I was having to how good it was. And it was like painfully accurate at the time. And I just kept going. I took that song and sent it to about 10 people only because it was it reminded me of my many years in corporate America and saying multiple things that got bingo cards. We actually had bingo cards back in the day. Oh, yeah. Because we'd go in knowing exactly who would say what in the strategy was, could you make them say the words? Totally. Yeah, it's like, have you seen the clips with where Jimmy Fallon got a bunch of the PGA tour guys to insert little random words in their interviews that week at the US Open? No. So it was like one example. This is not not correctly, but imagine like Ricky Fowler is up there giving his post event interview. He's like, yeah, you know, I just missed some shots here, a couple there. But you know, sometimes the turtle is just not poking its head out when you need it. It's like, what? What is that even me? But anyways, Jimmy Fallon like gave him these terms to to interject. So I don't know where I was going with that. If you had to give somebody advice today, that's why I brought up buzzword bingo. It's kind of a fun way. You find a tool. It's no cost. It's just your time and you get to have fun and send it to your friends or family to kind of get interested in. Wow, that was pretty cool. What's your advice to people to get started? My advice to get people started. You nailed it. It's pick something and have fun. But in a much more tactical way, the advice that I've been given more and more people is let's go with the chat GPs, the clods of the world. Instead of going to Google or instead of asking a friend or a co-worker, go to one of those tools first and just ask questions, give it things. If you feel stuck, go flip it over. I think that's a really simple mental thing to go and do where I think you get into like fun territory is if you've ever wanted to be a graphic designer or a musician or whatever, go find one of those platforms and just start playing around with that stuff. I think the one that most people probably could. So maybe a little bit of a stretch. Maybe it's a good thing is I think most people have thought of an app that they wish that they could have. Right. Like earlier, we were talking about a golf app for, you know, that's voice and a caddy. These vibe coding development tools are pretty fun. So Bolt.new is a good example. Cursor is a little bit too advanced. Replet. So there's a lovable, but I'd say Bolton lovable or the two. Go describe the app that you wish that you could have and let AI try to create it for you. And I think that's a really fun thing to be like, OK, really what I'm doing is I'm directing this AI with what I want it to do and what it's messing up and what I need it to fix. And just see how far you can get. Am I not get it perfectly here in the first time? But like, I think that's a really fun area to go and experiment. That's great advice. Adam, thanks for this conversation and always enjoy hearing your journey. I think it's a powerful mind of that AI success isn't about the technology. It's about the culture and what you think is possible. So appreciate the conversation. We'll put all the links in the show notes for people to dig in and try. And I highly recommend following Adam on LinkedIn. If for no other reason, he's dropping a new album. There you go. Thanks so much for having me, Paul. Appreciate it. Everyone, thank you for listening. Until next time, keep experimenting and stay curious.
