Subscribe
Share
Share
Embed
Future-proof your leadership with High Octane Leadership, a place where business leaders—whether by title or aspiration—share cheat codes for unlocking workplace excellence, lessons learned along the way, and insider tips for future generations of next-level professionals. With a career rooted in building people and businesses, Donald Thompson is an award-winning CEO, speaker, and author who empowers leaders to scale with purpose. Over the last 25 years, he has helped startups and enterprises alike drive cultural change, unlock performance, and deliver exceptional results through strategic leadership.
Find him on LinkedIn, and listen here to learn how you can become future-proof too.
Don Shin: [00:00:00] society never signed a contract or made a pact that all of humanity's data would go towards the building of these large language models that could potentially cause economic disruption for a whole bunch of people.
We just signed up for Facebook. We just signed up for Pinterest. We just signed up for all these, Reddits and, Twitter. and yes, there might have been terms of service, but no one had an imagination that we could potentially be contributing to superintelligence.
Donald Thompson: Welcome to another episode of High Octane Leadership. I'm very, encouraged about today's episode. Good friend of mine, CEO of CrossComm, Don Shin. Don, welcome to the show.
Don Shin: Thanks. Thanks, glad to be here.
Donald Thompson: I'm gonna give a little introduction, and then we're gonna get into it.
Don is a 25-plus year industry veteran in the technology space, Duke alum, he is a very sought-after speaker on all things, AI. Most importantly, [00:01:00] he has been an entrepreneur and business leader in this ecosystem for a number of years.
And so we have a shared affinity with helping other entrepreneurs grow, other business leaders to understand new technology and how that integrates with people. And so without anything further, we're gonna kinda jump in, to some questions. So one of the things we want to unpack, but before we get into it, AI is everywhere.
Independent of industry, independent of level someone is in an organization-
Don Shin: Yeah.
Donald Thompson: Yeah ... AI is a trending topic, if you will.
Don Shin: Yeah. Yeah.
Donald Thompson: What I would like to understand is, as an entrepreneur, as a technologist- Yeah ... running a technology company, how do you see AI changing the landscape? And how would you explain that to somebody that is really not living it every day like you are?
Don Shin: Yeah. So I'll start your, the answer to your question with a question.
(Clip - Abundance of Intelligence) what does our world look like with an abundance of intelligence? And what does the economy look like? What does our day-to-day work look like when we have an abundance of intelligence versus a sparseness of [00:02:00] intelligence?
So whenever we've seen technology change the abundance of a particular resource in our economy, it upends and changes and shifts and disrupts, changes what the economy looks like. And I think we're gonna see that but in a, on a different level because an abundance of intelligence leads to automation opportunities that were previously impossible.
So we've been automating, work ever since the advent of the computer. but all that work has previously been pretty deterministic and pretty much on the rails in doing a particular kind of work or task, and now with artificial intelligence, we're able to see the automation of decision-making and judgment- and possibly of subjective calls in new scenarios that were previously unforeseen. And that's going to open up the doors of possibility for automation for better or for worse because I think there are some areas where automation, should be embraced [00:03:00] with open arms, and then there are others where automation just because you can doesn't necessarily mean you should.
I would never want nuclear weapons to be automated, as an example. Fair point. An extreme example, right? I think that it's going to provide a lot of, a lot of debate and a lot of discussion in terms of, okay, now that we can, should we, and in what ways does this serve us or actually hurt us?
Donald Thompson: One of the things that, we've, we're both, I, would s- say in the similar generation.
So we've seen, so we've seen a lot- Yeah ... in terms of technology changes. You would remember the Y2K scare. When the internet started to become, a systemic part, systematic part of our infrastructure. All the malls are gonna be closed. No one's ever gonna shop in person again.
Why is AI different in terms of a lasting power? Why is AI different from a trend versus a very significant change in the way we work and live?
Don Shin: Yeah. [00:04:00] So I am one of those folks who would believe that AI is actually more significant than the advent of the internet. And because it's... Let me give you my own company as an example.
So I've been building digital product all my life. I started CrossComm as an undergrad in college, building websites on Netscape 1.1. Then I moved to building mobile apps. Then I started building augmented virtual reality apps, that lived on IoT devices and on the edge, then apps that levered machine learning and large language models.
And now I'm starting to realize, like the Avengers: Endgame, we've hit the endgame of building in terms of building the way we used to. It really is a true industrial age level event, where the way we did things before and the way we built things and automated things is now completely different with, AI-generated code, AI-generated, and automated [00:05:00] processes.
It's going to really change the nature of knowledge work in the same way that NAFTA and, the, movement of manufacturing capability changed blue-collar work 10, 20 years ago. So we're gonna see that on a much bigger level, and I think also the other difference is that it's happening much faster.
Yes. So John Keynes, econ- economic theorist, he talked about how, when new technologies disrupt the economy, society needs buffer time to accommodate those changes. And as long as there's buffer time, society can really benefit from those net gains in efficiency. But if the change happens too quickly, then we have, job displacement and disruption, in a way that can't be easily made up with just the balancing of numbers.
We've got people maybe your age and mine, who are now facing, [00:06:00] upskilling requirements and needing to consider whether or not to do something different. with the speed of economic disruption, I think the velocity is a real problem because it's not really giving time to society to adjust and for individual lives to adjust. You can look at it a macro level, but at the end of the day, does it really matter to an individual who has spent an entire career investing on a particular track and now having to be challenged to perhaps explore a different track?
There needs to be an accommodation for that along with the technological shifts.
Donald Thompson: I think you make a really good point of the people side of this technology shift, and one of the things as I chat with people about AI, they're both excited and afraid.
Excited about the productivity gains, what the technology, the tools can do- but then uncertain about what that means to me.
There's a group of people, 30,000- that woke up a week [00:07:00] ago and, were laid off, fired from Oracle.
Oracle's profits were up. Usually when a company lays off, you think their business is going backwards. Yeah. They're struggling, there's X, Y, Z.
But these folks were laid off so that this large technology company could put more money into their AI infrastructure.
And so when you mention upskilling, when you mention the pace of change, I am seeing that be a significant impact. I was reading, some data from Telus Health, an organization that, that I'm a part of, and we have a center of organizational effectiveness, and they were talking about employees that were surveyed and the level of anxiety going up around AI.
So my question to you is, there anything we can do to help employees and people who don't fully understand all this change- this pace of change, deal with it better?
(Clip - It's not just AI anxiety, it's AI resistance)
Don Shin: It's not only anxiety, it's also resistance that's growing over time. And w- I'm, I felt it at the most recent All Things AI conference, this sense of both excitement and wanting to be [00:08:00] there for the sake of relevance, but also a sense of are we starting to go into territory that we should be really careful about as a society?
I think if you're going to, and I, Look, I don't blame any business to explore how AI can transform-
Donald Thompson: Sure ...
Don Shin: operations, productivity, efficiency. Totally makes sense as a business owner myself. I think that there needs to be a pact of trust that if we're gonna go in-- If we're gonna go higher, that we go higher together, and that we figure out, and offer, opportunities for people to pivot to adapted roles or learning new skills- Yep
or potentially getting to ground where, automation has not yet occurred for w- whatever reason. because at the end of the day, the, the, worker and the team member has to be willing to adopt AI in order to net its benefits. But you're gonna get that ground level resistance [00:09:00] unless you show them how they're gonna be able to benefit in the process as well.
Donald Thompson: No, I really appreciate that point.
One of the things, Don, that we talked about the team members-
and managing the anxiety, and then you mentioned resistance, right? And one of the things that's really gonna be a critical business skill-
in my opinion, in the new economy, is adaptability.
people talk that change is the constant, but the thing that is changing is that change is coming faster than people can understand or control.
And so folks that are resistance to change by nature-
are gonna have even more of, a struggle in the new economy because change is gonna be the new partnership skill, the new survival skill, and the new advancement skill.
And, so I have a, a, strong feeling like, about that.
When we were off camera, and I want to give you space, you said you had a controversial take-
Don Shin: Yeah ...
Donald Thompson: on the way we should be [00:10:00] viewing AI from a high-level societal lens.
Don Shin: Yeah. I'm
Donald Thompson: gonna give you space 'cause I'm, I'm, here for it. Sure. I'm into it.
Don Shin: Sure. So when I speak on an individual level, I speak as an entrepreneur, but I also realize that at the end of the day, when we're all put in a situation, in a system where we all have to look out for our own interests, it's the collective that then suffers.
And so s- that means the system might need some adjusting. The system might be broken if we're all starting to look out for our own interests in a way that's perfectly understandable on a micro level, but actually does damage to us all on a macro level.
And that's an, that's a future where we're all living in game theory, and we're all missing out on a better outcome as a result.
(Clip - AI Leader Advocates for Token Tax)
Don Shin: And look, I'm going to use AI because if I don't, my competitors will, and they'll be more competitive, right? But, how can we all be incentivized to do [00:11:00] and create the reality and the future that we want our children to live in? ultimately, we've gotta make sure that the productivity gains netted from AI, are experienced and benefit all of society and not just a handful of folks.
So I would actually advocate for something along the lines of a token tax.
Donald Thompson: Okay ...
Don Shin: that rolls off the tongue, token tax, but what I really mean is taxing, electricity and water consumption of data centers. Because at the end of the day, society never signed a contract or made a pact that all of humanity's data would go towards the building of these large language models that could potentially cause economic disruption for a whole bunch of people.
We just signed up for Facebook. We just signed up for Pinterest. We just signed up for all these, Reddits and, Twitter. and yes, there might have been terms of service, but no one had an imagination that we could potentially be contributing to [00:12:00] superintelligence. And so if all of us were unknowingly contributing to the formation of superintelligence, then perhaps all of us as, in society should actually be benefiting and reaping the dividends from the economic gains of superintelligence.
Donald Thompson: That's an interesting take, right? Because if you think about the terms of service of a Facebook- or an Instagram-
Don Shin: Yeah ...
Donald Thompson: they're very clear in what they're going to do with your data.
Don Shin: Yeah.
Donald Thompson: And then we also know that we get served up ads when we're talking about a certain thing in our home, and that we're being listened to.
And so whether it's knowingly or unknowingly, we are the consumer and the product-
Don Shin: Yep, we are ...
Donald Thompson: of what is building these large language models. And so what I want to share with the audience, because I want to make sure that those that are familiar with AI follow what Don was saying in that moment, and those that, like me, that have to build from the ground up, when he talks about Reddit, what you want to understand is most of the language models have [00:13:00] certain sources they go and pull and train from, and Reddit is one of the primary sources that ChatGPT goes to pull information.
And so all of the data that we're delivering to these tools that are free-
are building the foundation of what we're now paying for in these different AI tools and different things. Is
Don Shin: that- That's exactly right.
Donald Thompson: Is that right?
Don Shin: Yeah. Yeah.
Donald Thompson: go ahead.
Don Shin: And,
so getting back to the idea of taxing electricity and consumption by data centers, like you might say, if the cost of tokens was higher, then maybe AI's not gonna take off, and then maybe China will win.
firstly, if tokens were more expensive, that would at least flatten the curve. We talked about flattening the curve in another context, right?
Donald Thompson: Yep.
Don Shin: To spread out change, to allow people to adapt, we need to flatten out the curve in terms of economic disruption for the sake of people and society. And actually paying what tokens actually cost could, [00:14:00] actually do that.
The reality is that we're actually paying maybe a fifth, maybe a 10th of what tokens actually cost through an LLM service, and that's actually speeding up, not slowing down, economic disruption for everyday normal folks. So just paying the amount that it actually costs would be better, and having a token tax, I don't think it would stop companies from automation.
I don't think it would stop companies from embracing AI. They might do it a little bit slower, and they might value the economic output of humans a little bit more, more thoughtfully as well. so I am all in on AI adoption. But I think that at the end of the day, we need to think about changing the incentive structure and externalities so that all of us are operating in a way that benefits society as a whole.
And also, the subsidization of AI right now is actually [00:15:00] causing an artificial distortion of comparison of value between what humans can do, against AI.
Donald Thompson: And so one of the things that you said right at the end was that value proposition between what the machine can do and what humans can do.
(Clip - You're Using Claude Wrong)
Donald Thompson: And one of the things I've learned as I've matured in my AI journey, is there's still a place for me.
And that was concerning 'cause AI's gonna take all the jobs, all the different things. I literally asked Claude, which for my knowledge work and different things, I use Claude quite a bit, is I said, "How-" confident are you in your answers for me-
in version one or two of what I see?
And Claude said, and I'm quoting Claud- Claude-
My first two answers are guesses, and I don't recommend you produce or send to a client, to anybody that matters," right- "the first one or two versions."
Don Shin: Yeah.
Donald Thompson: I said, Claude, what if it's the seventh or eighth version, ninth or [00:16:00] 10th version?"
Then it is me operating against your thought process.
And now the data to support it is what I research.
But the thinking is yours.
Don Shin: Yes.
Donald Thompson: And so what I encourage people to understand is that people talk to me all the time of, "AI slop in the system," all this stuff, and you're right, but you're using the tool wrong.
Because literally the tool said to me, the first two answers are educated guesses. You have to work with the tool to teach it your voice. You have to work with the tool to communicate at a human judgment standard. That is not what the tool's designed for, and that's not what it'll do. And so once I started to understand that from prompting and different things, I was like, "Oh, wait a minute.
This is instead of me cutting my grass with scissors."
This is me now having a John Deere supercharged lawnmower.
Don Shin: Yeah.
Donald Thompson: But it still requires an experienced driver- Yeah ... to get it right.
Don Shin: Yeah.
Donald Thompson: And that gave me more confidence in [00:17:00] as using the tool that everything about me wasn't gonna be replaced.
Don Shin: Yeah.
Donald Thompson: That judgment layer, I think is still gonna exist for a while.
Don Shin: Yeah. It's a force multiplier for you because you are senior, you are experienced, and you're an entrepreneur. With having four kids, I ask myself the question: What kind of opportunities are there gonna be for juniors and inexperienced folks who have yet to cut their teeth?
And what kind of place is there gonna be in the economy for them when AI potentially operates at a mid-level? So those are questions that w- we, can't solve individually, and that's where leadership, leaders like you and me, leaders like politicians, leaders like leaders of society, we all need to start discussing this and figuring it out together.
competition with foreign countries be darned
Donald Thompson: one of the things I've been really fortunate the last couple of months, I've done several speaking engagements at universities.
And so talking to young people has both, got me more optimistic- but they've [00:18:00] shared some of the concerns you just described in terms of what's, next for us-
Don Shin: Yeah ...
Donald Thompson: right when we get out in, into the world. So one, I certainly, encouraged AI adoption.
but number two, I encouraged them to take coursework and coursework that would help them learn how to think.
Yes. Yes. Because the judgment layer, the strategic thinking layer-
Don Shin: Yeah ...
Donald Thompson: is something that is not going to be defined longer, f- as fast as a AI stack as the routine.
Don Shin: Yeah.
Donald Thompson: So as, my good friend Greg Boone would say, we want AI to do the routine-
so that our teams can do the remarkable, right? Yeah.
And so that was my encouragement point, talking to young people, but it's a real risk to them because those entry-level roles that really can be more cookie cutter-
Don Shin: Yeah ...
Donald Thompson: Those are the low-hanging fruit for organizations that are looking for their productivity gains-
Don Shin: Yeah ...
Donald Thompson: relative to AI.
Don Shin: Yeah. I think your [00:19:00] advice is really sound, and I express it in the same advice in a different word.
I tell my kids, "You guys need to become entrepreneurs." if not entrepreneurs literally, entrepreneurs in mentality, problem solvers, thinkers, people who are proactive to figure out and grab a situation and guide it to the outcome that you want it t- to be. Those are the people that are gonna win the day in the next generation economy
Donald Thompson: I couldn't agree more.
Is CEOs typically understand AI's important. CEOs understand their competitors are gonna use AI. They understand they have to deal with this new change in the business environment.
How would you get them to understand how to really unpack it in their organization? What does AI adoption done well look like? How would you get them-- What would you guide that CEO that's sitting with you, asking those kinds of questions?
Don Shin: Yeah. So what we saw a [00:20:00] lot of in 2025 were a bunch of CEOs giving out a mandate from top down that you've got to adopt AI or else without adopting AI themselves, and that caused a lot of problems, a lot of wasted money, a lot of failed pilots because people were given the what but not the why behind the what.
And so what I-- what we're already seeing better in 2026 is an understanding that everyone on leadership needs to understand how to use AI. Once upon a time, when the internet came out, there were a bunch of CEOs that paid assistants to use the internet and send emails for them, and now it's just an unspoken requirement that, okay, doesn't matter if you're a CEO, just write your own damn email.
So prompt your own AI and figure out how to use AI agents for your own work-
Donald Thompson: Yep ... [00:21:00]
Don Shin: because everyone has work to do.
Donald Thompson: Gotcha.
Don Shin: And AI can help people at every single level of the organization, c-suite, knowledge worker, everyone in between, management, no exception. So that is the biggest change that needs to happen and is happening, and that's the advice that I would give, is you've got to use it day in, day out to get a sense of, like, where does it fall short, where does it really excel, and then lead the charge and lead by example.
Donald Thompson: I agree with that wholeheartedly. I'll give my personal testimony. I spent, time recently at, I went virtually 'cause we had a big snowstorm in Boston, but I was supposed to go on-site at Harvard for an AI strategy conference, and there were business leaders from nineteen different countries, eleven different industries, and we really talked a lot about AI adoption in the enterprise.
And one of the things I learned from this session that was really interesting data point, and I want you to comment on it, th-the number one use case for AI [00:22:00] was therapy and companionship And that to me was different than I expected, right? Why do you think that is? And I know that's a little off the pure biz, but why do you think that is?
Don Shin: firstly, because it's free and it's cheap, and it does it pretty darn well. It's all about token prediction, and so if there is any language around therapy, any kind of language around empathy, it's going to mimic that language and generate it really well. But I think the biggest reason is that we're lonely.
We're a lonely species right now. we had that COVID funk where we all came to realize just how important human interaction was to us, not just virtual interaction, but face-to-face. And yeah, we're out of it, but I don't think we've truly gotten out of the loneliness funk of people actually prioritizing relationships, over other things in life.
(Clip - The Danger of AI is Sycophancy)
Don Shin: And the other thing that I think is also really dangerous is sycophancy, and [00:23:00] that AI, if coached and prompted, will never disagree with you. Human relationships are tough, Don. You know that. Yeah. The older you get, the more you realize it's hard to keep friends, it's hard to cultivate friendships. Yeah. It's hard to keep them going.
It's hard to maintain connection over time. It's hard to find a mate these days if you don't have one. It's hard to date. It's hard to, expand your circles. It's hard to find third spaces that are still in business and still open- Yeah ... to the public, right? It's harder to meet people. And I think what's really dangerous for the younger generation is that AI can provide an all too easy way to simulate companionship and simulate empathy and connection without any of the difficulties and the hard work and the investment required to maintain human connection and relationships.
So there are some real dangers there as well.
Donald Thompson: No, I appreciate that very much. And what I share with people my perspective on AI-[00:24:00]
I encourage them to read the little bitty fine print that's underneath the prompt of every AI tool. "Don't trust me, I make mistakes." Every single one of the tools says that.
And so when you read those things, you have to recognize that it's a consolidation of information, but it's not necessarily the best representation-
of what you should do-
Don Shin: Yeah ...
Donald Thompson: standalone.
Don Shin: Yeah. Can I flip, the example on the head, though?
Donald Thompson: yeah.
Don Shin: so sometimes I use AI for therapy or at least a- Yeah ... form of self-reflection.
Donald Thompson: Yeah.
Don Shin: I've created, a chat environment called, Speak To My Future Self. That's awesome. And... Yes. And I've shared with it, I've shared it with it some of my dark, and deepest secrets, and basically prompted it to, assume the identity of Don Shin [00:25:00] from 20 years from now, and what kind of advice would you give to a younger version of yourself?
And then I'll sometimes chat with it and bring up, some kind of things I'm going through, and then just use it for food for thought and reflection. It's actually been a really great exercise. and it's really come up with some things that have challenged me in the way that I think about human relationships, my work, and et cetera.
And then I have a local version running on my computer, that is completely private, doesn't send any information to LLM services, and that is even more of my, vulnerable, raw self to process what is going on in my life at the moment. So-
Donald Thompson: That is good stuff. All right. So you're, you're sharing with our audience authentically, deeply.
So I'm gonna share one of my use cases with AI.
we'll see if this makes it to the episode. So I had a few adult beverages, and so I started asking Claude some just interesting questions, and, I'm messing with Claude. So I'll, [00:26:00] share one that's, is fun. Uh, I'll share two.
So one is I said, "Claude, I have this great idea. I want to stay connected with my friends. I want to share pictures. I want them to be able to, to, things that I share are good. I, want to be able to use it on my mobile device and different things. What do you think about my idea?" And Claude says, "I think you're talking about Facebook."
Don Shin: Yeah. "
Donald Thompson: And I think the way that you're talking to me, you might have had a little bit too many adult beverages- Oh my God ... 'cause this isn't a business you should pursue." And then I said, Claude, let's move off of business." I said, "Who's smarter?" Me or Jackie Ferguson, who's my wife and my business partner.
And Claude said, "I am not entertaining this conversation because there's no right answer, and it might make you switch to Chat." And it gave this whole funny... And I won't go into the thing, but what, it, aligns with, yours was a little bit more, serious, [00:27:00] mine was fun, is that the machines are built to lock you in to working with them So if you actually want good feedback on your presentation, good feedback on the things you're talking about personally, you have to tell the machine what mode you want it to be in.
You have to tell the machine if you want it to be very critical, if you want it to be the persona of a Hollywood director that could green-light your movie if you have a movie script. If you are doing a presentation to pitch to a venture capitalist, you should name the VCs that you admire, and then you can get feedback in the persona that can help you grow.
But left to its own devices, it is super reassuring-
Don Shin: Yeah ...
Donald Thompson: about everything that you do. And so you have to make sure that the compute- the machine is speaking to you in the way that you can gain value-
Don Shin: Yeah ...
Donald Thompson: not just false affirmation-
Don Shin: Yeah ...
Donald Thompson: of, what you're doing.
Don Shin: Yeah. Can I give a quick tip?
Donald Thompson: Yeah, please.
Don Shin: Yeah. Please, (Clip - PROMPT Framework) I created a prompt framework, conveniently [00:28:00] called PROMPT, P-R-O-M-P-T. P stands for persona. Assign a persona to the AI that you're speaking to. R stands for request. Make it clear and specific. Don't ramble on. Be straight to the point, clear and specific. O stands for output.
Specify the kind of output you want. Are you wanting a tagline, a paragraph, two paragraphs, a page, a memo, a research dossier, or an image, something completely different? M stands for mandatories, any kind of requirements or constraints, like you must consider, the material that's already in my client folder.
P stands for priority. Finish your prompt with an above all else statement that reminds the LLM what it should consider most because LLMs, the way that they're designed, have something, what's called recency bias. They weigh what's most recently in the chat thread more than what's in the middle of the chat bed- thread and sometimes the beginning.
And then T, if [00:29:00] you're into agentic prompts, men- mention or call out any tools that you want it to use, like searching the web or reviewing your Slack messages or checking on your email as part of the response. So PROMPT, P-R-O-M-P-T. if you follow all of those things, it's almost a certain, shot to get a quality response.
Donald Thompson: You, uh, gave the team one of your snippet clips. That's awesome because a lot of people are trying... I went to a 16-hour prompt engineering class, so I was following you as soon as you described it. But you did it in 60 seconds. that would've saved me a lot of time. Because that was a, you know how we have TED Talks- Yeah
or we have... that, that was a micro TED Talk on prompt engineering in that statement-
Don Shin: Yeah ...
Donald Thompson: that most people can do and apply. So that's really cool. And I think the thumbnail that, that I want to share with the team is that we're all gonna need our own personal agents, but don't make it more [00:30:00] complicated than it needs to be.
Don Shin: Yes.
Donald Thompson: That the tools are actually getting very user-friendly, and it's really a function of, y- anything you're educated about, you're gonna be more confident.
So then start playing with the tools.
Don Shin: Yes.
Donald Thompson: and it doesn't matter where you start, it matters that you start, and then you'll, push through that anxiety level-
Don Shin: Yeah
Donald Thompson: and then figure out what, AI means to you. as we land the plane, 'cause I could talk to you all day, right? And, and, I love learning and, and, our conversations. And I took a bunch of notes in, our last, meeting. We were talking about, there's the large language models. We were talking about the small language models, right?
And I was geeking out. What have I not asked you that you'd like to share with our audience as we land the plane for this discussion?
Don Shin: How are you going to continue using AI when token costs explode 5 to 10X? What are we gonna need to do in order to continue to benefit from the advancements of AI?[00:31:00]
Because as I mentioned earlier, the tokens and the, the, tokens being generated are being generated, subsidized from venture capital and investment. Sooner or later, these companies are gonna have to break even.
And you can't break... You can't, solve a problem on scale if you're losing money with every additional customer or every additional user.
This stuff is expensive, and it's gonna get even more expensive as adoption of AI starts- Increases ... increases and, starts to become accepted. So how do you navigate through that? I think that not only will everyone need to know how to make agents work for you, ideally when you sleep, but everyone is ne- gonna need to understand some basic context management and token management skills.
Understand just the basics of understanding how an LLM works, that, you get tolled for the traffic you send up, and you get tolled for the traffic you get down.
Donald Thompson: Yep.
Don Shin: [00:32:00] And also how to manage that context and refresh it when you change the subject. A lot of people will just, chat back and forth on the same thread, and that's going to just make the toll cost more up and down, So I think that as we enter into a world where, AI is gonna get more powerful but al- also more costly, w- everyone's gonna have to have a baseline intuition as to how to keep costs down because... I was speaking to a gentleman this morning He's a software engineer. He's been out of work. He's looking for an opportunity.
He's-- he says he's dabbling with Claude Code. What I told him is, "You gotta immerse yourself in Claude Code if you want to stay on this track." "You gotta immerse yourself in agentic engineering. If you want to stay an engineer writing code, you're gonna have to immerse yourself in this in order to stay up to date."
And he said, I only got the pro plan, the twenty dollar plan, and I run out of [00:33:00] tokens, and then I can't use it anymore." And so we're all gonna have to figure out how to be maximally productive with as few tokens as possible. that's gonna, require a little bit of understanding how the technology works.
Not enough to build your own LLM, but enough to use it effectively in the same way that people need to know that, okay, when you press the pedal on the metal in a car, you're burning more gas, That's exactly right. As I think and reflect on our conversation and all the things that we talked about, a couple things as we close really come to mind, right?
Donald Thompson: One is that, AI is a complex set of systems, but there's some simple rules that you can follow to get more, better outputs, and that's prompt. And so I really appreciated, that, and I'm gonna use that. I'll try to give you credit every time I use it, but, but I'm gonna use that. The second thing that really jumps out at me, and this is the CEO conversation, is you have to lead from the front.
Yes. you can't just [00:34:00] dictate, "We're all in on AI." You have to share with people what your learning experience and journey is, right? And then you have to look at the organization's readiness to adopt change at that magnitude, right? And I try to encourage CEOs to educate first, experiment, right? And then you'll make better decisions on what you execute against, is the way that I like to describe it.
And then the final thing that I'm taking away from this conversation is that even though we're talking a lot about technology, we're really talking about how to help people be more productive so that they can spend more time in life on things that matter. and so I really have appreciated, you taking the time, to spend with us.
I am hopeful that we're gonna continue to do more and bigger things, together in, the future. But, Don, thanks for being a part of it today.
Don Shin: Thank you. Appreciate it.
Donald Thompson: It was awesome. [00:35:00]