Intro to Steve Moritz – the perfect guest to join as 2x AI Founder and longtime IT implementation consultant
Marcus Schafer (00:05)
Welcome back to episode 14, where we are making better informed decisions around your life, business, and investing. This is with me, Marcus Schafer Greenspring Advisors, Director of Growth, and Pat Collins, our CEO.
Patrick Collins (00:23)
We had a great episode this episode talking with an expert Steve Moritz around AI artificial intelligence. There was really four things that I took away from it that I thought were great pieces of kind of nuggets that you can use in your business or everyday life. First is Steve really kind of got into the weeds about four business use cases that everyone can be thinking about right now on how to
make AI, how AI can make their business a little bit more productive. He then took it maybe to the next level and talked about some ways that companies can really innovate outside of maybe some of the generic tools that are out there that I found to be fascinating and obviously got the wheels turning around how that can help our clients and even our firm think about the future. The third thing I took away was how he talked about leadership.
inside of a company having to kind of endorse this and really bring it as a key topic for the firm and for the employees of a firm to kind of embrace. And then finally, trying to make it a balanced approach of how we looked at AI, Steve really talked about the cautions, the risks, and even the hype around AI and whether it’s warranted.
It was a pretty far-reaching discussion about all things AI. And I just found that there were some really fascinating takeaways.
Marcus Schafer (01:40)
Yeah, Steve was just an awesome guest because he’s obviously an advocate. This is his job is to go out there and help companies implement AI solutions now, but he was balanced around, here’s maybe some pitfalls. Here’s where maybe the hype’s a little bit ahead of the curve. So I thought that was great. A little bit more about Steve. he’s such a fantastic person to have on because…
He’s a technology implementation consultant, so IT consultant. And what’s fascinating, we were doing some research and looking at studies, trying to understand, is AI actually having an impact? And I’m reading about these different tech trends and the impact they’ve had, right? So it’s like, hey, internet, cloud computing, robotic processes. And then I’m reading Steve’s bio and he’s like, hey, I’m implementing cloud computing, robotic processes, machine learning.
So it was super, super fascinating just to see the correlation that he’s actually worked on a lot of these past trends. And then he’s a consultant to help people put this into practice. So he is a founder of two different companies right now. Those are Inspiration AI and Accelyst.ai And then he’s also connected with a VC, Oxygen Ventures. certainly, guy’s working.
Guy is absolutely working. He drops knowledge for us. So that was great. and then you and I, Pat, I thought this was also super interesting. We just stayed on after the, after the conversation to talk a little bit next level. Cause I think the natural way the conversation was going is, here’s how you can use this to increase your personal productivity. Here’s how companies can try to take this to increase their productivity, re: profits. And then the
The question that we had is, what’s kind of the evidence around, is that being factored into stock market prices? So what are the investing implications of this? And we stayed around to talk about.
Patrick Collins (03:37)
Thanks, Marcus. Great comments about Steve and the conversation we just had.
Just as a next step, if you’re a listener who really enjoys Greenstream enjoys this podcast, we’d love for you to subscribe to it so you can make sure that you stay abreast of any new episodes that come out. And if you’re feeling even a little bit more daring, we’d love to have you share it with your friends or leave a review on whichever platform that you listen to it. So without further ado, here’s our discussion with Steve Moritz
What is GenAI? – replicative and generative use of data to make decisions
Patrick Collins (04:10)
Steve, how do you define artificial intelligence in maybe its simplest terms?
Stephen Moritz (04:18)
As simply as I can, Pat, AI is the use of data and compute to candidly make decisions, do things, generate outputs and outcomes similar to the intelligence that we as humans have. So it’s replicative, but it’s also generative. And I’ll explain those terms a bit more.
Later on, I’m sure when we delve into it.
Comparing GenAI to Past Tech Shifts – the ability to generate a decision is wholly unique
Marcus Schafer (04:48)
Absolutely. I’d love to hear like, how is this different than maybe past technology shifts or paradigm shifts? know, think kind of home computing or the internet, just like you said, this seems like an evolution of that. So what’s the similarities and differences?
Stephen Moritz (05:06)
Yeah. I mean, it is transcendental mostly because of the generative aspect. I’ll briefly explain that, Marcus. So artificial intelligence and something very similar called machine learning have been around for quite a while. Process automation has been around for quite a while. The real leap here is, and is.
really being catalyzed by a learning model called chat GPT. And that was launched in February of last year and had more users than any other similar platform. I mean, the growth levels are ridiculous, but more to the point, the ability to actually make data, take information and not just make a decision.
and replicate workflow, but actually create something. So that something could be an image. It could be a video. It could be a decision about whether to underwrite a commercial loan. It can be actual contract terms for that loan that are reflective of both a number of data points, both internal to a bank, as well as external.
The external data could envelop well, what’s happening in that geography? Is this a good loan from an economic development perspective? What’s happening with interest rates, currency exchanges? So the ability to bring all this data together and then process it, understand it, and generate an outcome is really what is a bit of a different kind of a transcendental shift from previously.
Why the Breakthrough Now? – compute capacity has led to an exponential increase in workflow efficiency
Marcus Schafer (06:58)
And is the, I mean, people have been talking about machine learning as the second coming for 10, 20 years. So what was the, what actually enabled the chat GPT? Like what was the breakthrough that enabled this? Why is it important to businesses and to people right now, but it wasn’t four years ago.
Stephen Moritz (07:16)
Yeah. Well, I mean, automation has always been important and, Marcus, he kind of asked two questions there, right? Why is it important now? And how’s it different from other forms of automation? How did we, and ultimately get a Chat GPT Now, you know, we now have, you know, a massive number of competitors and you may have heard, some of the models like,
Anthropic and Google’s got Gemini and we’ve got Meta and the Meta platform Meta AI. We’ve got Microsoft Copilot and the list goes on and on. Much of that has been driven from an ability to compute at much higher levels to be able to take in vast and greater amounts of data than ever before. So.
You may have heard about training training models. Well, again, it’s synonymous with training employees, training athletes. And what’s enabled chat GPT and all these other platforms is something called GPUs, right? And I’ll try to stay away from the acronyms. There’s a much greater ability to process.
structured data, so structured data are zeros and ones and nines and et cetera, and unstructured data. So to be able to go through a procedures manual or to go through call recordings of a customer service center. And you’ve probably heard of Nvidia, Which no one really had ever heard of 10 years ago, but they continue. And then there are competitors to create.
processing capability to underpin these models and their ability to go out and collect, harness, aggregate, and then to create value in the form of generative outputs through all of this data. So a lot of these firms are basically using the internet, right? Where there’s a vast amount of both legitimate and not so legitimate information. And they’re creating
outputs, generative capabilities for all of this data and the compute associated with it. That’s what’s different. That was one question, right? So another one that you asked is why is it important right now? It’s important in a sense, no different than other technology ways. So I think there’s an analogy here to the internet. The internet, believe it or not, for many folks,
probably can’t imagine a world when it didn’t exist. And for those firms that felt it was important to their business, their operation, their ability to sell and service, they have thrived. So as an example, can tell you Amazon stock is over the last 10 years up about 18 fold. Well, they originally sold books.
But they really understood the power of the internet to connect suppliers, to create marketplaces, to disintermediate sellers, to customize service and that type of automation. And now with the power of generative capability, you’re now expanding the aperture of what this technology can tackle from a workflow standpoint. It used to be, it’s really good at handling.
one of the transactional aspects of maybe accounts payable and accounts receivable, right? Well, now it’s also able to actually, as an example, create the terms that make sense, say, for a new vendor, right? And it’s using a number of sources, again, back to this term generative, to be able to generate a term sheet for a company that’s looking at using a number of different new vendors.
Right? So now it is even more profound in terms of its productivity potential than any of the other automation technologies that have come before.
The Biggest Misconceptions of AI – it will not take over the world, but it will reduce clerical jobs
Patrick Collins (11:31)
You know, Steve, it’s interesting. You brought up Amazon. I remember at the onset of the Internet, people saying, I’ll never put my my data in to buy something on the Internet. I’ll use it to surf and kind of find some information, but I would never put my credit card information in there. And now you have, you know, one click and it shows up at your house the next day and nobody cares at all about their credit card information. It’s become kind of common. But that I think was a misconception back then.
I have people now that I talk to that think AI is going to be the downfall of civilization. What do you think are the biggest misconceptions right now of AI?
Stephen Moritz (12:11)
that it will take over the world.
And let me briefly explain why I don’t necessarily think that is the case. And I’ll use some data to do that. This is another technology wave. And like any wave, a little bit unpredictable and they are of different sizes. This is a big one, right? This is tsunami-esque, if that’s a word.
The reason I don’t think that it will take over the world is that there are predictions, right, that it will eliminate jobs and it will, right? But as an example, Goldman Sachs says over the next, I believe, 10 years, it’ll eliminate 85 million jobs in the United States. I’m sorry, that’s global. But also it will create 97 million new ones.
Now that means that there will need to be some retooling of the workforce. So people who are in predominantly workflow type jobs, I am looking at a commercial loan application. I am taking that data that’s also got, maybe it has equipment schedules. It has balance sheets. has income statements.
And I am now taking that data and it’s sitting on my desk and I’m either copying and pasting and maybe I’m putting it into a loan underwriting system. And that job will go away because there really is no incremental value to shareholders to other employees in that job. So pseudo clerical, some of these white collar jobs, which is a lot of data transcription, data movement, those jobs will go away.
But the jobs that will be created will be ones that are associated with people’s ability to harness these AI platforms. So let me give you another example, right? As to why it won’t gobble up the world. And I’ll make this a quick one. And I’ll stay in banking. There’s a lot of data that once a borrower gets a commitment from a bank has to produce, it’s called covenant data.
And in banks, have a lot of people that are fetching this data, analyzing this data and moving the data around. Right. So with AI doing that, there’s one level of automation to that. But more importantly, the AI is also looking and analyzing that data and it’s making suggestions about which borrowers might we actually want to expand our relationship predicated on their financial strength, the performance.
of both the risk and the loan and the properties or whatever. So we’ll continue to see people who bring value through leveraging these AI tools and understand how to use these tools in their jobs. They will be more productive.
Pat, you asked about some misconceptions, too. Let me crisply, more crisply try to respond to that. So you mentioned the story about people a bit afraid of putting their data up there and the big cloud, the internet, whatever. And I’ll tell a brief story. I have a friend and we were talking about like,
I send you some Venmo, Venmo you some money. And he’s like, no, I don’t, you know, I don’t put my credit card out on the internet ever. I was like, you know, I respect that. And I said, Mike, do you ever go to a restaurant? He goes, yeah. He’s like, do you always pay cash? No. Well, you’re giving your credit card to a complete stranger. That stranger is taking your credit card and then going to another room where I’m sure there’s paper and pencil and cell phones that can take pictures of your credit card front and back. That doesn’t worry you? He goes, well.
I never really thought about it. I said, well, compare that. And this isn’t necessarily an AI thing, but compare that to the world where in the internet and in commerce, you have encryption. You have that data is called encryption at rest, encryption movement. have multifactor authentication. Now we all have read the stories about those databases with credit cards and social security numbers getting breached that can happen.
but it happens a lot less kind of in the cyber world than it does in the physical world.
Patrick Collins (16:49)
Interesting. It’s great. Yeah, it’s, it’s, it is always interesting, you know, the, the move to the cloud, for many businesses, whether it be their CRM platform or other operating systems, you know, people sometimes don’t, don’t equate the fact that you’re putting a lot of your business information. You could be on QuickBooks, for example, your financial accounting information is on an online system, but people don’t sometimes equate that to
The same thing with AI, for example. Well, they’re still both out there. That data is out there somewhere. And hopefully it’s being secured, as you mentioned, and probably much more secure than a piece of paper sitting on your desk. it’s a great story. Just maybe shifting gears a little bit here to get into, I think, probably the meat of the discussion, which is personal and business use cases. A lot of our audience.
The 4 Ways Every Small Business Should be Using GenAI Today – to enhance productivity and impact
obviously work in businesses, run businesses, own businesses. And I think what we’d like to do is try to give them some information on maybe, if they haven’t started using AI yet, or just scratching the surface, maybe what that could look like. maybe starting out is, where are you seeing maybe some of the best areas where small mid-size companies can really adopt AI to improve their overall operations?
Stephen Moritz (18:01)
So, you know, kind of on that simple level of AI adoption, I’ll speak to four different use cases or examples. And none of this requires massive investment. You don’t have to get a certificate in machine learning. But let’s quickly walk across this fourth hat. So meeting summaries, You know, thanks to the pandemic.
We all spend a lot of time doing zoom and teams calls. so using a meeting summary tool, there’s one called Otter, O T T E R dot AI or fireflies dot AI. So it doesn’t mean you should stop listening and figure like, I’ll just look at the otter.ai summary, but it really is helpful for people that spend a whole lot of time doing the conference calls to be able to kind of come back. Those summary.
⁓ engines can not only, you know, condense an hour or two hour into, you know, something that’s scripted, but also, identify what were the next steps, what were my next steps, what were Marcus’s next steps. that’s one, writing assistance, which is the bane in academia, right? But there are tools actually to assess generative output. That’s actually not, truly built on manual means, but there was writing assistance. You may have heard of
Grammarly, Hemingway Editor. Again, these are tools that, know, like, hey, I’m going to write a nasty email to a customer, a really good customer who’s behind on payment. And then you write it and then you ask these tools to say, can you make it a little more friendly? And I’m going to give you one other really quick example of writing assistance. ⁓ I was doing a video.
And I had to, for my business, and I had to write a script to fit into a 90 second video. And I sent my script to the firm that was helping me build this video. And they said, you know, this is way too long, right? You got to cut it down by roughly 30%. So I took the entire script and I said to the AI engine,
I need the messaging to be intact, but it needs to be reduced by 70%. Right. We got to cut out a lot. It brought back a new script and I was like, this works. Right. So using a writing assistant server as a second to third is image generators. So if you’re building marketing content or you you’re even trying to, maybe you’re a contractor.
and you’re trying to develop renderings or your interior design, you can describe, I want you to build me a sofa or a house. I want it to look like this. I want, I want it to evoke the following emotions. I want a palette color of this or that. And so there are tools like a Dall-E Canva, Stable Diffusion There’s a lot of places that you can just describe something where you can say,
Here’s a picture of something, you know, can you break it down into a bill of materials? Because I need to start figuring out how to do maintenance on it. So generative again, using that word again of, you know, images and videos. Um, it’s interesting. And this is maybe related to kind of image, but let’s also do voice music. There is a, where’s a group on Spotify that has 1 million users called Velvet Sundown.
And it was just recently revealed. There’s nobody in Velvet Sundown The entire, everything about the band is AI, right? The “people” in the band, right? The music. And again, it’s awesome. I’m not saying that that’s necessarily for a use case for a business owner, but it’s all kind of under the umbrella of.
image and voice in this whole generative capability. And then the fourth is just presentations, creating outlines, creating content. So I review a lot of pitch decks in the world of venture capital, and I have to produce a lot of pitch decks and, you know, we all unfortunately have to live in a world that is over-saturated with PowerPoint.
But being able to, again, rely on the tool, if you’re a business owner and you’re making a pitch, you’re trying to win a contract to do interior design. And you can say, take a look at this presentation and make it better or make it less fluffy, or let’s really emphasize some of these characteristics in our culture and our organization. I want to really make sure we highlight trust or creativity. And then you just hit the button.
And you can use tools like Gamma AI for Google slides. There’s an app called Tone These are going to make your ability to produce your communications material and do it faster. And again, it’s still you, right? Because there’s still something called human in the loop, right? That’s another acronym. So across these four use cases, there’s still human in the loop, right? It’s not like.
AI is going to take over your business, you still, but you’re using it as a tool to augment, automate and help create and generate maybe faster than you ever could on your own
Are Employees or Employers Benefiting from Increased Productivity? – this is a key business opportunity
Marcus Schafer (23:36)
Yeah, those four use cases are super, super fascinating because it kind of gets to this question of where are the benefits of AI going? Are they going to the individuals or are they going to the companies? Because I look at those four use cases and kind of some of the research I see is, maybe your productivity is increasing by about 3%, some more in different job descriptions, but are the benefits actually
coming down and accruing to the businesses in those use cases or are employees just getting some more time back?
Stephen Moritz (24:08)
Both. And again, let me tell you a story to give you this example. So I was working with a firm that produces speakers and lighting in the automotive aftermarket. And they have to produce what’s called firmware. And firmware is basically software that sits inside, maybe it’s a
an equalizer system for a Jeep or it’s a lighting system for a Ford F-150 pickup truck. And the firmware that they have to create has to kind of thread the needle between the hardware that they’re designing and the specifications and standards that come from these auto manufacturers. So they had a team of
Uh, 10 firmware engineers. it’s just relatively small firm. It’s about a $200 million top line. And it could take them as long as, uh, three months from start to finish to produce the firmware, then to get the product out to market. By using AI that was actually generating, right. using that word again, software. Now there was a human in the loop, the team analyzed it.
found some errors in the code, it wasn’t perfect. They were able to compress basically time to market and creating that firmware from three months to about one week. Now, it didn’t mean that that firm fired any of their firmware engineers, but it totally changed their capacity and their throughput.
and their ability from getting requirements to actually having tested, complete, and in the market firmware dramatically. And that put them at a competitive advantage. So to your original question, Marcus, that was really good for the firmware engineers, right? Because now they’re spending a lot of their time almost in a supervisory capacity, kind of as a human in the loop, ensuring that what was created by these bots was legit, was going to work, et cetera.
And again, ultimately it means that firm is worth more value. If it’s worth more value, there’s growth. If there’s growth, there’s more opportunity for compensation uplift, right? And, know, and that’s why, for example, McKinsey, the big consulting firm believes that this phenomenon is going to contribute about 4.4 trillion uplift to corporate productivity. That’s a huge glob.
that can be arbitrage both for the corporate or organizational gain. Cause it’s not just, it’s for profit. And I can speak to some for profit, not for profit use cases, but it’s good for the person. It’s good for the employee and it’s good for the company. It’s not good for somebody whose job it is, is just to move numbers around from one spreadsheet to
Prompt Engineering in Hiring- that helps future proof your business
Patrick Collins (27:09)
That’s great. I have one other kind of follow up to that question around how businesses can use AI. I’ve heard now from a business center who I think is somewhat tech forward saying that they’re actually interviewing people and part of their interviewing process around AI is something called prompt engineering where they’re asking people or testing to see if people can
do this, I guess. Can you explain what that is and what that looks like if you’re a company who says, I think this is the future and I want to find employees that want to embrace this and can embrace this? What is this prompt engineering and how would you go about thinking about finding employees that know how to do this kind of stuff?
Stephen Moritz (27:58)
Yeah. So, you know, it’s fascinating, Pat, that like, especially kind of in the technology space, there’s, there’s a lot of fancy words that are used to kind of obfuscate meaning. Like, so prompt engineering is just your ability to use an AI platform, to speak to it, but then also based on the response that you get to have the right type of iterative
Next question. And these models are built in such a way that again, let’s, let’s, let’s use an example. Like, so as I might be preparing for this presentation and I say, give me some metrics to explain what the business value of AI is in the United States. I can get an answer and it says it’s $4.4 trillion dollars. And then.
As a prompt engineer, I’m looking at that response and then I’m kind of thinking about the what’s the next question predicated. And in a sense, it’s kind of like a pipeline, right? It’s a funnel. So you start out and you prompt the AI model and you’re getting something that’s this wide. then because you really know where you want to drive that model, you say, well,
That’s too general, $4.4 trillion. I want to know is in the United States. And then you get another answer and you’re like, well, I want to know in the United States in the next 18 months. And then you get another answer. I want to know in Maryland and in Pennsylvania, and you get another answer. So you’re engineering this AI model prompt to continually winnow
and refine to get to the point where you’re like, all right, I think, I think I like this answer, but in a business setting, you can do something similar, right? You’re using an AI model at the bank and you’re asking it based on all this data that you’ve input to the model to give you a term sheet. And then you may look at the answer. You’re a human in the loop.
Patrick Collins (30:01)
Hmm.
Stephen Moritz (30:23)
prompting the model to say, you know, I’m uncomfortable with the length of term. So I want you to take another approach. Give me another answer with an assumption that interest rates are going to go up by 25 basis points because I have a confidence factor of that happening. And then you’re going to get a different response and so on and so forth. So you’re just iterating your way through what’s generated.
to get to something that you view as being more accurate and relevant and pertinent to the challenge that you may be confronting as an employee.
Patrick Collins (30:59)
Yeah, it’s almost like I equated to being a good delegator. Like you can’t just say, go just go do this project for me. Like you need to give more specific instructions. And as you get more specific, you get better results. And that seems like it’s almost part of being a good prompt engineer, if you will, when you’re asking these models to return data back to you.
Stephen Moritz (31:22)
Well, exactly. mean, if someone was running a wealth management firm and they went to their head of growth and they said, I want you to grow the business. And that person, if they’re smart, they’re going to say, well, in what products, where should we be emphasizing our growth across what demographic, which geography, right? So that head of growth is a prompt engineer coming back to you, getting more clarity and therefore their ability.
to ultimately respond to what you’ve tasked them with continues to get more accurate, aligned and relevant to the very challenge that you brought in the first place.
Marcus Schafer (32:03)
That sounds like somebody you’re talking about in specific, but we won’t name them. ⁓ I think Pat brings up this great point where a good analogy, think a lot of people are talking about AI today is it’s kind of like an intern. You give it a project, you turn away, come back, you get kind of an output, and you have to double check it because what I notice when I use it, you have to be knowledgeable enough to know when it’s just flat out wrong.
AI is Like an Employee – humans in the loop processes help plan for, detect, and solve for wrong outputs over time
Marcus Schafer (32:29)
And some of the use cases when you use these broad-based tools, sometimes they’re flat out wrong. I’m doing a mortgage refinance and I’m, hey, help me calculate if this is worthwhile, what’s the net present value of this decision? How long do I have to stay in this house for it to make sense? And it gave me answers that were just wrong. And when I asked it, hey, it should be this. And I told it why, like, why is it wrong? It’s like, yeah, that’s my mistake.
So some of it, just have to be super, you have to be knowledgeable enough to understand some of the limitations. But in a business context, we’re also seeing kind of some people, some employees using off the shelf solutions for business, which might carry certain risks. So maybe just talk about the pros and cons between kind of building something gated within your ecosystem for employees who probably should be trying to leverage these tools on their own.
What are the trade-offs that you think about?
Stephen Moritz (33:28)
Yeah, interesting. It’s a lot to unpack there, right? So, well, so, you know, maybe some of you have heard of the term hallucination, right? That you get hallucinations. And again, I might hear, you know, pitching that AI is perfect and accurate and always right. And again, the analogy is, it’s no different than other employees in a sense, right? It requires care and feeding. It requires continuous education. I’m sure Pat, that’s something.
Right. That in your firm, you know, you have HR and people working on, you know, continuous knowledge. But, um, so the, you raise an interesting point, whether you intended to or not, I’m not sure about your question about privacy and what kind of lives inside of an organizational ecosystem and what doesn’t and where’s, where’s the
paranoia and this also relates, Pat, to a question you asked earlier about some of the misconceptions and fears, right? So let me kind of break that down and as best possible use stories to do that. So let’s maybe for the moment talk about healthcare, right? So yeah, you certainly don’t want an AI bot taking in your personal health information and then
developing a diagnostic or a treatment plan and having it be intently inaccurate, right? I don’t think any of us would like that. However, that does not eliminate the value and the necessity of using diagnostic AI in the healthcare provider space. Now, always, well there should always be a human in the loop.
I’m going to tell you again, a brief story. I’m on the board of a health tech company that has a device that generates information, health information, personal health information about people with COPD, chronic obstructive pulmonary disease. And it moves that data to clinicians so that these people don’t have to come into the office, which is the way they used to have to do it.
They’d have to come in and use something called a spirometer. And in working with that company, I said, you know what, that’s great, right? Because that brings a lot of value. It saves time and money. But I said, we need a diagnostic capability, right? We need an ability to kind of generate a, so what? I have all this data, so what? What should I be doing with this patient? How are they doing? How do we improve their life? How do we automate and improve the velocity and productivity
of the clinicians in terms of managing treatment plans, right? So that AI that was built is an awesome saver of time. And that brings value both to the patient and the clinician. However, there’s a massive amount of scrutiny and testing and retesting to ensure that it is accurate. And then that is all backdrop by having a human in the loop to ensure that
that treatment plan as is suggested by a bot makes sense for any particular patient.
Shadow IT – the difference between leveraging off the shelf, internal, and hybrid tools
Patrick Collins (36:55)
Do you, when you talked about kind some of those use cases earlier on, you had to get four of them around meeting notes and some of the other kind of tools that allow you to create presentations and whatnot. Is there a point for a company, it sounds like for this healthcare business might’ve been the case, where you would say, you should really develop your own AI tool or your own bot internally.
versus using these off the shelf products and is there better use cases or times when you would expect a company to start thinking about that or how do you kind of go through that in your mind?
Stephen Moritz (37:32)
Yeah. And I think the headline to answer your question, Pat, is it really often is a hybrid architecture. so let’s, know, those examples that I gave you earlier for, you know, for how a, maybe a small ⁓ owner can use those tools. Those are for the most part dependent on, you know, these commercially available GenAI platforms.
And they’re very valuable, right? But what I’ve seen and what I’ve worked with is a number of companies and organizations that kind of have a corporate mandate, have a corporate policy, a corporate approach. In the world of IT, there’s something called shadow IT. And what that means is that you have the non-IT people doing IT because they’re frustrated on how slow it takes.
for IT to do this stuff. And then it becomes a shadow and everything’s fine till it breaks or it doesn’t work or it’s inaccurate, right? Where you get blue screen or black screen and then everyone goes running to IT going fix it. And IT goes, I don’t even know what this is, right? So to have a policy and approach to have an organizational control over the use of these GenAI platforms makes sense. It’s not a heavy lift.
But to your original question, I see more and more of my clients using kind of a hybrid approach. And let me explain briefly what that might look like. I’ll use an example of a healthcare payer, right? A healthcare insurance company. And that’s working on or we’re building an artificial intelligence bot to help them with something called
pre-authorizations, which is very much in the news. And I think unfortunately for any of us that have had to get insurance to pay for something that’s significant in cost, the insurer has to generate a prior authorization. So that’s reviewing a lot of data, but back to the point about a hybrid solution in this space. using a GenAI platform that may help with something called natural language processing. So
Wouldn’t it be great to be able to take what might be medical records and understand those that are literally on a page. It’s not hard data. Wouldn’t it be great then for someone working for the health insurance company to say, hey, to the prior authorization bot, I’m going to send you this information. I’m going to ship you some medical information. But I also want to tell you a little bit more about this, this account, right?
This is a major account for us. So we’ve got to be really careful with, with any denials. And so you’re speaking through natural language and there’s an engine that’s understanding everything that you’re saying. And maybe you’re using Anthropic to do that. Right? So these major billion dollar investment corporate enterprises, they want to make money by selling subscriptions, right? To corporate users. But now it’s a hybrid solution because
In this particular example, you may actually want to tap into a lot of data that you have in this. Maybe it’s a Blue Cross Blue Shield, one of the many Blue Cross Blue Shield agencies. You have a lot of data about that particular policy. You have data about that particular patient, about their family, about where they are in terms of maybe copay. And there’s a lot of other data that’s internal. So meshing.
the internal applications and the internal data. And then also then using these public commercially available platforms is an architecture that you see more and more. Because to your point, Pat, what people want to do is to be very careful in using all this automation and using these AI platforms, but staying compliant. Right? So in the world of healthcare, you have
private health information compliance, have HIPAA compliance. If you’re doing this in Europe, have GDPR compliance. So you have to remain compliant. That’s one concern, going back to one of earlier questions. And two, you also want to keep your secret sauce secret. So a bank that has many more rating factors than their competitor, which makes their pricing more accurate and reflective of risk, wants to use AI.
but they don’t want the rest of the world to know that they’ve got 27 % more rating factors than their competitors. So they build these systems, they go out through something called an API, which is another IT three-letter acronym, to use these commercially available systems. And then they often architect the answer and the response into an environment that is secure.
Now that can be on their servers and their data centers, or it can also be in the cloud, but it is contained and separate from anybody else and not to get into the architecture of the cloud. But that is capable, the ability to build these complex architectures and do it in a secure way. So you’re not violating compliance or HIPAA, or you’re not revealing secret sauce for much, much more of that. And that’s fueling this expansion of AI adoption.
AI Privacy, Security, and Transparency – the evolution of the cloud means that it is often safer than employer secured networks
Patrick Collins (43:07)
And it sounds like from some of the client stories that you’ve told, it’s probably in the most either regulated or, you know, areas of data confidentiality that you have to be healthcare, banking, like these are areas that, you know, it sounds like you’re seeing AI adoption with your clients there. So I guess the follow-up, and I think you’ve answered this, but it sounds like there’s a high degree of confidence, at least with your clients, that this data
even though you’re using a third party kind of commercially available platform that’s hybrid with your internal data, you feel confident, your clients are feeling confident that this data is protected. And we know that this AI tool or agent is not sharing it with other, you know, in other ways.
Stephen Moritz (43:52)
Yeah, I think there’s growing confidence. And I think in part, Pat, the reason that is the case is because of the cloud, right? So the cloud is to be as simple as I can. The cloud is just somebody else’s computer. It used to be our computer, now it’s somebody else’s computer. And so that was…
that was slow in terms of adoption for a while. And then it just became a snowball, right? Tsunami. And people got very comfortable with it. And one of the reasons people got, one of the reasons being, so if you’re a $200 million company, you may have one person whose job it is, right? You may have one person who is your chief information security officer, your CISO. That’s a one person.
organization at this $200 million company trying to keep the network secure, the data secure, the application secure, whatever. Contrast that with them maybe getting rid of their data center and getting rid of their servers and putting that out, whether it’s the Microsoft or Google or Amazon, where there literally are thousands and thousands of people whose job it is to provide security. And by the way, the reason they have to make that investment is
If they can’t portray security, if they can’t delivery security, deliver separation. I mean, you have HIPAA clouds, you have GDPR clouds, you have all of these safe environments. So that’s one reason there’s greater comfort with it. And, you know, I got some, I think there are interesting statistics about how that comfort is translating into demand. Right.
So every day you read more articles about, you know, new adoption and your use case. So again, this is from McKinsey over the next three years, 92 % of companies will increase their AI investments. Only 1 % consider themselves mature in the AI space and 47 % of C-level executives believe they’re moving too slow with AI initiatives. So that’s a reflection of the demand.
Patrick Collins (45:59)
Hmm.
Stephen Moritz (46:03)
right? And the dynamic for more.
And last point, one thing that
Still, think, coming back to the earlier questions about analysis and misconceptions, AI can still be a black box. It can be opaque as to why did the bot decline this loan. And that can get companies into trouble. So it’s very important not just to be secure, not just to design accuracy.
but also to have a developed ability to articulate all the transparency of what’s happening inside the body. It is, I think, good corporate organizational responsibility to be able to
to tell the story of what’s happened inside the black box.
Data Risk is a Top Concern – garbage in, is garbage out
Patrick Collins (46:59)
So, obviously the last 30 minutes or so, we’ve been really focusing on the use cases and the benefits of AI, maybe just transitioning a little bit. What do you think are the major risks associated with incorporating AI, I guess mainly in a business operation? So small, mid-sized business, they decide we want to adopt this. What are some of the biggest risks they should be thinking about when adopting it?
Stephen Moritz (47:26)
Um, so one, uh, data risk. And so hopefully, and Pat, the story that I was talking about kind of these hybrid architectures where you’re really relying a lot on public data, internet data, large language models like Anthropic, Claude, Chat GPT.
The data that sits in there is not necessarily clean and legit. And there’s, you know, again, we can spend another hour on IP protection and, and challenges around copyright, but more importantly, and more around a risk that can actually be addressed from an internal perspective, right? We’ve all heard the term “garbage in, garbage out”. If you’re training a model because you want to automate some customer service functions.
You have to ask yourself, do we even have the data? Right? So if you got 50 people in a customer service organization and you don’t record calls and you’re not able to digest or kind of digitize, what was the issue that our customer had? Let’s what was the root cause? What was the resolution? What was the resolution quality? If you’re not creating that data, then there’s.
nothing to learn from. So there still remains a lot of work that companies have to do. This is not an easy button, right? That you’re like, hit the button. Now there’s some really great platforms that are called low code, no code, where you can kind of build some AI agents fairly quickly. However, and I will continue to emphasize to your question about risk. Do you have the data? Is it healthy? Is it accurate?
And also associated with that is have we been able to take unstructured data and make it — unstructured means it could be on a page, right? It could be a picture. It could be a conversation between a call center agent and a customer. — Have we been able to kind of transition that into something that can be understood by either our model that sits inside or a large language model? So that’s one, you know, there’s always the risk of accuracy.
When you think about the risks of employees going rogue, employees not necessarily doing their jobs well, making mistakes, we have the same risk here, right? Now, the difference is if you deploy a bot to replicate the workflow that used to be done by a hundred people, you now maybe have a hundred times more risk, right? So,
How do you do that? You test, you test, you test, you make sure you’ve trained and you’ve trained well. So that’s a risk. Transparency versus opacity. That’s a risk, right? For compliance purposes, employee purposes, customer relation, maybe investor relations, being able to articulate, you know, how and where and why you’ve deployed automation.
Um, you know, it’s really important, but again, I like analogies, right? Um, nobody really cares about the software behind, uh, an ATM until you hit the button, you ask for a hundred dollars and it gives you $80. Right. So, you you got to get accuracy built into this as much as possible. So that’s risk. And then we mentioned before, Pat, you know, kind of unsanctioned.
Patrick Collins (50:57)
you
Stephen Moritz (51:11)
organizational use of these models. because again, you could have an employee who is producing marketing content. And I, you know, I’ve worked with companies that are using AI for marketing. It’s a beautiful use case, right? Cause you can do what’s called hyper personalization, right? If you have a thousand employees, can, or I’m sorry, a thousand prospects or targets, you can create a thousand different pieces of content.
You can have AI determine where that content gets distributed because each of those prospects maybe has a different consumption pattern. Some are heavy on LinkedIn or other social media channels. Some are snail mail. However, to the question, are you sure that you haven’t infringed any copyright or IP with what you’ve produced in this unsanctioned use of a learning model? So those are about four or five considerations. I think each one.
can be mitigated, but not necessarily taken down to zero.
Vertical and Horizontal Integration – how AI is changing the job market
Patrick Collins (52:12)
Got it. Yeah, it’s great. staying on kind of maybe risk disruptions. You know, I have two boys that are in college. Marcus has got two much younger ones that are going to be coming up. I think a lot about what the workforce is going to look like in the next five or 10 years as they go into their careers, whatever that may look like. So for anybody listening,
that is thinking about that part of how is this disruption going to impact industry- specific jobs? We talked a little bit about kind of specific job functions, clerical, data, but is there any comments or thoughts you have around, you know, some of the listeners thinking about what are the things I should maybe avoid or be really careful of, or maybe there’s also outside of getting directly into AI and computing and whatnot there.
Is there other areas that you think won’t be as impacted by AI as others?
Stephen Moritz (53:10)
So the first part of the question, I’ll use your boys as an example, right? Like, so, you know, what I would say to them as they’re, know, endeavoring on their educational pursuits is I said, you know, I think it’s really important for you to understand how you can use AI to augment, underscore, right, your educational journey.
which is very different than saying, really hope you can figure out how to use AI to, to, to kind of replicate or be a replacement for your educational journey. Right. So those are two very different activities. So yeah, they need to learn how to use sanctioned AI to complement the things
that they will do in their life, things that they need to learn if they’re in an educational setting. And by extension, when they go out into the workforce that they’re working for a 501c3 or investment bank, as Marcus, you had said, their ability to say, look, maybe they’re finance majors or they’re creative arts majors, but their ability to articulate how they can do
prompt engineering, their knowledge of these AI platforms and which one is good for certain features and functions may be a differentiator.
you could have asked him this question 25 years ago about the internet. You could have said, you know, it’s kind of this high tech thing, right? Is peer to peer network or whatever, but I guess that’s banks, healthcare, right? And, ⁓ you those, so it’s really kind of everything, but let me, let me drill down a little
So there are both industries and specific use cases that I think maybe could have come to the top. from a use case perspective, recruiting, you know, nobody wants to look at a hundred resumes recruiting in the world of IT writing software. Right. So there is a lot of platforms that are very good at.
at, you know, again, you know, to toil over is that a semicolon or backslash and I put a backslash, but it should have been a forward slash and through rigorous testing, you reveal that. So software coding. So there’s data that says that actually right now, the various platforms that are out there right now are generating about a 14 % reduction in labor hours associated with, with, with software development. So there’s the horizontal use cases.
software development, marketing, HR, customer service. There’s a lot of data that says the majority of people actually would rather have an accurate conversation and resolution through a bot than a person. So those are some horizontal use cases. Now from an industry perspective, mean, you know, financial services, ⁓
healthcare, ⁓ believe it or not, manufacturing and you might like, well, that’s bill of materials and you got the cost of goods sold. But, you know, in that particular case, you’re seeing, for example, an automotive and aerospace at 50 % through using AI, 50 % reduction in time to market, a 30 % cost reduction. And then I want to mention one very specific thing. It’s kind of a personal passion of mine, which is agriculture. You might think.
You know, ag, right? You put some seeds in the ground. So what? So 90 some percent of the farms in this country are owned by basically small farmers. And that leads of course to 10 % for the large, highly scaled commercial growers who are crushing it in terms of productivity and their use of IT and AI. So actually I’m working on an initiative called Keystone Agritech
Initiative. which is really to bring a model platform to all of those 90 some percent small farmers who are struggling that will involve robotics, AI, a lot of data. And again, to bring some life to this. Imagine the power of being able to do a soil test or to take pictures of images of crop health or particular
insects or weeds and be able to leverage an AI supported platform that is taking into consideration gigabytes and millions of other farmers, right? And it’s giving them abilities. So they’re applying pesticides and it’s telling them actually on each particular meter of property, how much to be putting down where it should go equalizing.
kind of their ability to compete. And that particular initiative, of course, is not only very interested from the governmental perspective, but we’ve got large retailers who want to reduce their dependencies, who don’t want to necessarily have to navigate some of these terraforms — and they want to do controlled environmental ag. But basically,
vertical integration. So you have Walmart, have Kroger, they’re going to own their own buildings. And in those buildings, they will be producing the agriculture that they themselves sell. again, there’s a really, I would challenge anybody who might say, you know, this industry is just not going to be touched. And last point on that, I think I told you earlier.
I did a speech about 20 years ago when I was talking to folks about the internet and I had a number of folks kind of with their arms crossed in the back of the room and they’re like, this is just the fad, dude. This is not really going to be important. And I think they were wrong. And I think history kind of proves that out. And I really do think that, you know, this again, it’s another technological wave into your question.
I do think it’s pertinent and they’re both horizontal use cases and specific workflow agents that are highly applicable and potentially very productive virtually in every industry.
Is AI Overhyped? – it augments your business but doesn’t change it
Marcus Schafer (59:40)
Do you, ⁓ last question for me, because I know we’re kind of bumping up on time. I don’t think AI is wrong, but do you think it’s overhyped in any way? Like it can end up being a distraction for certain businesses that are trying to pursue kind of evolutionary enhancements, not just we can make a presentation a little bit better, but hey, we’re really going to change the way we do business. And the parallel I think about is self-driving cars. They’ve been promising that since 2015.
And it’s not like we don’t think it’s gonna happen, it’s just taken longer than expectations. So do you think it’s overhyped in any way? Or how do you think about expectations for timelines?
Stephen Moritz (1:00:18)
Yeah, I think it can be overhyped. here’s why I say that. The fundamentals of running an effective organization.
grounded in concepts and precepts, strategies and culture that have nothing to do with AI. So if you’re running a wealth management company, your ability to connect with your employees and motivate them to foster strong relationships with your customers.
to make them a priority, to treat them as if they’re the only customer you really have. Your ability to cultivate and curate your products and services so that they resonate and meet demand to be able to understand really what your customers’ pain points are, whether you’re in not-for-profit or you’re selling widgets.
to understand your costs and ensure that your revenues and your costs, like again, this could be an MBA class and it could have nothing to do with AI. AI is fundamentally a technology that can uplift the productivity of where you’ve already invested in your people, in your, you know, there’s a lot of AI for,
for maintenance, for manufacturing plants, right? That get ahead of equipment before it actually breaks down and takes down the whole line. So it helps that. But if that manufacturer is making the wrong product and doesn’t have the right alliances and sales and distribution agreements and is pricing their product wrong and is buying raw materials ineffectively.
they will not succeed. AI might be able to help them with supply chain. AI might be able to help them with how to arbitrage, hedge currency risk, but it doesn’t change the fundamentals of organizational achievement and success. It’s augmentation. It’s generative capabilities are still generating things that people used to generate. I used the example Velvet Sundown,
the group that has a million followers. So it’s not a bunch of people, it’s a band, but they’re generating music, it’s meeting demand, right? So, you I think some people can get intoxicated by the technology and forget, you still have to take care of your customers and your employees and your investors. And you know, no bot is really going to be as good as some of that personal relationship.
necessities that underpin all those relationships.
Small Business Benefits – by leadership endorsing sanctioned uses and finding workflows where data is in motion
Patrick Collins (1:03:14)
Great. So my final question is, if we have listeners that have listened to this whole thing and says, and they own a business or they’re running a business and they say, I definitely see where this is going. We haven’t done anything yet, you know, other than maybe some of our employees are maybe, you know, using chat GPT or something like that. What do you think the best way to get started for a business is just
you know, how do you start with this? where would you start if you were just running a business and realizing we need to start to adopt some of this technology?
Stephen Moritz (1:03:46)
Yeah, I run down two complementary tracks, Pat. First, go down those four use cases and I would again try to legitimize and sanction and from a central perspective say, this is the ordained way, we’re gonna buy a license.
We’re going to, we’re going to not just keep using the free version of chat GPT or Claude, because they all have limitations. So I would invest in an organizational license for a large learning model. And I would be very specific as far as the use cases. I would run some training. I would try to make my people prompt engineers as best possible and say, look, here’s, here’s what we are going to use. And here’s kind of how I would like to use it. Let’s show how we’re going to use it when we do teams calls or zoom calls.
And I I would run down that track. And I think that can, that’s kind of some low hanging fruit. think you’ll get some real nice ROI from that effort. Then the other track is I really look across my employees and the workflows. And, and I really be circumspect to find where a lot of data is kind of maybe, you know, being asked for and solicited.
And kind of, you know, what’s just kind of getting moved around a lot where’s the copy and paste. Where’s the person who’s, know, getting a loan application or grant application or, you know, a request for a CT scan, right. In the healthcare setting and then going to a user manual. So there’s your manual intervention, but there’s not really a whole lot of ultimate value to the organization.
There’s a lot of rules-based processing and I wouldn’t scare that person as I try to understand their job in the way that I I didn’t scare any of those portfolio managers of the bank that were spending so much of their time just managing spreadsheets because they wanted to get out and play golf and take these folks to dinner because a lot of their income could be made more on commission than base pay and I’d say
I want to understand this role. want to understand this workflow because I really want to try to unencumber. It’s a word I’ll use a lot. Unencumber you from things that maybe aren’t as valuable in your journey, your career journey. I want to unencumber you to bring more value to the organization. Now, without making a promise, that ultimately mean they may actually earn more money.
So I would look for those jobs and those workflows. I’d create a back– a backlog of those workflows. And then I would look for some of these platforms that are very good at building agents, right? That replicate the workflow. And I start looking at how, you know, and some of them are, they’re based on what’s called low code, no code. So you actually don’t have to be a computer programmer.
I also might see if my own IT shop feels that they’re capable of turning some of these workflows into agents. If not, I’d say maybe we should get some outside assistance.
Patrick Collins (1:07:07)
Great. Well, I want to thank you. I this has been fascinating for me to just learn about AI for an hour here, just to really dive deeper than I think what probably a lot of our listeners have done. So we appreciate you sharing your knowledge with us. And thank you so much.
Stephen Moritz (1:07:27)
it’s been my pleasure and in case anybody wonders, I’ve stumbled a couple times with these answers just to prove to you guys that I am not an AI bot.
Marcus Schafer (1:07:37)
you
I love it. Yeah, when Pat was asking about impacts for my kids, my number one fear is, hey, they’re not going to make more Cocomelons (a popular kid’s TV show) with these AI tools, which you guys might not understand, but any young parents listening will definitely understand that that is, would be a travesty for all of us. All right, Steve, thank you.
Stephen Moritz (1:08:58)
My pleasure. Hey, really enjoyed it. I appreciate you giving me the opportunity to chat.
Patrick Collins (1:09:12)
Thank you.
How Does AI Affect Investing in the Stock Market? – if the promise is a productivity boost across most industries, AI should be a rising tide that lifts all boats
Marcus Schafer
Great conversation with Steve. That was awesome to have his perspective. And it just made me think, you know, one of my big takeaways is the individuals, a lot of the AI benefits seem to be accruing to them, their productivity is increasing. Research, I was reading somewhere between 3 and 12%, depending on the discipline. It’s a little bit more uncertain around, is AI affecting businesses?
Marcus Schafer
in general, which is obviously what we care about as investors. So I thought it might be good for us just to continue the conversation and talk a little bit about, hey, is AI actually affecting stock prices and how to think about investing going forwards?
Patrick Collins
Yeah, the time savings piece of it does seem real to me. mean, as a user of AI and somebody who, you know, will use it for certain tasks and things like that, it does feel real. So if that’s true, kind of on a large scale, and I would assume just gets kind of grows over time. Question is, yeah, who does that value that extra time? Who does it accrue to? Does that mean employees are working a little bit less hours?
Does that mean that they can take on more from a capacity standpoint? You know, those are questions to me that still remain and how that impacts stock prices. Obviously, you would think that would translate into higher profitability. If I can get one person to do twice as much work because AI is allowing them to do that much more, then that should theoretically mean I have higher profits. But there’s all sorts of things that go into that. Does that mean I have to pay that person more now because maybe they’re more
They have more expertise, they’re more specialized, they need more training on AI, they have more, you know, all these features that kind of come with it. So I think it remains to be seen to a large degree. Obviously, we’re in the early stages of AI and companies adopting it.
Marcus Schafer
Yeah. And I also, know, so much of investing is, are you meeting, exceeding or missing expectations? Right? We believe all the evidence shows you should act as if the market’s efficient. So the market probably has it as close as anybody else should have it. And then the real question is how hard and how difficult is it to meet those expectations? And what time has kind of shown us history in the past is.
when you get a lot of hype into something, it is just really tough to meet these expectations. Now, if they are met, then it’s amazing, but it really is tough to meet expectations continuously, to exceed expectations.
Marcus Schafer
And investing is all about, are you kind of, as I said, are you beating that consensus expectation? So I think I always start with that perspective that the price is right.
Marcus Schafer
If you have crazy high expectations, to me, that’s a harder hurdle to clear than if you have more realistic expectations. But I kind of have like two questions when it comes to AIs investing is. Number one, is it going to be a winner-take-all market where you have to find the one firm or the two firms or the oligopoly of firms that are really going to dominate the returns?
Or is it more, hey, rising tides lift all boats? So those are kind of the two frameworks I’m thinking about with AI as it comes to investing.
Patrick Collins
Yeah, a couple of things come to mind. First is we talked a little bit about this with Steve in the sense that when I kind of asked him, hey, what are the industries that are really going to get impacted in a negative way? And he kind of had a hard time to articulate that because in his mind, the rising tides lift all boats is definitely the thesis that he would have. But the other thing that comes to mind to me is
what I started, you know, kind of lived through in the early parts of my career was the internet. And I think it was just absolutely transformative to businesses. I know I could say that for, Greenspring, but I know every other person that knows a business could not operate the way they do today without the internet. but there were clear winners and losers in that. mean, there was lots of hype.
there was lots of companies that went out of business that kind of rode that fad. I have no idea if that’s what the AI kind of revolution might look like is just a few winners. But in reality, the internet boosted profit or productivity across every single firm. And so it’s hard to say that there was only a few winners. It was, oh it was only Amazon. It was only Google. There’s there’s countless winners that are in the internet space, just like there’s Nvidia and other companies that are really dominating in the
AI space, but every company got to benefit from the internet. And I think every company will most likely benefit from AI. So back to your earliest point, I wholeheartedly agree that trying to outsmart the market on this to say, will we know something that all the market participants don’t about AI and how it’s going to impact profitability of companies and therefore we are going to buy these companies or not buy these companies.
Patrick Collins
That one’s tough because price does show, give us a lot of data around what we think is going to happen. And when you look at a company like Nvidia, the price is very high compared to its earnings. So the market is expecting very, very robust growth there. And so to think that you are going to outsmart the market by buying Nvidia, it’s kind of already known and reflected in the price to a large degree.
Marcus Schafer
Yeah, yeah. And to think about the internet revolution, the winners take all market. It changed over time, Like Cisco was a winner for a really long time and then it really wasn’t. Same thing about mobile communications, right? Verizon and the wireless operators, the hardware people, people thought those were going to be the winners. And then it turned out to be the Apple and the Google and
different people. So I think that’s one of the real dangers is, if you become too concentrated, these are really catastrophic changes on, you know, volatile on the up and on the down. So I think that’s a word of caution. And then also I just, it’s just like, when you read a biography, you only read a biography of a successful person.
Right? So there’s this survivorship bias in the books that you read and there’s survivorship bias in the companies that we follow. And Steve gave a fantastic example of Amazon who’s done this incredible transformation. They went from selling books on the internet to selling cloud services, to trying to ride the AI wave and the cloud service is where they make most of their money. But if you go back in time and you compare them, okay, what were other internet companies selling?
Pets.com as the extreme example, and I know I’m cherry picking on both ends, they completely went out of business, never got the opportunity to pivot multiple times. And you have to wonder going back in time, could you really tell that, Amazon is going to be the clear winner and Pets.com is going to be the clear loser? I think that’s really, really difficult. But again, if we can get all these little enhancements and some of the transformational
innovation that Steve was talking about inside of all of these other companies, then they’ll become more productive. They’ll become more profitable. And as investors, you don’t even have to take the gamble of trying to find the one. You can just try to be broadly diversified. So that was one of my big takeaways. And I was just thinking about that as he’s talking like, I’m viewing this from a business owner, but how should I view this as an investor?
Patrick Collins
Yeah, to me, it’s just an increasing kind of arc of innovation that happens just with humankind in general. We are looking at ways to make things better all the time. And so if you go back 100 years ago, obviously the way we did business was completely different. The types of companies were different. But we innovated. We created automation. We created automobiles. We created airplanes. We created computers.
The Future of Work and Investment Strategies
Patrick Collins
And this is just the next thing from an innovation standpoint. And could be very transformative, just like there was a lot of transformative technologies that have happened in the past. It helps increase productivity. It is this creative destruction that happens in capitalism. We talked about this study. I have no idea if the data will turn out to be true, but whatever it was, can’t remember exactly. 84 million jobs are going to be destroyed by AI, but 97 million are going to be created.
And that has been the kind of the story of time and capitalism. it’s not a bad, mean, it’s difficult to go through it if you’re one of those jobs that are getting destroyed, but in the arc of making kind of the world better, this is kind of one of those things that will continue to make us better in my belief. So it’s an interesting kind of investment discussion, but ultimately I think we still come back to
Well, what should you do as an investor that understands these AI trends and what’s happening in AI? And I think our thought, or at least our recommendation is not much. You should still own a diversified portfolio of companies. You don’t know exactly who are going to be the biggest winners and losers. We think there’s going to be lots of winners. There probably will be a few losers. It’s really hard to pick them in advance. And even if you could,
and you knew the future, it’s still hard to understand how the market’s gonna react to that. And so in our opinion, continue to own the market, stay diversified, you’re gonna get the benefit of productivity gains when they ultimately happen with AI.
Marcus Schafer
That’s great summary. Thanks everybody who listened. I think this might be one of our longest episodes. So I have noticed, Pat, one last comment, that some of our most watched episodes on YouTube are when my beard is the biggest. So we’ll see if that’s correlation or causation. Don’t invest off of that perspective though.
Patrick Collins
Well,
there was always, there was a period in the first few episodes where your facial hair was changing on an episode by episode basis. So I thought maybe our early listeners were really just tuning in to see what’s Marcus going to look like today. But you’ve been pretty consistent the last few episodes.
Marcus Schafer
Yeah, I think some like 400, 500 people got to see that our most watched was the biggest beard.
I don’t know, could be something to it. Alright, thanks everybody.
Greenspring Advisors is a registered investment adviser with the SEC and only transacts business in states where it is properly registered, or is excluded or exempted from registration requirements. The social media platforms associated with this name are solely for informational purposes and do not offer advisory services or sales of securities. Investing involves risk and possible loss of principal capital. Comments by viewers and/or recognitions are no guarantee of future investment outcomes and do not ensure that a viewer will experience a higher level of performance or results. Public comments posted on this site are not selected, amended, deleted, or sorted in any way. If applicable, certain editing of personal identifiable information and misinformation may be deleted. Content may be dated. Links to third-party sites, or information is from a reasonably reliable source. Information presented on this program is believed to be factual and up-to-date, but we do not guarantee its accuracy, and it should not be regarded as a complete analysis of the subjects discussed. Discussions and answers to questions do not involve the rendering of personalized investment advice, but are limited to the dissemination of general information. A professional advisor should be consulted before implementing any of the options presented.