← Artificial Intelligence Podcast: ChatGPT, Claude, Midjourney and all other AI Tools
Artificial Intelligence Podcast: ChatGPT, Claude, Midjourney and all other AI Tools
AI Can’t Go Solo: Why the Human Touch Still Matters with Robert Brown
00:46:47
Loading summary
Transcript66 lines
- [00:00]
A
You always need to keep a human in the loop when using artificial intelligence. We're going to talk about it today with very special guest Robert Brown. Now welcome to the Artificial Intelligence Podcast where we make AI simple, practical and accessible for small business owners and leaders. Forget the complicated tech talk or expensive consultants. This is where you'll learn how to implement AI strategies that are easy to understand and can make a big impact for your business. The Artificial Intelligence Podcast is brought to you by Fraction aio, the trusted partner for AI Digital Transformation. At Fraction aio, we help small and medium sized businesses boost revenue by eliminating time wasting non revenue generating tasks that frustrate your team. With our custom AI bots, tools and automations, we make it easy to shift your team's focus to the tasks that matter most, driving growth and results. We guide you through a smooth, seamless transition to AI, ensuring you avoid costly mistakes and invest in the tools that truly deliver value. Don't get left behind. Let Fraction AI help you stay ahead in today's AI driven world. Learn more and get started. Fractionaio.com.
- [01:07]
Thank you Robert. I'm so excited to have you here because as we were just talking about before the show, the biggest problem I'm seeing right now is that we're doing everything backwards, which is that we're doing here's a tool, I bought a hammer, now I need to find some nails and we don't. And on a deeper level it's that most people don't have a written down decision making process. We were talking about this with some other people I worked with last week, is that sometimes our clients don't have a template against which they measure. Is this go in the direction of our business or does this not go in the direction of our business? So even when you're deciding this is when you see like mission creep or they have like add ons to the brand and I always say this isn't a logical progression. How did you decide to do from this? I just had a vision. If people think you're a car brand and now you do helicopters, that's a huge jump because they go, wait a.
- [01:53]
B
Minute, they're very different.
- [01:56]
A
So that's what I'd dive into first. As an expert in decision making, this idea of processes, where should someone start when they're thinking, oh, I don't have a process for decision making.
- [02:07]
B
I think the first thing to recognize is that first of all human beings are really bad at making decisions. And when I say bad, this doesn't mean that everything we do is just A disaster. But we are, by adaptation, by evolution, whatever, we have developed certain mental heuristics that allow us to live in, let's say, more primitive environments. And so the environment that we're in today, of course, is not the African veldt that we evolved from.
- [02:40]
So the type of mental processes that we use to evolve in a highly complex, threatening, fast paced world is not quite the same type of world that we live in today. So we need, I think, more structured ways of thinking through decision making situations that led us to short circuit or lead us to short circuit some of those cognitive illusions that we face that arise as a result of that evolutionary process that brought us here today. So I think that's the first thing we need to understand is that human beings actually really do need, let's say handrails or guide rails to get us to the point of making good decisions, which also then should raise the question in our minds what is a good decision? What constitutes a good decision? And that's the other thing that we need to understand first before we even get into process. And that is a good decision is not necessarily one that gives you the outcome you desire. It's one that conforms to a quality of standard or a standard of quality. If I can maybe reverse my language there a standard of quality associated with decision making. And if we can conform to that standard of quality, the closer we do it, and the more frequently we do it, the more often we do make decisions that lead to the outcomes that we want. So it's, I can always tell you beforehand whether or not you made a good decision before you ever experienced the outcome. In fact, I can guide you through the process and I can guarantee you that you will make a good decision, but I can't guarantee that you'll get the outcome you want. And of course the reason for that is that we live in a world of uncertainty and risk. There are many factors that occur outside our control. And so ultimately we make a bet, we make a series of bets, we make multiple bets in a portfolio. And what we're trying to do, of course, is increase the likelihood that we more often than not achieve the outcomes that we want, not that we achieve the outcome that we want on every single decision instance. So I hope I set that up well for you.
- [04:51]
A
Yeah, that's great, because I'm actually thinking about what a lot of us deal with is we have a client or a boss or someone we're working with, who they give us instructions, but the vagueness make them impossible to achieve. So one of the things I deal with a lot is someone will say, I want you to build an AI machine or a bot or automation that does this. And I go, great, can you show me an example output? So I need, and I call it an ideal output. I said, can you show me what would be if the AI gave you this? It would be a perfect result. And they go, no, but I'll know it when I see it. And I go, yeah, that. You're not gonna. Whatever comes out of this is you're not gonna. Because you're asking me to add in guesswork. And every time you add in, and this is the problem, maybe I talked to Dago. Every time you add in an impossible variable or a new variable, the odds of success go down dramatically. So when I have to guess, I'll have to. I said, we'll have to iterate many times because every time I'm building it, I'm guessing at the output. And I deal with this so much where whether it's building an AI that will outline a book. And I'll give you an example of a perfect outline. I don't have one right now, but I'll know when I see it. Anytime I hear, I know when I see it. I go, oh no, but I'm gonna have to raise the price.
- [06:02]
B
It's the classical find me a rock dilemma. Somebody says, go find me a pretty rock. Can you tell me what you think of is pretty? No, but I'll, as you say, I'll know it when I see it. And you're exactly right. It's an impossible task nearly to accomplish. You're lucky if you actually do find the rock that people want. But you're not smart, not you specifically. When people do this kind of thing, when they experience that sort of lucky outcome, it's not because they're smart.
- [06:29]
So you're right. We absolutely need to have a pre structured or a pre framed idea about the outcome that we want. And I think you're right to describe it in terms of can you give me the ideal output? Now let's work backward. What does it take to get to that ideal output? And we can have a process that we can, that can guide us through that to go from ideal output back to inputs and then work our way forward to figure out how to optimize that along the way.
- [06:58]
A
Is there a way to get someone to switch into that mode to realize how important it is? Because the two areas I deal with a lot of vagueness are description of the final result when they want it.
- [07:10]
B
Sure.
- [07:10]
A
So I was just talking to someone recently and I go, do you need this by Monday? And they go, I'm just like doing a lot of things right now. And I'm like, but do you need it by Monday, yes or no? And they like. And it's like, I need to know if I have to pay my team to work through the weekend extra if this is an emergency. And sometimes people have real challenged with it's an emergency or not emergency. And this is something, a lesson I learned from my dad, who was a lawyer for many years. He goes, it's always an emergency until it's on the client's desk. And suddenly it's, you have to work through the weekend and work all night. And then they don't look at it for two weeks.
- [07:43]
B
Sure. So I think the first instance is how do we convince people that they need to change their behavior? And I'll be honest, this is almost impossible unless a person is feeling or aware of the fact that they've been wasting time and resources or that they've had an unnecessarily large number of failures associated with what they've been doing. So unless that sort of exists beforehand, it's a little difficult to convince people you need to change the way you're making decisions. Because most people think of themselves as good decision makers, right? Particularly leaders in an organization. Because they convinced themselves, of course, that they wouldn't be a leader in an organization unless they were good decision makers. The fact of the matter is, if you were to go back and do an audit trail like of all the various decisions they made, it's probably a 50, 50 chance that anything that they decided to do actually worked to what they wanted. But now once you're in that mode of dealing with a person who does want to change, then the way to start is to, I think, frame out the objectives, the preferences and values that person has. And actually there's a really good book about this and there's. It's a little dated now. So when I say dated from my point of view, oh, that was really in my career. But From I think 1996 book by Ralph Keeney called Value Focused Thinking. And I think this actually flips the whole paradigm upside down of the current idea behind data driven decision making. What you have to do is convince a person that they need to be values focused in their decision making. So start off first of all by framing out what it is that is important to you in terms of your values. And when I say values, I don't necessarily mean your morals. And ethics. But what is it that you want to experience? What is it that is valuable to you? And what are your preferences for those values? Once you can get that framed and clearly articulated, then you can begin to start to think about the decisions that you would make to achieve that. And then once you've identified the decisions you can achieve that you can use to achieve those objectives, then you can start to think about the various uncertainties and risks that could prevent you from. From actually achieving your objectives. That gets into a more lower level, quantified level of the decision making. But I say always start with identifying your goals and objectives and values and preferences first and follow that outline that Ralph Keaty gave us back from 1996.
- [10:25]
A
Yeah, I am. It's interesting because when I talk to my kids, especially my younger kids who are like young ones that are under six, I'll say, do you want toy one or toy two? And they'll go, I want both. And that's how kids think. But, like, when I deal with clients and they're gonna go, which one do you want first? I want both. Yeah, but that. It reminds me of high school when people like, I have seven best friends. I go, then that you don't know what best means.
- [10:51]
So something interesting. In my 20s, one of my friends, Ollie, one time said to me, he goes, whenever I'm ask a girl to be a girlfriend, I say to her, what do you think girlfriend means? And I was like, what? That sounds crazy. And then he goes, every girl gives you a different answer. And I realized that it's true. Every person I've dated, I said, what does girlfriend mean to you? And it always, no two women have given me the same answer. I'm very married now, but my wife's definition of married is probably different than every other woman's. And it's really surprising how much words have different meanings. So now I've learned. So when someone says, I want an automation or I want this, I go, what exactly do you mean? And I often go through this, like, Dr. House phase where I'm like you. I know what you said, but I have to figure out what you meant to say and what's the problem we're trying to solve? And sometimes people use words. I think this is actually a big problem with AI is that the word agent means 50 things to 50 different AI people. We don't have any standard definition. So it's really hard. I talked to someone last week who's agentic is different than agent. And I was like, okay, that's really tough. So even to get to that point of what's important to you or what is the most important thing? And this is another area of definition because I've worked with people before who they go, what's the most important metric to you? And I go, dollars.
- [12:08]
B
Exactly.
- [12:09]
A
I don't care about followers, reach, engagements, likes, nobody likes post and it makes a thousand dollars. I like that way more than a thousand likes because I can use those thousand to buy food likes. You can't. And it's always fascinating to me again when you. I have to go through this process of. Let's let me explain what's the most important metric to me. What are the values here? And yes, I was explaining someone the other day, like the thesis of my company, which is AI simplified. I said, if I teach a process and it's more complicated than another process out there, I can't do that. I have a. And it took me a long time to get to that thesis, like a year of thinking about how can I really specify what I do. So when I. When someone is trying to figure out what their values are in this context, I realize we're not talking about moral values. We're talking about what's the most important metrics for your company or what are you trying to achieve? I think this is a really important question because this is when people go into mission creep. Because then you go, does this decision go match the core value of my company? And this is like when a company has a mission statement. I think a lot of companies miss the point of a mission statement is to say no. It's not to tell everyone how good you are. It's to create like tool to measure your other decisions against. To go, does this fit inside that?
- [13:29]
B
It's your heading. It's the heading that your company has the directional heading. What you're describing, you're describing this from the point of view of talking to particularly in a single individual and then running into the ambiguity that they give you with the words they use. But imagine now that you're in a. It shouldn't be hard, right? You do this all the time. But working in a large organization where you have multiple stakeholders that presumably are all working for the same sort of outcome. And yet if you were to go and ask them, tell me what your company's goals or objectives are or what is, what is it that your company does, they will all give you a different answer. And then you wonder, how in the world does any company make money at all? I'm actually pretty convinced that companies after they get to a certain phase, make money in spite of themselves. And this is not really to denigrate anybody, but. But the fact of the matter is people set up systems and then they get stuck in various silos and within their organization. For good reasons, by the way. There's some really good reasons why companies develop silos. It's because there's a common language, there are a common set of practices and things like that that allow that particular department or organization to accomplish something with more efficiently if they were having to constantly explain themselves. Right. But at the same time, when you get so siloed that the silos aren't talking to each other anymore, then you've got a real pathological problem. But this happens all the time in organizations. You ask one particular operating unit within a company, what's the most important objective about this decision? Let's say you've got an overarching decision to be made. They of course are going to give you the metric that they are measured on. They are not going to give you the metric that the overall company. And I'm saying as if they do this all the time with 100% certainty. But the propensity is much more in line with the idea that they focus on the thing that they are responsible for as opposed to thinking about what the overall organization is actually trying to achieve. And so this whole process, you mentioned how important it is, and I really cannot overemphasize it is so important when you embark on something like implementing a large artificial intelligence system or a significant one, let's say maybe the word large isn't the right word, but a significant system that involves the input or the use of multiple stakeholders. Getting that framing right, what are my values and objectives? What are the thing that I'm trying to achieve and why am I trying to achieve it? That is absolutely table stakes. Most important, if you don't do this, you will literally fail.
- [16:17]
A
It's a really good point because I always think about how like marketing hates sales and sales hates marketing and marketing like all these. I was talking to some people I worked with the other day and I was like, we should just fire all the other department to hire more engineers. And they're like, that's exactly what every engineer thinks. We just need keep making the product better and that's the only thing that matters. And they're like, but we have to generate sales and customers and the other things. And I'm like, yeah, but if the product's better, it's like, I have that field of dreams thought it's if the product's really good, everyone will find it. Yep. Every engineer will come. Yeah. So talk about large scale decision making or using large amounts of data because I find that the more people whose input you get, the murkier the data gets. And I think about you have this idea that if we get everyone's opinion, we'll get a good answer. But then if you ever watch Family Feud, there's always the numbers at the bottom that are wild. Three people said this and two people said that those are not useful pieces of information. You never get. 80 people said one thing like it's not 80 people agreed on the number one answer. There's this idea now and I think this is really popular because of AI, because AI is good at analyzing large amounts of data.
- [17:28]
B
Sure.
- [17:28]
A
That we have more data, we'll get better answers.
- [17:33]
B
I not entirely convinced of that yet. Particularly with the use of artificial intelligence for certain kinds of things, and particularly, let's say, strategic setting, strategic intent. I think, and don't misunderstand me, by the way, I think artifact, the advent of what we're seeing today with artificial intelligence is huge. It's going to, obviously it's already changing things immensely. It's going to continue to change things immensely. And I think on net for the good. Right. I think we're all going to benefit from the trajectory that we're on with this.
- [18:11]
The problem is that in the short term, the failure rate for these artificial intelligence initiatives is extraordinarily high. And I think that's related to the discussion we've been having through all this. In fact, it really reminds me if you go back, maybe starting 10 years ago, when the whole data science movement really began writ large. Right. Everything needs to be data driven decision making.
- [18:38]
In 2017, actually, maybe going back a year before that, 2016 or so, Gartner reported that something like 65% of all data analytics initiatives failed. And you think, oh my gosh, that's a huge failure rate. And particularly when we're talking about the field of inquiry. Right. That is supposed to make us better decision making makers. Right. So why does the physician not heal itself? Why is it that data driven initiatives can't be more successful than all other types of initiatives? Because in fact they were at that time failing at about the same rate as just about all other IT initiatives. And then a year later, Gartner came back and said, oops, we were wrong. The data failure rate wasn't really about 65%. It's more like 85%. It was even Worse. And if you track back the causal reasons for that failure rate, it all starts with what we are exactly talking about on this whole thread. And that is it's a failure of the executive function to clearly identify the right problem to be solved. In other words, these, all of these large data driven initiatives were really science fair projects. Large, complex science fair projects that didn't really have a problem to solve. They had, they came up with a solution, then they went looking for a problem. The same thing is happening right now with artificial intelligence initiatives. In fact, if you go back last year, late last year, Rand published a paper that showed that something like 80 to 85% of all of artificial intelligence initiatives were failing. And then when you look at the postmortem explanations for why they failed, it was the exact same reasons, as we saw with data analytics, it was that failure of the executive function to clearly identify the right problem to be solved with an understanding of why you wanted to solve that problem. Again, in other words, people are solving irrelevant problems. And there's really, in my mind, no bigger waste of resource than to solve an irrelevant problem.
- [20:52]
This, I think, has got to be something we fix in the industry, by the way.
- [20:55]
A
Yeah. I feel like so many companies are just doing something that's cool.
- [21:00]
B
Yeah, exactly.
- [21:01]
A
And clever. But it doesn't solve a problem. I always think about the best inventions are you solve a problem you have and then you find out other people have the same problem.
- [21:11]
B
Exactly.
- [21:12]
A
Then you're on to. But what we see is people go, there's also this thing. You ever heard, oh, they don't know what they want and I'll show them. And it's like, okay, that does work very rarely. And that's the problem is it doesn't work never. It just almost never. And you're like, one. It means there's a. There's more than a million people. It means it's going to work for me. And it's. I think this is a really important lesson. It gets back to first principles because people, when I teach people entrepreneurship or starting a business, they say that you want to find the three Ps, which is, are there enough people who have the problem you solve and they're willing to pay for it? Those are the three elements. Now, not a lot of people, but they'll pay a ton of money. Like surgery, you don't need that many customers. If it's something they'll pay very little amount of money for, like a magazine, then you need a lot of customers.
- [21:56]
B
That's right.
- [21:57]
A
And that kind of determines things and, like, how big the problem is for them means how likely they are pay for it. And I see that we're bypassing that so much right now, both in people investing in these AI ventures. And they reach out to me to solve problems or to help them to use a really cool tool. And I can always tell, because I really am aware of the news, which article someone read based on the question they asked me. They're suddenly really interested in something, and I go, okay, let's talk about the good idea fairy. Like, great, when they visit, but let's make sure we have a reason for this. So I go through this really complicated process when I'm making a purchasing decision. And I teach AI, right? I teach AI. I publish an article, everything on LinkedIn. I. And yet I only use seven tools, and I would only use two if it wasn't my job. I just have to do a couple of sometimes comparative tests.
- [22:49]
That's why when I see people that they're like, I use 50 AI tools. And I go, no, you're not good. There's nobody who's. I do drywall and tile and I paint and I do electric and plumbing. Yeah, you're probably bad at all of them. And you're not a master. That's the difference is that it's the person who narrows their focus who's really good. And I think that this is exactly what's happening. Just that.
- [23:11]
Yeah. And that's fine. But it's. Then say you're a handyman. But there's this thing that's happening, and I think you're exactly right, which is that we think more information is always better. And as I was saying earlier, like, I think I think of this all the time, which is that now we transcribe every meaning and we store it on cloud servers and think about how many servers there are around the world, hundreds of terabytes of transcripts that nobody will ever use. Listen to. And just. It's like hoarding. Like, now we're doing digital hoarding and all of this data's out there. And it's just like, we don't. We think because it's in the cloud. Like, the cloud's not a real place. It's. No, the cloud's a computer just at someone else's house. It's still using electricity, taking it made out of compute parts. And we're. More data. And I tried to say this, like, more data that you're not using doesn't help. And if more of the wrong data doesn't help and accelerating broken process just means you're crash faster, take it away.
- [24:05]
B
Yeah, and I think that's really good distinction is really to make good decisions, whether it's developing a new gas reactor, if you will, versus implementing some significant AI solution in your organization. It doesn't matter if you're using irrelevant information or if you're, you're trying to mine just irrelevant information to help support your decision making. You'll probably find something to justify the decisions you've already made. But rather than starting from the top like we've been describing, identifying your goals and objectives, then creating the decision strategies that could help you get there, and you need multiple, consider multiples. Then you find the data or the information.
- [24:53]
That would help you to make distinctions between those strategies of moving forward. Right. You could have one artificial intelligence initiative idea in mind in terms of what you want to achieve, but you probably can come up with three or four different ways to get there. They all would have their, their benefits and their costs and risks associated with them. And you need to have the kind of information that gives you the relevant means to make distinctions between those pathways. So that's the really important thing of distilling the information, the data you have, down to what's relevant and not just using all the possible data, all the data that's out there. We've run into a, I think a significant problem with this issue in that before the age of being able to house and warehouse all this data, of course we said humans need data to make better decisions. We can't necessarily rely just on our intuition, which by the way, is true.
- [25:56]
We went the other direction. We went the other extreme and now we have so much data that we now have a duplicate of the problem we had before. Right. That is, the world was large and complex and somewhat scary because we didn't have data that we could access in an easily accessible way. But now we've got so much data that we don't know what to do with it. And it's just a duplicate of all the information that exists that surrounds us in the universe, naturally. So which data set do we refer to? That's going to be the big problem we face. And fortunately we are getting better at collecting information. But the problem, of course, as we've just been discussing, is that we now have a lot of irrelevant data we have to parse through first before we can find what's relevant. I think that's one really big problem. The other really big problem with this is that all of the data that we have of course, is a reflection or an imprint rather of what the past looks like. It doesn't tell us anything really about the future except for short term trajectories. It's very difficult to actually be successful at strategic thinking if you're using data that's already representative of the past or only representative of the past. Right. The real concept behind strategic thinking is that you do something different, which means that you're generating a counterfactual to something that has never existed before in the world.
- [27:25]
If that makes sense. So you're actually thinking about a future that has never existed, which means that you need information about that future that you won't have. This is the conundrum of strategic thinking, right? So it forces us, instead of relying on data, we have to rely on a different kind of information. And that information actually is probabilistic reasoning. We have to use probabilities and our willingness to make bets and our willingness to actually assign probabilities to various outcomes. Even if we don't necessarily have data to support us, we can at least quantify our propensity to believe certain things and then we can make trade offs based on those probabilities that we assign. But this again brings us back down to that level of looking for the relevant information to support the decision making. We need to get to the outcomes that we're seeking. The only way to do this, by the way, is to develop a structured business case analysis of what you're trying to achieve so that you can make these sort of informed trade offs. And that I think that's actually one of the biggest problems I'm seeing myself with people that are going down this pathway of developing large, significant artificial intelligence solutions to problems. They aren't necessarily stopping to ask themselves what is the business case rationale that supports this decision making?
- [28:49]
A
It reminds me of that saying often people are fighting the previous war instead of the current war. Like you look at the data for the last one and it's no, this one's different. And it's the same thing that one of the things I find interesting is people will often say, oh, I admire this business version of that billionaire. And I say, their wife left them and their kids changed their names, their family hates them. Do you want that out? Like you have to look at the totality of someone's result and we just look, I want this family from this person and that business from that person, that income. And it's like you can't pick and match wherever you model, you're going to get that result. And like maybe you can set the new world record for the biggest divorce settlement. That can be your dream.
- [29:28]
B
Fine.
- [29:28]
A
But it's we. I see that so much where. And it's. That's obviously a more personal decision. But even at larger levels, we model decisions or we look at things that people did in the past, or we're looking at the wrong data, we don't look at enough of it and go, everyone used to love that book good to great, right? About those 10 companies. And then I think eight of them are gone. I don't know if that book is the reason, but it's like this is the example of resilient companies. And it is. I think that's one of the things that's really interesting is that large companies are slower to make decisions. And it's really challenging because you get locked a certain way of thinking. And if you model that, you lose your advantage as a smaller company, which is agility. But I think that it is this temptation of going, have a bunch of data tell me this is a good decision. And I often look, have you qual. What's the quality of the data? Have. One thing I find really interesting, people make a decision based on like online opinions. And I go, whose opinion did you ask? Did you ask your customers or just everyone? Not exactly. Very different. Existing customers. When I get feedback from an existing customer, I pay a lot more attention than when it's just a blind email from someone who's never bought something from me.
- [30:37]
B
Right.
- [30:38]
A
I wait them differently, but I see a lot of companies that weight them the same. And I don't care if you were never. A lot of people will never buy any of my products, read any of my books. Most people won't. The majority of the world isn't interested in what I do, which is fine. So I need to narrow it down, just existing customers or potential customers and then make the decision about what they want. Because that's another way you. And there's just one of many ways that data can be thrown off. But I want to dive a little bit more into this kind of predictive process. Can you explain that? How you can develop a system that stops relying so much on the past, but starts to factor what are the most likely outcomes or what's the likely future.
- [31:18]
B
So this really gets us a little bit into the weeds, if you will, of the concept behind decision quality. But I think it's a useful pathway for discussion here because I do think that people think, particularly when they have to implement a decision process like we were discussing earlier on in our conversation that this does lead to a reduction in agility. And by the way, I've seen the exact opposite. When you have a structured way of moving and thinking with each other, it actually increases your agility. So oddly enough, the effect is that at least in shorter periods of time, you're slowing down. So there's like this stoic saying, we go fast by slowing down. What happens is that by having a structured decision making process, you're actually taking the time to think through what can happen before it happens. And that way you can develop mitigation plans or the sort of the converse of that maybe instead of thinking in terms of the things that can happen to you that you don't want to happen, you think about the things that you want to happen and then you develop a quote mitigation plan to help promote those. Right? So first thing is understand that a refined or a defined decision making process actually helps to accelerate your decision making when it's structured the right way. That's there. There is a bit of a caveat to it. Now to get to the predictive part that you were just asking me about, the way to start this is actually to map things out. I like to do this visually. I'm a very visual thinker and I find a tool like an influence diagram is a very powerful way to think through the sort of, the predictive, the structure of any sort of models we might need to help us think through the decision making. But if we start with what is that objective that we originally identified that we want to achieve? Then we disaggregate or decompose that objective outcome into the forces that led us there or the events that would lead you there. And so actually, to make this sort of simple to understand, think of it like any generic business. What is it that you want to achieve? From the perspective of most businesses, at the end of the day, it's that maximization of corporate value, right? It's money.
- [33:47]
How the measure for that over time with most organizations is net present value of cash flow, right? So that's something we can model. Net present value of cash flow would be our objective in our influence diagram that we end with. Then we might break this down. What causes net present value of cash flow? It would be revenue minus cost. So now we have two decompositions of net present value, revenues and costs. And then we can decompose of what's the source of revenue. And we can keep decomposing each little node in our influence diagram until we get back to a certain level of there are Inputs now that we have to think of as, let's say, assumptions, right? And those assumptions are the sources of, or would be the placeholders for the data that we need to get to help support that decision making process. What we want to do, of course, is to align that overall influence diagram with each of the various decision pathways that we could take to achieve that goal. We want. Then we change the assumptions for each of the pathways and then test the effect of those assumptions. Right. Now a key part of this though is that the assumptions cannot just be single point values. Because we don't deal in a world that basically the future always delivers us a certain outcome. We have to get really comfortable with thinking in uncertainties. That means we have to think of these assumptions as distributions. We have to, and we have to get comfortable with representing our assumptions as probability distributions. And there's some easy ways to do this, right? We don't have to have a huge amount of data to support building a distribution. We can actually go to a subject matter expert and ask them to give us their belief about how something might look in terms of its overall probability distribution given the, the underlying qualifiers that describe a given pathway for creating value like a decision pathway. I've been doing this for about a little over 25 years now and it's astounding, honestly, just how replicable this is. Once you are working with a subject matter expert who can think clearly about the problem that has been given to them to think through, they're very good at finding 80th percentile ranges that are fairly accurate about the world that they'd be facing. But it's really important to get this right, to support this predictive making capability that you're just asking about. You have to ask the subject matter experts to describe the reasons why a particular assumption can vary, why it has range to it. You can't ask them what the outcome will be. This is actually, by the way, this is a really important understanding about the quality of a subject matter expert that you might go to for information. Subject matter experts are actually constrained by the same types of biases that all human beings are. In fact, they're pretty bad at making predictions about the future in an unaided way. But what subject matter experts are really good at, this is what makes them an expert, is that they can give you fine grained explanations for why a system varies, like what causes a system to behave the way that it does. That's why they're experts. They understand those really fine grained details, the possibilities that can happen that you know an amateur or maybe somebody that's not quite so sophisticated in their understanding of a specific type of event in the world, they can't give you very detailed explanations for why that event might. Might vary. Subject matter experts, on the other hand, are very good at giving you detailed explanations. Now, once they're able to think through those detailed explanations for why a system can vary, then they can give you their probabilities for the ranges that you might see and sort of the central tendency that you might see across that assumption as represented as a distribution. Then you have to use something like Monte Carlo simulation to then tie those assumptions with their ranges back to an outcome that you're predicting. And this takes a little bit of sophistication, I admit, but there are a lot of tools on the market that really help people that are novices in this area at least pull their bootstraps up a little bit by themselves to get started. You don't have to have a massive data science force to help you through this. There are a lot of resources that are available.
- [38:42]
A
Yeah, I think that it's just so interesting because now we so often shoot from the hip.
- [38:48]
B
Yeah.
- [38:48]
A
I get the most decision. When I was a kid, 20, 30 years ago, if you want to start a business, you had to have a business plan, go to the bank, get a loan, start the business. And now you don't need a business plan, so no one makes one. So no one has a plan. And it's this.
- [39:02]
Kind of decision make. I'm not always perfect with this, but a friend of mine the other week said, oh, I want to start a side business. And I go, how much money do you need to make every month? And he goes, what?
- [39:10]
B
Yeah.
- [39:11]
A
And I was like, isn't that. Let's start from there and work backwards, because that number will determine what you need. And the factors I always look at, and I try to explain that, I look at how much money to make and how much risk are you willing to put into it.
- [39:25]
B
Exactly.
- [39:26]
A
And risk management is something that people really struggle with. And I try to explain as there's two ways you can pay me. You can pay me a flat fee up front or commission on the back end. Flat fee up front. Like more risk for you and less risk for me. If I write a book, I only make money every time the book sells. If you pay me to write your book, you pay me up front. I make money no matter what happens. So it's less money. Royalties be more money over time. But if you never release the book, which many of my Clients don't do or you never finish the project or you mess up. That's. And I try to explain it's risk management. So I try to always have fast money and slow money projects and risk management. That's another element of it. How much money to make then how much are you willing to. If you need to make it fast, you need to go high risk. Right. If you have one day it's very risky. You have to do something really extreme versus if you have six months or two years. So it starts to bring in these other factors. And I think it's really important because we often hear everyone's a self declared expert and I have to deal with this all the time. Me, people overestimate my level of expertise and I try to explain, listen, here's what I can do. Everything outside of that, as we get further away, I get more and more terrible.
- [40:44]
B
Yep.
- [40:44]
A
I'm very good in a very narrow thing with AI and it's all. People are always surprised when I don't claim I know how to do everything. And I say listen, I don't want to make a promise and I can't keep sure. I go down the path and I'm not very good at a lot of things. Most things today I'm not good at. I'm good at a couple of really good at two or three things. But there's thousands of things that people can do and I think that is where you start to know. A real expert knows the limitation of their knowledge. So I always look for people like what they go outside what they're good at because we see so many experts on TV and then their predictions are so wrong. I look at that all.
- [41:20]
B
Yeah, you're right. And that kind of goes back to the point that I was making earlier is that actually experts are terrible at making predictions. They're terrible at making predictions in an unaided way. They're shooting from the hip. And let's be honest, it seems that thinking now has become declassee.
- [41:35]
A
Right.
- [41:36]
B
Nobody is doing this anymore. Everything is driven by our intuition it seems. Or else we're now become self proclaimed experts because we're relying on artificial intelligence in the background and not saying it upfront and explicitly. But yeah, I, I agree. Experts are actually they're subject to the same sort of biases and mental failures that all of us are subject to. But they serve a, they do serve a very good purpose when they're utilized in the right way. And that's something we have to under. We have to figure out how to Understand better. How do I use an expert the right way? Don't ask for predictions, ask for explanations.
- [42:18]
That's the way to get around it.
- [42:21]
A
I think that's really good. I think this is going to help a lot of people who have been caught. I often talk about Double Dutch with making a decision like you're trying to want to jump in at the right moment. The two jump ropes are spinning. I don't want to make the wrong decision. So you wait longer and longer or you jump in too fast and get all tangled. And I think this will help a lot of people because I think my feeling is that decision making before you implement is like where people are really struggling now. And for people who want to learn more about what you're working on and possibly get some of your help to help them to make better decisions and do that free AI decision and build a strategy. Where can they find you online and find out more about the amazing things that you're doing?
- [42:57]
B
Sure. The best place probably to find me online is LinkedIn. Robert D. Brown, the third. I know that sounds a little highfalutin, but the name Robert Brown is actually such a very common name. I have to have a way to make a distinction so that people can find me.
- [43:14]
A
Oh, I know.
- [43:16]
Yeah, deal with it.
- [43:17]
B
It's like, yeah, those of us with colors for last names.
- [43:22]
And names like Robert and Jonathan, we ran into this problem. But. But yeah, LinkedIn is the best way to find me, I think. And I maintain a sort of a, an open network attitude. If a person is interested in having a truly professional, collegial sort of relationship with me on LinkedIn, I answer emails that I get there. Uh, I'm not really all that interested in people just jumping right in my inbox and trying to sell me something. Probably one out of a thousand times has that actually ever turned into something where I go, oh, I'm actually glad you did that? Maybe even less than one in a thousand. But yeah, that's a good place to reach me. And also I'd give my. I'll give my email address robbrown at Resilience, or I'm sorry, I get these. My email addresses here. Confused. Robbrowniberresilience.com but currently I work for an insurance company called Resilience. And, and I'm actually working within this company as an internal consultant, if you will, as the, the senior director of Cyber Resilience to support making these kinds of decisions that we've been discussing. Not necessarily AI decisions per se, but decisions related to new product Development, if you will, new supporting, strategic planning, those kinds of things. But they're all related. Just because artificial intelligence is certainly a new technology that's available for us and like the Internet was and many other really cool, powerful things that have been developed, in the end, really, the decision making around it is the same again. It goes back to identifying what we want to achieve and why. And then what can I do to get there? And then finally taking into account what are the risks and uncertainties that would prevent me from doing that? Then how can I structure my risk management efforts to maximize the livelihood that I get, the outcome that I want? But also keeping in mind, just because you do make that good decision, that is, you've taken into account that information the way we've described it, in that hierarchical way that doesn't guarantee that you get the outcome that you want. It just increases the likelihood that over repeated trials at making decisions that you do increase the likelihood of getting what you want. That's a really important thing to understand. It's. It gets you out of what we call resulting, that is looking at the final outcome to determine whether or not you made a good decision, to initiate the decision, the process that you're on.
- [45:56]
A
I think that's gonna help a lot of people make a little better decisions and hopefully some people who are interested in what you do reach out and only people that are interested in what you do. I get a lot. Far too many sales messages in my LinkedIn inbox too. But thank you so much for being here for an amazing episode of the Artificial Intelligence Podcast. Thank you for listening to this week's episode of the Artificial Intelligence Podcast. Make sure to subscribe so you never miss another episode. We'll be back next Monday with more tips, tips and strategies on how to leverage AI to grow your business and achieve better results. In the meantime, if you're curious about how AI can boost your business's revenue, head over to artificialintelligencepod.com calculator. Use our AI revenue calculator to discover the potential impact AI can have on your bottom line. It's quick, easy, and might just change the way you think about your business. While you're there, catch up on past episodes, leave a review and check out our socials.