AI, Actually – Episode 15: How Businesses Can Actually Get Started with AI

Andrew Sweet (01:22)

Yeah, I love that question, Pete. Let me just start with, you know, with everything that’s coming out into the market, it is very easy to start with the tool versus the outcome you’re trying to achieve. Even if you’re doing something personally at home, start with what am I trying to do? Not start with the tools. So, you know, I think that that’s critical. And again, the distractions are are more prevalent than ever. But but having said that, it feels like and maybe this is well known that it

feels like we crossed some sort of chasm in December of last year. And just the level of productivity that you can get now out of tools is amazing. And so for example, mean, being able to quickly as a product person, quickly show a concept, not in PowerPoint, not in Figma, but actual code using Claude code or Codex is incredible. So I just think it’s not just the technology, but thinking about how you work.

how you work with colleagues, how workflows in your organization is critically important.

Pete Reilly (02:20)

And what have you seen that’s been surprising to you?

Andrew Sweet (02:20)

What do you think is the biggest problem

You know, again, it feels like this is going to sound funny. It feels like just six months ago, maybe even four months ago, when you were building, let’s just talk about the coding side of things. You really had to babysit the coding agent and it would frequently head off task and you spent all your time trying to figure out how to get it back on. Or you could build 80 percent of the solution really quickly, but it took forever to do the 20 percent.

Those days are quickly going away or have gone away. And the ability to quickly build, for example, prototypes that we’re doing even in the physical AI world where you’re interacting with cameras and quickly showing concepts to clients, to the client’s clients is incredible. And so what’s happened now is you can almost do that with one shot. And it’s very quick in the code it’s writing.

has significantly improved.

Pete Reilly (03:14)

Yeah, so one-shotting software where six months ago, it would have taken a much larger number of iterations, is your big perspective. Yeah, that makes sense. Hey, Jim, how about you?

Andrew Sweet (03:22)

Yeah, completely.

Jim Johnson (03:27)

Well, my surprise is from maybe a little bit of a completely different angle here. And the constant shock is companies, potential clients, people we talk with who have no idea what’s going on and that this is happening and the threat that it poses to their business and the opportunity that they’re missing.

It’s amazing. I mean, we have this sort of running joke that all of us are living inside this bubble and we’re reading and listening to podcasts and reading and talking about it. And that’s what we’re doing here, making our own podcasts. And there are other people, our 12 listeners who sort of follow along with us. We live in that world and then we go out and talk to somebody and the business problems that they have, the potential for sort of

to be readily addressed by the power of AI is unbelievable. And they’re just ignorant of it. ⁓ And I don’t know whether it’s purposeful or accidental, but the opportunity for them is incredible. The opportunity for us and what we do as a result of that is incredible. You know, I think I’ve said this before, when people start to think about AI and heading down this path, I am extremely reticent.

Pete Reilly (04:23)

Yeah.

Jim Johnson (04:43)

to ever say the words, this is an AI project. I’m even starting to sort of arrive at a place where I don’t believe in the idea of AI strategy necessarily. There is business strategy. are business problems. There are business opportunities. And it’s how can AI enable it or facilitate it. Now, there are certainly worlds where our clients can’t.

think about how to change their business strategy or can’t think about how to solve a business problem in a new way because they’re just not aware of what’s possible. And that’s sort of the obligation on their part is to start to get smarter. And how do we sort of face up to that and help them take this journey and get a little bit smarter? But just to circle back here, the shock to me is the divergence of understanding of what may be the

Pete Reilly (05:28)

Yeah. Interesting.

Jim Johnson (05:31)

certainly the biggest thing in 2013.

Pete Reilly (05:33)

I’m probably, I’m probably going to get a little more geeky on you and then we’ll go to Stew who will out geek me, I’m sure. But my, two, two surprises for me were it is ⁓ how, like I sat down to do, build a list of companies and a spreadsheet and do some research on some things. My normal playbook would have been a fire up a spreadsheet.

Jim Johnson (05:39)

Yeah. It’s easy for Stew to do.

Pete Reilly (05:55)

Start filling out every row of the spreadsheet, put in links and do a bunch of things. And I thought, you know, I’m going to try to build an app that does this. And I built an app that is hosted. It’s got a SQL backend. It’s got a modern front end with tools like, you know, using words like node.js and TypeScript and SuperBase and none of which I actually know by the way.

But I knew enough to be able to say, here’s the problem I’m trying to solve and to work with the coding agent, in this case it was Claude Code, to build a working solution that actually solved my problem within doing a little bit of work over a handful of days, something I never would have attempted before. That kind of thing now becomes possible. And the other one, I started using Claude Cowork recently.

And quickly was able to set up, download my Gmail, download my calendar, download transcripts from Fireflies and be able to turn that into a personal assistant. What am I to do? What’s on the calendar? What do I need to do to prepare? Build a presentation about the meeting that Jim and I just had and that capability to have, to build that level of assistant with very little technical knowledge, but just a willingness to just get in there and do things is just.

mind blowing.

Jim Johnson (07:18)

Pete, I’ve stopped even saying, here’s what I want to build. It’s more of, here’s my problem, or here’s what I stink at, any ideas. And then it’s sort of, go further back upstream with the question. You don’t even need to know that this is what you want to build. Just sort of say, is a problem for me, or this is something I wish I’d do better, or whatever. And it starts an interesting dialogue, and you head down that.

Pete Reilly (07:24)

Right.

How about you?

What’s your mind blowing experience?

Stew (07:44)

I mean, I think the biggest thing we’re starting to see early this year, that’s a big change from last year. I think a lot of the mind blowing things that are still mind blowing by the way, from late last year, were what I would call co-pilot scenarios, right? And so those are the scenarios where you’re…

you’re having a lot of back and forth conversations with an agent to accomplish a task. That would be like the sort of things we’re talking about with Claude Cowork. Now those have gotten a lot better over the last few months and continue to get even more superior. But I think we’re moving to this world where there’s additional, you continue to have the co-pilot scenarios, which continue to be very high leverage and keep getting better. But now we’re looking at colleague scenarios.

Pete Reilly (08:15)

Yeah.

Stew (08:31)

So this is a situation where the colleague, you know, where a, where you’re interacting with the agent, very similar to how you would interact with an employee and you’re giving them very long running tasks, tasks that may run literally for a few hours before they come back. And they’re giving you finished deliverables basically as part of that. So two examples from today on that, you know, one this morning,

on one of the things we’re working at, someone reported an issue.

in Slack, in the thread, at Claude, hey, go fix this. And it went off, its own, read the thread, got the context, understood what it needed to do, spun of its own cloud instance, solved the problem. I looked at it, created a PR, and we got it published, right?

And we got it published after other agents, by the way, picked up that PR and did code reviews, found one or two things on that code review and we fixed that up in a loop. So you’re interacting with it the same way you would an employee in Slack getting real work done. It’s using its own compute power to do so. So that’s a clear example of how these things are going. Another clear example that I’m doing right, I’m literally watching an agent right now.

that’s on a multi-hour run to start from a data set, go through in a large research problem that’s given it, that is not just, you know, find this thing or create this two page document. This is like create a really well thought out 30 slide deck that is, you know, very insightful.

Well, I’m not telling you what that deck should look like ahead of time or what reports go find the interesting stuff and create this deck from this research problem. it’s, and you know, it’s going to run for a couple of hours. It’s going to create that deck, but it’s also going to create two or three spreadsheets that back up everything that it says in the deck, right? Points it back to ground truth. Let’s you understand where the truth of that is coming from.

and, and a few other really interesting deliverables, from it. So getting your start in, in, know, but that literally is a multi-hour kind of flow. So it gets back to this colleague idea where you’re starting to get this world where it’s you’re, we’re really on the cusp of having effectively AI employees, right? ⁓ people that are doing that you’re working with, in the way you might a remote employee.

Jim Johnson (10:58)

I was just going to say, Stew, if you asked a really smart analyst familiar with that data to go attack that problem, your best guess, how long would it take for them to come back with the kind of qualitative result that you’re going to get here? And then of course, understanding your digital employee in this example may come back. It may surprise you with sort of insights that you’d

wouldn’t have thought about or it may have sort of wandered off in a different direction and you may have to get some coaching and say, go fix this. But sort of draw the analogy to if you had been working with a smart, experienced analyst in that.

Stew (11:33)

You know, it’s

jagged edges. And what I mean by that is like, sometimes it just completely blows your socks off. And then sometimes it does require more of that coaching at the moment. But one, just like this couple of hour run definitely would have been days, right? Even for a very experienced analyst. Two.

This is going to look at things that that experienced analyst really wouldn’t do, even if you looked at it for a week. And it’s because what we’re, you know, what we’re able to do through this process is it’s actually one of the things it does is it after looking at the data, it generates a bunch of hypotheses and we can just test a bunch of hypotheses that scale in a way that a human analyst honestly just wouldn’t be able to do.

And so that probability of getting like a really interesting insight or connection, I think goes up. you know, obviously the human analyst is bringing their own deep context around, history and then the industry and the business and other sorts of things that we have to, you know, it’s continual effort to try to get that into the agents, context. it, it can take those same things into account, but.

but it’s literally found things that, know, I don’t know that I’m a smart analyst, but I had looked at this data for, ⁓ months, very similar data. And when we point in this thing to it, it started, looking at the data in ways where I was immediately like, well, I guess I was a dumb ass for never thinking of that. You know what I mean? Like,

Like, you know, it really came up with some pretty smart ways to think about the data and look at it.

Andrew Sweet (13:05)

Yes.

Pete Reilly (13:11)

we’ve got all this backdrop of these things that have been surprising us and impressing us. But that’s sort of on the edge. And what we said at the beginning of this episode is 90 % of the people that are out there right now, we’re finding are nowhere near that. that may sound like pretty wild examples that we’re stating there. So where do we think, I want to shift the topic to

We see companies are stuck. Where are they stuck? Why are they stuck? Is it a tools thing? Is it a governance thing? Is it a leadership thing?

Andrew Sweet (13:41)

You know, I think there’s probably two or three sticking points. One is just getting started. It can be so intimidating that, you know, it’s hard to even figure out where to start. And so coming up with just, okay, let’s just start crawling and let’s start with individual use cases. How can we make each employee individually more productive and do that in a way that’s safe for our business by getting things like enterprise licenses, but getting started.

and leaning into it. The most important thing is to lean in quickly and understand the capabilities because I think once you do that, once you kind of break through to the other side, maybe it is, it’s literally breaking through to the other side, you start to see the possibilities. And then the AI itself can help you figure out what should I be doing next, right? So it’s not just the thing doing the thing, it’s the thing helping you understand, okay, where else should we go as it starts to understand the context.

of your business. So it has that ability to help there. And then there’s, you know, places where we can get into enterprise use cases, but starting small, starting with individuals, and then focusing in on business outcomes.

Pete Reilly (14:46)

Jim?

Jim Johnson (14:47)

Well, Pete, you hit me over the head all the time with crossing the chasm. And this is such a mindset shift. Again, we’re all in this every day and we’re continuously challenging ourselves to make the mindset shift. Don’t do anything without asking yourself how AI could help you. And only a certain number of people are those sort of

out in front edge early adopters. And there’s a tendency to simply say, hey, like all of technology or IT, I want to throw this to IT. If we’re going to do anything, let me just sort of make this IT’s problem. And there is a lot of this sense of let’s go check the box and we rolled something out, maybe.

Pete Reilly (15:24)

Okay.

Jim Johnson (15:31)

You know, I get it. There’s plenty of security issues, there’s plenty of governance issues, a lot of things to worry about, certainly. But setting that aside for a minute, there’s a little bit of IT, check the box, roll it out. And we’ve talked about that before, but I think Stew just hit on the mindset change, which is so hard to make, which is if you have at a reasonably low cost sort of infinite intelligence, not infinitely intelligent, but an infinite, you can buy it now.

at a low cost intelligence and the ability to take that intelligence and turn it into work. What would you do with it? And that’s a total change over, okay, how can I apply a technology to my existing problem set? Even if somebody’s there. Again, I think there’s a lot of people who are intimidated. think there’s a lot of people who sort of, I hesitate to use this phrase, but I will. There’s a little bit of whistling past the graveyard.

we’ll just see how this all works out. Or I tried to to you a year ago and it gave me a stupid answer. It’s impossible for people to recognize that even in the last three months, things have doubled and tripled in terms of capabilities. But I think if you can get the right business executive to make that change on the mindset of, what would you do if you sort of had viable intelligence?

and that intelligence can do work and we’re there, Stew just described it, how would you apply it? But that’s a

Pete Reilly (16:49)

Yeah. So you’re bringing it back to

really a leadership for you. It’s a lot about leadership and the senior execs in the company really diving in.

Andrew Sweet (16:58)

And so I think that’s think that’s a problem even in our industry. We tend to talk about chapter 10 when everybody’s still back on the intro and trying to figure out chapter one, two and three. And I think that’s that’s critically important. Yeah.

Jim Johnson (17:01)

You gotta use it.

Pete Reilly (17:11)

We’re there.

Stew (17:12)

And I, yeah, I mean, that actually hits on exactly what I was going to talk about, which is, think in order to be effective with these agents, you have to have even the very sophisticated ones. Now there’s a little bit of, of, of foundational knowledge. have to build up over time that some of it kind of needs to become instinctual. And if you don’t kind of get certain

It’s like math, you know, like if, I kind of like really struggled on, functions and algebra, I’m going to have a really hard time with calculus until like rock, whatever I was not understand, you know, until I connect the dots around those algebraic functions, I’m probably really, really not going to understand calculus very well. And there’s a little bit of that foundational knowledge that you have to build up.

which really takes immersion. you know, and I do, you mentioned the example of someone tries one thing and it gives a stupid answer and it comes, you know, and you, kind of say like, okay, that confirms my priors. This is all just a bubble and, and it’s stupid and whatever overhyped. Let me go back to the normal way when the reality is like to get the agents to work well for you.

is its own set of work and you, the immersion is necessary in a way actually to start to understand the theory of mind of the agents. And what I mean by that is like, you know, theory of mind is like a concept of. can, I can have a perspective on what you’re thinking, right? I may be right or wrong, but I can, you know, maybe guess what Pete, you know, Pete’s saying, like, what the hell are, you know,

Like I’m talking about whatever I’m talking about, but Pete’s theory of mind is like, what are you talking about? I got to get onto the next topic. Hurry it up. You know? So I, so with that theory of mine, I’m like, okay, I can move. can do whatever, right? You can kind of put yourself in the shoes of this other person, understand what context they have that is different from yours so that I can understand that person’s perspective and maybe get my point across. That’s theory of mine.

Pete Reilly (19:03)

Yeah.

Stew (19:22)

The same thing is true with these AIs. And you have to learn what they’re good at, what they’re bad at, what they’re good at if you tell them certain, right? If if, when you give them the appropriate context and understanding, but where they struggle if you haven’t. And it’s really, really hard to do that without some amount of immersion now. And I think that’s gotta flow all the way up to the executive level. That doesn’t mean the executives out there

you know, doing too much crazy stuff, but really starting to interact and get real work done with these AIs fighting through the initial struggles you have, starting to understand it, getting some of those lessons so can start to connect dots and see how these tools flow together.

Pete Reilly (20:04)

Yeah, I was going to maybe pile on to all of your comments. I do think it boils down to in most of these companies, the leaders got to lead. And I don’t think it unless you are experiencing this technology many times a day and really leaning in. I said something that somewhat controversial internally the other day.

Andrew Sweet (20:11)

Good.

Pete Reilly (20:24)

I said, I think CEOs should aspire to be using Claude code, should be using coding agents. internally we had people challenge me like that. Are you saying they literally be doing this? said, yeah, I do because I think it really is still exactly what you just said, because if you don’t do that, if you don’t at least aspire to that, you’re not getting in the flow. You’re not riding the bike.

Stew (20:49)

By the way, think it’s so much easier than people think. Executives, in a way, have an easier time with this than people who have never been an executive. you’re just asking someone to do something in English, right? executives tend to be really good at, good ones at least, to be really good at putting in, you know,

Pete Reilly (20:56)

Yeah.

Jim Johnson (21:02)

Right.

Stew (21:13)

building context around what they, not only exactly what they want you to do, but why they want you to do it, what they’re trying to achieve, all of those other sorts of things. So executives actually are often very, very good at this once they kind of run, get over any technical barrier they feel and just think of it as like I’m having a conversation with an employee.

Pete Reilly (21:36)

What do you think, Andy, what do you think is a good on-ramp? I’m an exec, I buy all this, I need to get on the bike, I need to start riding. What should I do?

Andrew Sweet (21:45)

Yeah, so I love that question. And I think you start in your own personal life. Like don’t even start on the business side. Start in your own personal life and say, what is a problem I have? So one of the problems I have and had, now I’ve solved it with an agent, is I have trail cams on my property, right? And I was getting all these squirrels and not seeing like deer and fox. And so instead of saying, you know, there’s probably an app for that.

Instead of saying that, just build the app for that and go out and be very prescriptive, just like Stew said, lean into that. So when you start to see your own personal productivity improve, then you can start to apply the analogies to the business. And so I think it’s a very logical place to start. And I would say just have the discipline, whatever you do on your weekends, pick Saturday morning, Sunday morning, whatever it is, and just spend two to three hours because you don’t need much more than that.

frankly, to really start to lean in and start to solve problems in your own personal life and then start to try other tools. And I think you break through in that and it’s safe, right? You’re not doing anything with the business and you have to expect, it’s just you and your coding agent, your favorite one, and a problem that you have that you’re solving. And when you solve it, there’s a real feeling of, now I understand what people are talking about.

You know, we’ve said this before, sometimes technology really hard to approach. If you wanted to be a C developer and you know, you as a CEO, you weren’t going to go get a C compiler and start churning out software. But as AI, you can do exactly that. So that’s where I would start Pete.

Pete Reilly (23:20)

You know what’s interesting, Andy, I agree with all that. I find people though, if you said solve a problem, they don’t even know what’s feasible, right? And so I would even back it up further. I would back it up even further and just say, and this is gonna sound a little odd to Jim probably, because he may disagree. But I think to some degree you have to say, I am going to just use this technology. have no idea. I’m just trying to explore.

Jim Johnson (23:30)

Yeah.

Pete Reilly (23:45)

the parameters of it and what it can do and its capabilities. And I give you an example. So my kids, I keep telling my kids, I’m like, guys, you need to get in here and learn. These are my adult, these are mid-20s sort of age. And so they said, hey, dad, let’s have a, and I was telling them all the stuff I’ve been doing with Claude code and so on, and they don’t understand anything I’m saying. They said, dad, we got to start at the beginning, AI-based. So I said, fine. So we had an AI.

Jim Johnson (24:09)

Isn’t it, by

the way, isn’t it supposed to be reversed? Isn’t it supposed to, you know, it’s technology.

Pete Reilly (24:14)

It is, and we’re a little backwards here, but

just showing them in ChatGPT like they’re used ChatGPT like Google search, right? Which is fine, that’s a good entry point. Showing them how to speak with it, right? Like, oh, you’re driving, you want to brainstorm on a topic or you want to learn about something, you can just turn this thing on and talk to it. That was new news to them. Using the camera.

to either the video, share a video of what I was doing and get feedback on that, or just take pictures and get feedback on that. Sharing my screen on my mobile device to get help with a technical problem with my mobile device. The idea of projects and uploading documents to this project, those kinds of things, once we had that conversation and just sort of teach them how to use it in their everyday lives, and what I would propose is you just need to embrace it. You need to say, hey, I’m gonna figure out how to use this thing.

and ask the model itself, what are the things I could use you for? just because otherwise the habit doesn’t really form. I think you just need to force yourself to get in there and start using the technology and see where, just to understand what the parameters are and the possibilities.

Andrew Sweet (25:21)

I made the same mistake really quick. I started talking about chapter eight and how you get started versus like the chapter one. And so I love that, Pete. The best spaghetti sauce I’ve ever made, Chat GPT helped me and I was taking pictures all along the way. Not only did it give me the recipe, but I was taking pictures. It was giving me feedback. You need to do this, taste for this, do that. So you’re right. It’s just the everyday life examples and anything you do anyway. I love that. I love that point.

Jim Johnson (25:21)

So, go ahead.

So for years and years and years, my wife has listened to phone calls, all these calls, and often I ask her her opinion of people, problems, et cetera, what’s going on, and I get stunningly good feedback. It’s shocking. Sometimes she points out things before I see them coming based on listening to these calls. I have a lot of scary stories about that. But so I decided to just hook Claude to Fireflies’ meetings and my email and my calendar.

and simply start asking questions. What problems do you see with work? What can I do better? How can I manage my time better? How can I help my team a little bit more? And sure, some of the ideas that came back were sort of rudimentary, but some of it was pretty darn interesting, especially about where it thought I was spending my time and where it thought problems were going to come.

really interesting conversation. And you know what? was just about me and me trying to do my job, but wow. It was a fun place to start. And I told my wife, I was now gonna have a second wife in the form of Claude analyzing my work.

Pete Reilly (26:54)

You

Excellent.

Jim, I’ll start with you. You’re walking into a 500 person org, right? What’s your general, and they’re saying, hey Jim, how do we get started? What’s a framework that you would use to advise them and guide them?

Jim Johnson (27:14)

You know, Andy said it earlier, and we’ve made this mistake in sort of jumping to chapter eight with companies that are super early in the journey. ultimately we realized we needed to circle back. there are, mean, here’s the bottom line. These are businesses, they’re functioning, they’ve got revenue, they’ve got expenses, they’re out there trying to do whatever they’re trying to do in whatever industry it is.

And we need to sort of respect that and start from that point. But there are some three, there are three or four or five basic conversations that we need to have very early with those clients. There are some things about security and governance that need to be dealt with. And we’ve run into companies that are just making some really scary mistakes in terms of what they’re putting out into the sort of open, I’ll call it open chat GPT sphere, if you will, of stuff that they really don’t want to. So there is some.

So there’s some rudimentary basics to deal with. Then there is a topic that needs to be addressed, which is just individual enablement. If we’re gonna encourage people to do some of the things that we just talked about, if we’re gonna encourage them to experiment a little bit, if we’re gonna encourage them to sort of learn how to apply it a little bit in their own life, maybe apply it in their own work life, their own personal workflows, you here’s what I do every day, here’s the reports that I need to create every day, how can…

How can a large language model help me? They need to be enabled for that. So there is this idea of basic end user enablement that needs to be addressed. And you’ve to figure out who are the right people in the company and what level they need to be enabled. But that’s just rudimentary. Let’s go ahead and do that. Then on top of all that, there is this idea of, I’ll call it, enterprise use cases. Again, go back to this idea of we’ve got this.

tremendous amount of intelligence now that’s purchasable. And it can do things. It can do work. This is the point of agents. We talk about it all the time. Do work that you’re doing today. Do work that you wish you could get to or do work that maybe you haven’t thought of yet. Spend time thinking through what those opportunities are. What’s the potential value of doing that work? And what’s the viability? Some use cases would be awesome.

But maybe they do exceed what we would want to turn over to an AI agent today. Or maybe we’re just not ready to go all the way there. But I guarantee you, there are slices of work that can be turned over to an agent today. again, sort of recap. Some basic security and governance, and we just got to deal with these things right now. And that’s a roadmap. We can mature there. Dealing with end user enablement, that’s a roadmap. And helping people become productive individually. And then figuring out

the sort of where we can get the most value at the enterprise level and think through those enterprise use cases. That’s really oversimplified, but that’s the way we need to think about it, I think.

Andrew Sweet (29:56)

And just to build on that, love, know, if you’re especially an SMB, I think if you’re a small medium sized business, you have a tremendous opportunity. Sometimes people think this is only for the Fortune 1000. It’s absolutely not. And, you know, a very logical place to start what we’re seeing is with the customer journey. How do you optimize that customer journey from both the customer acquisition perspective, but also how do you optimize the employee experiences you’re going through that there’s mundane parts.

of the customer journey that you can free up your employees to focus on higher value things, building relationships with potential clients, building relationships with long time clients so that you keep them. So you’re freeing up your team to optimize the relationships that they have, but you’re also improving that customer experience as they go through the entire process.

Pete Reilly (30:48)

I think there’s a massive opportunity, especially for more mid-sized companies, to free up capital to grow. And what I mean is that these tools are actually starting to be able to do the work that humans could do or tasks, let’s say, that they would do.

Jim Johnson (30:53)

Uh-huh.

Pete Reilly (31:06)

which then enables you to focus on growing the business, right? How do I acquire more customers? How would I support more customers and so on? And the framework that we’ve talked about is we call this crawl walk run, but also thinking about it at the individual level, maybe the team level and the organization level, and really pursuing all those really at once. to really get to the crawl stage through a two to four week,

sort of if you’re really focused, it’s possible to get through that in sort of a two to four week pace or timeframe where you’re learning the basics of it, assuming you haven’t done these things already. ChatGPT, how do you set them up for enterprise or Claude and or Gemini, depending on the space that you’re already in. How do you start at the basic level with those tools very quickly within a month or so you want to really move on to sort of the walk phase. we want to, and to me, those are things like

How do I use these tools to generate presentations, to generate financial models, to evaluate financial models and to do much more sophisticated work that we all do as individuals and our respective companies, you would be, your mind would be blown if you could see if you try, for example, Stew mentioned Claude Co-Work and its ability to build presentations and spreadsheets and take a bunch of spreadsheets and put them together very quickly on the fly is massively impressive.

And then the final stage, this ultimate stage that we really weren’t trying to get people to, is around the run stage. And that is having agents that are proactively helping us sort out financial allocations, or agents that are helping us handle support calls, or agents that are doing work, that are doing tasks and accomplishing meaningful tasks. And it’s important to approach that thoughtfully.

and to look at not just slapping AI and maybe an existing process, but reinvent, looking at the capabilities of that, these platforms and reinventing the process in a way that really takes advantage of the capabilities of that platform. So that’s sort of the gamut of, and the series that I would take people through to make their way from just getting started to really maximizing the value from these tools.

Jim Johnson (33:11)

So I love that well-framed. think one of the things we’re going to be thinking about here is organizational design, organizational change. And Stew said it earlier, if sort of the listeners who are picked up on that, which is this idea of the digital employee, where we’re going to be giving that individual their own.

You know, that, that, that digital agent, their own email and, you know, their own, the, the, right tools to do the job. It is completely analogous to, hiring an employee, but it’s not going to be a one-to-one. I know we’ve said this before in other discussions. It’s going to be, it’s, going to map differently because of the, the sort of scalability of it, this, the speed of work of it. It’s going to require a lot of.

of sort of evolution and sort of the human roles and how we supervise and lead and take responsibility for the results of those digital employees. And it’s not different than you would today in terms of taking responsibility for an employee of yours and sort of their output and what they do and the actions they take, but it’s gonna be different. It’s gonna map into the organization differently in terms of the tasks they’re taking on. Again, the speed, how we review that work.

you know, how we sort of ensure that it’s on track. It’s just going to be different. It’s going to be exciting. But I love that metaphor, that model of the thinking about it really as a digital employee, and we’re going to be expanding their capabilities over time.

Andrew Sweet (34:38)

there’s so many things in people’s heads. You you just have a simple approval process. Well, it flows really easy for 80 % of the cases, but it’s that 20 % was no, no, you got to go talk to Sally and accounting and she knows the rules on how to override and she does this and she goes to this spreadsheet in this new world that traditional

Jim Johnson (34:39)

yeah.

Andrew Sweet (34:58)

You know, you could almost call it traditional consulting, pulling knowledge out of people’s heads, re-engineering the process so that an agent can operate in that world and semantically the language of it, they understand that language, I think is going to be critically important in building that kind of roadmap too for these new digital employees. And I’m not trying to be chapter 10ish, but I think that’s where the world’s kind of headed is understanding that semantic blueprint.

Pete Reilly (35:25)

Maybe to close it out and to hopefully get people fired up about the possibilities, Jim, how should companies be thinking about return on investment payback here? Meaning where should they be looking for? What kinds of, how should they measure it? Cause it’s not just as we said at the top of the calls, this is not technology for technology sake. I think we all believe there’s legitimate business value here. Even if none of the models improved a scotch beyond where they are today.

We have an enormous opportunity to leverage this, we believe, to achieve some real business benefit. How should companies be thinking about that, measuring that as they embark on this?

Jim Johnson (36:02)

This is a controversial one and, uh, you’re sort of, but, I love the topic. So let me, let me maybe address it in a couple of slices at the individual productivity level. This is a challenging one because we’re saying to, you know, if you’re running a 500 person or a thousand person organization, you’re saying to your knowledge workers, your smart, smart people there.

Hey, I want you, I need you to get up to speed. I’m gonna give you the tools, I’m gonna help you, but you’re gonna have to invest some time in this. People are looking at it saying, you where’s the white space? I’m overwhelmed with my current job. We need to support the company leaders, the executives need to support that. And that is a challenging walk. The other side of that equation is if all of your team members become…

10 % more productive or 20 % more productive as a result of this kind of enabling technology, how do you capture that value for the business? And that is a very challenging one and unique to the enterprise. there are lots of different ways to do it. There’s some that are sort of top down and some that are bottom up, but I think it’s a worthwhile discussion organizationally and department by department. Now.

in terms of what I’ll call enterprise use cases, that gets a lot cleaner. There are definitive opportunities to say, hey, there’s something we’re doing today, it’s probably transactional oriented, that we now have this intelligence we can apply given the right tools, it can get work done. That’s sort of easy, low-hanging fruit. And you know what? It’s…

Doing it today is, we can do a ton of things that we couldn’t do a few months ago, that’s awesome. I tend to believe that you should shoot for very quick payback in that space and use that payback to likely fund more sort of opportunistic or aggressive use cases to maybe do things that you aren’t doing today. Hey, we can apply more muscle, more effort, more labor now onto these other things.

and fund the next thing and the next thing. Ultimately, you’re trying to do two things. One, you’re trying to create a flywheel of AI here. And once you start to get it, it puts you ahead of other organizations in terms of that return and that value and the reinvestment of that value. The second thing though, is there is a certain level of accountability that you need to put in place to say, yes, we’re not just mucking around with this technology or experimenting because that

bluntly tends to be sort of what happens with IT led initiatives or technology led initiatives. You got to put the accountability on the business to show that payback and to go after those business cases. Maybe things that they’ve never been able to do before, but can have real measurable hard dollar impact. You know what? I get in a conference room and I talk with clients about this for hours. So that one’s a tough one to hit in a few minutes, but I think it’s a challenging thing to think about at the individual level.

⁓ I think it’s a little bit clearer on some of the enterprise use cases to create that flywheel of value and reinvestment.

Pete Reilly (39:07)

Jim, I almost think about it this way at the individual level. Welcome back Stew. The way I think of it is you’ve got to get, this stuff has to come, a lot of it has to come from the bottom up. Yes, top down, you’ll have a perspective, you need to get everybody in the business sort of pulling for, hey, I see this great opportunity here, or I see this great opportunity here. And I think that individual level empowerment,

Jim Johnson (39:11)

So, we missed you Stew.

Pete Reilly (39:30)

It’s almost just like, yes, maybe we get some benefit from that in the short term, but it’s almost about, think, Stew, you called it the model of mind. How do we teach everybody how to use this and what the possibilities are? Because if everybody understands that, then you start to get this grassroots capability where everybody’s identifying, here’s a really big pocket of return that we could go get. And in that point, Jim, I agree, there has to be a very distinct, here’s the plan, here’s the return we’re expecting, we’re going to measure that.

and ensure that we’re getting those things across the finish line and building the habits that can repeat that over time.

Andrew Sweet (40:02)

Yeah, and I was going to come at it at a slightly different angle. I think sometimes businesses make the same mistake I made when I planned my Vienna trip using chat GPT. I asked it to do all the fun stuff and then I had to go do all the hard work, right? Make all the train arrangements, do the hotels, do the flights. And so it got to do the fun, interesting, go visit this place, go do that. And I think we should flip that on its head.

I think we as humans should free ourselves up to think creatively and actually have the AI do the mundane hard tasks. as part of, you as you’re looking at your organization, where could we take those tasks, automate them, free our humans up to do work that they were really built to work? And I heard that from an organization we were talking to this week. It’s like, I want our humans to be doing the kinds of work that we were meant to do.

not the mundane tasks. And so I think I would challenge everybody to look at it from that perspective as well.

Pete Reilly (41:01)

Stew, we were talking about a return on investment, payback, and how we recommend companies think about that. Anything top of mind for you there?

Stew (41:10)

Yeah, I mean, I think a lot of it goes with how do you implement, right? So some of it is start small, try to find a good, hit, ⁓ be able to demonstrate ROI. think continually having that re-grounding yourself back to, is this actually creating business value or am I just creating more slop faster or whatever, you know, is a…

Pete Reilly (41:20)

Good advice.

Thank

Stew (41:33)

is a continual thing to do. I do think there, you know, the small word of warning is, you know, it’s, there is typically a little bit of a J curve and you want to compress the J curve. But what I mean by this, this has actually always been true in productivity around technology changes. Even if you go back to, you know, the early nineties when it was the web coming on and things like that.

There’s a little bit of J curve on the adoption of these technologies, which means you do have to, and you know, when you first start to use this technology, your productivity may go down for a little bit because you’re trying to figure out how do I use this technology? What’s the best way to do it before you get, know, before the J turns up and it goes to dramatically better than it was before.

⁓ so, what you want to try to do is compress that J right. as much as you can stay focused, find quick wins. and I think as we’ve been working with clients, you know, I think it’s very possible to, you know, get a flywheel going where, you know, the last initiative is paying for the next initiative, through the ROI and you continue to get, continuous improvement, but, but you definitely just have to stay focused on it as a goal.

and ground yourself on why are we here? We’re not here to use AI. We’re here to solve problems effectively and, and yeah.

Pete Reilly (42:50)

We’re exactly where.

Jim Johnson (42:51)

Stew, there

should be no such thing as an AI project. There’s a thing we’re trying to do, and maybe AI is a great enabler or accelerator.

Stew (42:54)

I agree, yeah, no.

100 % yeah and you always got to reground yourself I think back on what are we here what are we trying to do yeah you’ll get me off on all sorts of theory I actually have a theory I think a lot of things well just that we used to think we’re important but we’re probably performative

will actually go away in the age of AI when it’s no longer performative to do them. And, and, or, you know, you know, when everyone can do it, it’s not performative to do it. And so the answer may be that just, you don’t do it anymore. Right. Like it actually can force you to reground on like, what am I actually trying to do? What’s the business purpose? How is this helping a customer better their lives? You know, I think is, is something you always have to go back.

to

Pete Reilly (43:44)

So we’ve got about a minute left. Jim, I’m going let you bring us home in terms of maybe just bubble up what your advice would be to these mid-sized companies that are maybe a little bit on the side.

Jim Johnson (43:57)

Well, my advice is, if you can sort of make this mindset shift, this is the inflection point and it’s, I say the inflection point, we’re in an inflection time, the next 12, 24, 36, 48 months, where small and mid-sized businesses have more of an opportunity than ever before to unsettle

rethink their industry, maybe jump the sort of large scale competitors they’ve typically dealt with in a way that just hasn’t been possible before. If you’re willing to sort of, if I’m talking to the leader, if they’re willing to be a visionary leader and say, okay, how can we do this just dramatically different at scale, at speed? It’s just not been possible before. So to me, this is just an opportunity for companies like that that just

hasn’t existed before. Now I get it, there are constraints, right? If you’re in a manufacturing industry or something like that, the plant is still the plant and the supply chains are still the supply chains. But this presents an opportunity that we just haven’t seen before to maybe make a leap forward and just jump into it.

Pete Reilly (45:08)

Great. With that, we’ll close it out. Good to see you. Look forward to the next time.

Andrew Sweet (45:11)

Good.

Jim Johnson (45:12)

Thanks again. Bye.