At the last few marketing and customer experience conferences I attended I wandered through the trade show displays and noticed that almost every company claimed to be using artificial intelligence (AI). Yet when I asked them to explain how they were using it, most were unable to answer. They’d talk some gobbledygook, and/or what they’d describe it as doing wasn’t actually AI at all.
So I was feeling a bit disillusioned about the whole AI thing when I picked up a copy of Jim Sterne’s recent book, Artificial Intelligence for Marketing: Practical Applications.
Read this book! My copy is covered in sticky notes. It will give you a much better understanding than most marketers, customer experience officers, or executives have about how AI can help you now and in the future.
AI is not a magic bullet for marketers, but it will revolutionize every industry over the next few years. It may even have more impact on how you do business and how customers interact with businesses than the Internet did.
In today’s Frank Reactions Podcast on Customer Experience, Sterne makes several important points.
Perhaps most important is this: your organization’s experiences with AI in the near-term will be like teaching a baby to walk. At first you are all excited, because you see the huge potential. That soon gives way to frustration (for you and the baby) as the baby keeps falling. But, like that baby, if you don’t give up, it will learn to take one step, then several, and before long, as Sterne put it, you’ll be chasing after it to stop it from running into the street!
Unfortunately we are still at the magical thinking stage: because most marketers and executives don’t understand AI, they believe it can do anything.
And, according to Sterne, it will always require human intervention.
“A human has to decide what question are we trying to answer, what problem are we trying to solve, which data should the machine consider in order to come up with a recommendation, is the outcome logical, reasonable, does it pass the smell test.” – Jim Sterne
The “smell test” is particularly important. All AI starts with algorithms and a data set fed into them. The results of that combination create a model of how the world works, but no model is 100% accurate. What makes AI exciting is that as more data is fed in the model can continually be tweaked automatically so it gets better and better at predicting outcomes. But the predictions can lead you astray. They may result from hidden biases, or reach conclusions that are unethical, for example.
What's The Role for Humans in this AI Process?
Let’s break down his thoughts on where humans are needed.
- Humans need to decide what question the AI should be used to answer. You can’t just throw a whole pile of data at a machine and say “solve all my problems.” Even saying something as seemingly specific as “optimize our profitability” is actually too vague.
- What data should the machine consider in coming up with a recommendation? Some structure is needed. If you want to find out what are the characteristics of your “best customers” you need to have a definition of “best.” Are you talking about lifetime purchases? NPS scores? The purchases they’ve made in the past 12 months? The relevant data will vary from company to company.
- What biases are built in? Whenever you are using past data to predict the future, bias will creep in. As I’ve noted before, looking at historical data might conclude that companies should only promote 35 – 50 year old white men into senior executive ranks. But that is because historically very few women or non-white men in many countries have been in those roles. There is now plenty of research to suggest that limiting your pool to that one demographic will ultimately harm your chances of success, not help them. You need humans to think through what unintentional biases that algorithm might be basing its recommendations on.
- Does it pass the “smell test”? Bias is one aspect, but there are other ways in which AI systems could lead you astray. For example, the data might suggest that making your staff work 24/7, 365 days a year would optimize profitability. But we humans know that that won’t work. Humans aren’t robots. And even robots need downtime for repairs and maintenance.
How Can AI Help Marketers Most Today?
In the short-run, the most exciting potential of AI for marketers is the ability to do massively complex optimization of marketing messages. It makes it possible to test so many more variables, and so much more quickly, than was ever possible before.
Of course, for that benefit, you need a lot of people being exposed to your messages. Otherwise you won’t be able to get statistically significant findings before the world in which you are doing the testing has changed. Peoples’ attitudes can change a lot in a few months.
And even when the model tells you that purple people eaters are your most profitable customers, you may conclude that you’d rather not sell to purple people eaters. There are other human factors that enter into decisions. Despite what they taught us in business school, profit maximization isn’t the only thing that matters. Or, at least, it shouldn’t be.
Episode 127 – Artificial Intelligence (AI): Hype or Help for Marketers?
Posted on 07/20/2018
JIM: If you just trust anything that the machine says, you will fall down a rabbit hole, you’ll never come back, and you’ll hate it. It won’t work.
[Transcription starts at 0:00:43]
TEMA: Today’s guest is Jim Sterne who, like me, has been exploring what works and what doesn’t in digital marketing since the earliest days of the Web. Now, I’ve got to warn you. This episode gets a little geeky at times. I mean, I am talking with the guy who founded the Web Analytics Association back in the 1990s, but it is so worth your time.
0:01:06 We explore the hype versus the reality of artificial intelligence for marketers and there are some super important lessons in there if you want to make AI pay off for your organization and avoid the blackholes that it can suck you into. Sterne believes that AI, today, is where the Internet was in the 1990s, massively overhyped but it will, in the next couple of decades, change the world as we know it. He thinks it could even be bigger than the Internet was in changing things. Marketers, customer experience practitioners, and senior executives all need to understand it better and this episode gives you a crash course on some of the things you need to know about AI now.
TEMA: Now, it’s time to listen to my interview with Jim Sterne.
[Interview starts at 0:02:37]
JIM: My name is Jim Sterne. I am the founder of the Marketing Evolution Experience Conference, co-founder of the Digital Analytics Association, and author of a dozen books on online marketing, analytics and, now, artificial intelligence for marketing.
TEMA: I have read a lot. I’ve been following you for many years, since the original digital analytics days. I was really fascinated by the fact that, in your new book, Artificial Intelligence for Marketing, you’re getting realistic because one of the things I’m finding very frustrating is that an awful lot of people seem to be talking about, you know, they’re all claiming they’ve got artificial intelligence in their products and stuff.
TEMA: A lot of the time I think they don’t really.
TEMA: Let’s just talk about the hype versus the reality of artificial intelligence. When it comes to marketing, how far along that path do you think we really are? Do you think most of what we are hearing is unrealistic or overblown?
JIM: Curiously, no. This is a different kind of software. I think that’s the clearest way to think about it. It’s just a new type of programming.
JIM: It’s not magic. It’s not scary. It’s not undecipherable. It’s just different and much more powerful. As such, it has the possibility of changing everything that a computer does in some pretty fundamental ways.
0:04:12 The hype about it is coming from the fact that, since we don’t understand it, well, it could do everything and anything. Well, that’s not true.
JIM: But, you’ll remember back when the Internet first came out. I started writing books about online marketing in 1994. Then it was just so much hype. It was the cover of Time Magazine. It was everywhere.
TEMA: Yep. I put up my first website in ’95.
TEMA: Yeah, so I remember those days.
JIM: It was pretty unbelievable. The hype was outrageous. Yet, it turns out they were not wrong. It has upended finance, medicine, commerce, marketing, and-and-and. I think artificial intelligence, in the broadest definition, has the opportunity to have as much or even more change than we’ve seen with the Internet.
TEMA: Okay. I think you’re probably right about that. I guess it becomes a question then of figuring out where it is currently effective. Can you give me some examples of places where it’s actually working and working well?
JIM: First of all, let’s provide a definition. Artificial intelligence is an umbrella term that includes things like computer vision, natural language processing, robots and self-driving cars, machine learning, and all of that is the stuff of science fiction to start with. It turns out that they are all very difficult problems, but we finally have: one, enough data; two, enough computer power; and, three, enough experience to test out the theories that we’re actually getting a handle on it in improving our capabilities dramatically day-to-day.
0:06:08 Now, you can actually have a conversation with somebody in a foreign language on your phone, and it will translate for you in real time. That’s science fiction in my world, but it works.
JIM: The ability to look at a photograph and say, “That is a cat,” or, “a dog,” or to be able for a car to drive down the street itself, these things are working.
JIM: Now, for marketing, I’m interested in machine learning, above all, because that’s the tool that marketers will use, and it is incredibly valuable at segmentation, classification, clustering. Here are all of my customers. What do they have in common? Or, what do my best customers have in common?
JIM: It’s up to me to describe what is a best customer and then, if these are your highest lifetime value or whatever that might be, great. Can I go out into the world and find other people like that who will be my prospects? This is a powerful tool.
TEMA: Yeah, absolutely. Now, you made the comment that it all begins with models and all models are wrong.
TEMA: Can you elaborate on that?
JIM: Certainly. The statistician George Box, that’s his quote.
JIM: It means that a model is just that. It’s not the real thing. The map is not the territory and it is an abstraction of the territory to help you think of things conceptually. A model is a mathematical construct that is intended to draw a picture of the reality.
0:07:54 I have a model that says, “If I increase my marketing budget by 10%, I should be able to increase sales by 15%.” Well, that’s a model. It might be true. It might not. Chances are, absolutely 100% certain that it is not 100% true. Therefore, all models are wrong.
JIM: But, if I look at a year’s worth of data, I take ten months and I build a mathematical model that describes those ten months exactly, and then I ask that model to predict November and December and then compare it against the real November and December, I can judge whether or not that model is useful. If it is, ooh, now I can use it to predict January, February, and March. If it doesn’t predict November and December, I’ll go back and tweak it a little bit until it does. Now, it’s even more useful. It’s useful for a certain period of time and then life changes; the economy changes; the competition changes.
JIM: It’s payday, or it’s not payday; it’s seasonal. All models are wrong, some are useful, but all have a limited time value.
JIM: This is where machine learning is so much more powerful because it looks at the data, it creates its own model, and it changes its mind, over time, as new data comes in.
TEMA: Right, so you’ve still set the parameters in terms of what your desired outcomes are.
TEMA: Okay, but it’s adjusting based on experience.
TEMA: To make it more and more likely to predict–
TEMA: –when you’re going to get the desired outcomes.
JIM: Mm-hmm. The human has several responsibilities, and this will never change. The human has to decide what question are we trying to answer, what problem are we trying to solve, which data should the machine consider in order to come up with a recommendation and, then, is the outcome logical, reasonable, does it pass the smell test?
JIM: “Which data do you feel it?” completely changes the output.
TEMA: Oh, yeah.
JIM: Let’s say I’m using salesforce.com and you are using salesforce.com. We have exactly the same data structure, but we have different data.
JIM: If we take the same algorithm, and algorithms are many, varied, and new ones are popping up all the time–and they are open source, so knock yourself out–if we take exactly the same algorithm and exactly the same data structure with different data, we’re going to end up with different models that will be useful to us individually, but we can’t share them.
TEMA: That’s interesting. I mean that’s a really good point. Does that imply then that there’s no practical or useful sort of industry benchmarking capabilities?
JIM: Well, you can benchmark how much better you’re doing today than yesterday.
TEMA: Right, but not how much better you are than your competitor.
JIM: From an outcome perspective, yes. We spent the same amount of money on outdoor billboards, direct mail, or social media advertising, and we got different results. That’s benchmarkable.
TEMA: Okay. Yeah, that makes sense. Some of the places where you suggested that there is real potential for using this type of thing, some of the ones that caught my eye was its use for testing of marketing creative.
TEMA: Can you talk a little bit about that and how that’s useful there?
JIM: So, this is where large companies have a leg up because they have more data.
JIM: I’m a small company. You’re a large company. We can both take, say, 100 different headlines to promote our product, whether it’s a subject line in an email, it’s a tweet, or it’s the headline of an article for inbound marketing. Take 100 of those and put them in a database. Take 1,000 pictures of our product, our customers, happy faces from stock images, whatever, throw that in the database, and hand it to the machine.
0:12:20 You ask the machine to put together the outbound marketing. Let’s just say tweets for now. It’s going to tweet a variety, a random combination of these headlines and pictures until it’s got enough results to say, “Oh, well, this combination works best.”
JIM: You will look at this company. You have more people you can show that to to get results from.
JIM: I have 10,000 people in my database to show. It’s going to take me a while to get statistical significance in results. You have 10 million in people in your database. Oh, you can do it in a couple of days.
JIM: The machine, just by literally trial and error, will figure out, will learn which combination of headline and picture work best for what cluster of customers.
TEMA: Hmm. You’d still, though, need a fairly large database then of customers to do that kind of testing. But, are you saying it lowers the amount that you need, the number that you need?
JIM: No, the more the better.
JIM: It is always the case and, with machine learning, even more so. But, even if I only have 10,000 customers, I can have the machine cluster them and say, “Well, you’ve got 20 different unique groups here.” Then I say, “Great. Go out and take the group that has the highest customer lifetime value.”
JIM: “Grab their cookies. Look out on the advertising networks where those cookies have shown up so that you can create an even bigger picture, a bigger persona, if you will, of that cluster, and then go find other cookies that exhibit the same behavior.” Ah, those are my likely prospects.
0:14:08 Even if I have a small customer base, I can still make use of an almost infinite amount of third-party data to, “Well, gee, you should really advertise towards these people on Facebook and these people on LinkedIn. You should put an ad on that publication because your customer lifetime value people tend to visit those websites the most.”
TEMA: Cool. When you talk about taking a look at other cookies out there, not just your own, is that information accessible to everybody?
JIM: Yes, that is what Facebook, LinkedIn, and Twitter do for a living.
TEMA: But, they do that internally, so you’re saying you buy the data.
JIM: You don’t buy it. It’s like a direct mail list. You rent it. You send them your ad, you send them the attributes that you’re looking for, and they match up who they have in their database, so you have no idea who they are. But, we know that they’re sending to the same kinds of people who match these attributes.
TEMA: Right, so your model, then, has determined what are the attributes you’re looking for or the combination of attributes that you’re looking for.
TEMA: Got it. Okay. It’s exciting from the viewpoint of multivariate testing, the ability to test for multiple things at once. It makes it so much faster. There are a couple of areas that I’m curious about your thoughts on. One is chatbots. In the customer experience business where I play, there is so much hype about chatbots.
TEMA: Yet, the chatbots are still learning, very much still learning. It seems to me that there is a risk of customers being turned off because, essentially, we’re having to train everybody’s chatbots.
TEMA: What do you think about that? To what extent do you think chatbots are going to become truly useful quickly versus just being a gimmick that turns into the voicemail menus of today?
TEMA: Is it just going to become the same sort of frustration for customers?
JIM: They are going to be wildly valuable. They are going to be a preferred way to communicate. It will not happen quickly.
TEMA: Right. In the meantime, customers still have to train the tools. [Laughter]
JIM: Or, companies have to invest.
TEMA: Yeah, it would be nice if they did. [Laughter]
JIM: There are 400 ways that you can ask the same question.
JIM: Either you can hire a bunch of people to sit there and brainstorm 200 of them or you can force your customers to be the trainer, which will upset them to no end and turn them all off. There is a happy combination in there.
0:17:05 But, in the long run, I have a question for a giant corporation, and I have a variety of choices. I can get on the phone, be put on hold, and then go through the interactive voice response hell and then, finally, ask somebody a question who doesn’t know what I’m talking about–
JIM: –or I can go to the website and click around like a mad person for hours and not be able to find the answer, or I can pop open the little chat in the corner.
JIM: Now, there are two choices. Either it is a human on the other end who does not know the answer to my question or it’s a chatbot that’s never heard this question before.
JIM: But, the 27th time somebody asks the same question, the human on the corporate side will be notified that this is one that somebody should sit down, have a meeting, and come up with the answer. The next time it gets asked, boom, the answer comes up instantly.
JIM: Man, thank you. I didn’t have to count on a person. I didn’t have to click around. I went to a chatbot and said, “What about this?” and I got exactly the answer I wanted. Bingo. Great customer experience, but it’s not going to happen quickly.
TEMA: Well, and I wonder because these systems all say, “If that’s not what you wanted, let us know what you were actually looking for.” I wonder how many people are willing to do that versus just hanging up or turning and going to something else, feeling frustrated.
JIM: Enough, I hope.
TEMA: Yeah, exactly. I do it, but that’s partially because I work in this industry.
TEMA: I think, okay, I can see the long-term potential, but it can also be–
JIM: I’m the same way.
TEMA: Yeah, exactly. Another thing that you talked about that I found truly intriguing was this whole issue of dynamic pricing, which companies would love to be able to do.
TEMA: But, as you point out in the book, there’s some real potential for backlash when they try to customize pricing to the value of that particular client.
TEMA: Can you talk about that, please?
JIM: The dynamic pricing that got a lot of attention about two years ago was travel sites that, if you show up on a Windows machine, you get a certain price range of offers for hotels. But, if you show up on a Macintosh or an Apple Air or whatever, you are shown more expensive hotels.
JIM: Now, this was not price fixing. This was not dynamic pricing. This was dynamic offers.
JIM: Because people who buy Apple products have more money and, statistically, they want higher end hotels. Perfectly reasonable and, yet, that was mistaken as, “You’re charging me more money!” No, it’s a different product. We’re offering you a different product to match your persona.
TEMA: [Laughter] Except, it isn’t a different product, though, if they’re offering you the same hotel room at a different price.
JIM: Oh, that’s a different animal.
JIM: That’s not what happened.
JIM: Although, Amazon tried that about three years ago and caught a lot of flack for it.
TEMA: I remember, yeah.
JIM: Because that is the same product at a different price. That is breaking the social contract, if you will.
JIM: I came across that experience on an airline website that will go nameless where I looked at the ticket price. I came back an hour later. It had gone up. I came back a day later, it had gone up again. I deleted my cookies, and it went back down.
TEMA: Yep, and that still goes on. I mean I’ve seen that sort of thing too.
JIM: That, to me, is lying.
JIM: I don’t have any patience for it as a consumer.
JIM: That sort of dynamic pricing and then, of course, there’s the reverse of all of this, which is consumers gaming the system and putting the expensive pair of shoes into the shopping cart and then abandoning the shopping cart knowing full well they will get an email the next day with a 10% off and free shipping coupon.
JIM: We’re both guilty.
TEMA: Yeah, and I think that we, in a lot of ways, are training customers to do things like that.
TEMA: Do you see any ways that companies can then usefully use dynamic pricing?
JIM: Well, let’s talk about airlines again, which use dynamic pricing in a competitive way. It’s perfectly reasonable to expect a “discount website” to monitor how much Amazon is selling things for and then price it at $1 less just to bring in those of us who are adamant about saving a dollar.
TEMA: Right. Yeah. What about small businesses? Is there any realistic way that they can be using artificial intelligence at this stage?
JIM: Yes, and it’s really quite interesting to see. This is an area where midsized companies are going to struggle. Big companies are hiring their own data scientists, and they’re building their own systems. God bless them. I hope they’re successful.
JIM: Small companies are able to take advantage of startups. This is the frothy area of venture capitalism is anything that has artificial intelligence attached to it is having money thrown at it.
JIM: The result is a whole bunch of small companies that are trying out new things. Some of them are working pretty well. Now, there is a sweet spot of, “I’m a small company, but I’m big enough to have some money to experiment with. I’m a small company, but I have a large enough database for it to be statistically significant.” And so, there has to be a culture alignment, a financial alignment, and a data alignment. Then, “Oh, boy, stand back. We’re going to do great things.”
TEMA: Mm-hmm. What are the most common misperceptions you see when companies think about artificial intelligence for marketing?
JIM: Turn it on and let it make up its own mind.
JIM: All of my problems will be solved. It’s magic.
TEMA: Right. [Laughter] Yeah. Yeah, that’s kind of what I figured.
TEMA: And, you know, that’s partially the fault of vendors who claim that.
JIM: Absolutely. Yeah, the difficulty is that I come across companies who are not far enough along the analytics maturity model to take advantage of AI, really. They say, “Oh, well, we have Google Analytics, and we kind of look at it every now and then, and that’s all we know. But, we’re not going to bother investing in predictive analytics or statistical analysis. We’re just going to wait for all of this AI stuff to solve these problems for us.”
JIM: We go back to the human being responsible for what problem to solve, what data to use, and does the answer make sense. If you don’t have people with a firm grasp of statistics, that last part will fail.
JIM: If you just trust anything that the machine says, you will fall down a rabbit hole, you’ll never come back, and you’ll hate it. It won’t work. The other misconception is that it will happen in a heartbeat.
JIM: Again, there’s no such thing as a machine that has learned everything, so an algorithm will have to learn from your data, and it will make recommendations to you from your data that will be wrong. That’s how it learns.
JIM: It is very much like a baby learning to walk. It is going to crawl, which will be very inefficient and expensive, and why are we bothering with it? Then it will take a couple of steps and fall down on its nose and howl.
JIM: You’ll think, “This was a serious mistake.” Then it will take four steps, and you’ll be amazed because you didn’t think it could. Then, suddenly, it’s running out into the street, and you’re chasing it saying, “Wait, wait! Come back!”
JIM: I have seen this happen. I refer to it as the machine bump where a human is doing sort of a natural slope up of optimization. Then you bring in the machine learning, and it just takes a nosedive. It makes horrible recommendations, and you go, “No, no, no, no, that’s not right.” The machine learns, and then there is this amazing growth of return on investment, this amazing, positive outcome that, because it’s leapfrogging what humans can do.
TEMA: That’s right.
JIM: Then, it will taper to just a regular sort of normal rise in optimization because it’s taken that machine bump, but it can’t continue to improve exponentially. But, it can improve bit by bit by bit. Look for it to fail horribly, succeed massively, and then just bubble along just fine, thank you very much.
TEMA: [Laughter] In that bubbling along phase, as you said at the beginning of the interview, of course, competitive situations change over time, so do these systems automatically fine-tune based on changing market conditions?
JIM: They automatically fine-tune based on the new data that you give them and the responsive systems, the reinforcement learning systems, are the ones that are hooked up to outcomes.
JIM: It learns, and it does things. You give it agency, so it can send out emails, then automatically monitor response, and then change its mind a little bit and send out different emails and see how that works and continuously improve.
TEMA: If a company were thinking that they want to sort of dip their toe into artificial intelligence for marketing, where would you suggest they start?
JIM: A couple of things. First, if they are using a large system, if you’re using Adobe Analytics or their customer experience suite, if you’re using salesforce.com, if you’re using IBM customer analysis tools, these companies are bolting on machine learning systems that already understand the data structure, and so they are learning some meta-lessons from other customers and can be quite useful.
JIM: On the other hand, there are these startups, and some of them are small companies that have only raised $50 million and only have a couple hundred people working for them.
JIM: But, have really interesting tools and methodologies. They are worth pursuing. That’s on the tool side.
JIM: Internally, deciding what problem to solve is the major focus. You want to test something out. You want to try it and learn how to use it. For that, you’re looking for a problem that is high volume of data with low risk. Advertising on Twitter, sending out emails, display advertising through AdWords: these things are high transaction numbers, so we get lots of response back, and they’re pretty cheap.
JIM: It’s okay to experiment there.
TEMA: Right. What about the issue of algorithms building in existing biases?
TEMA: I mean that is a big concern, so what are your thoughts on that and how should companies be handling that?
JIM: That is a gigantic concern. That is where the smell test comes into play. Just to briefly outline it, there is already bias built into the data. If you were using a machine learning system to help you with recruiting–you know, we get in 10,000 resumes and we just want to sort out which 50 of them should we actually look at–it’s going to say, “Okay, well, show me all of your corporate employment records and what are the attributes of somebody who has been successful at this corporation?”
JIM: You define success as, well, they’ve received so many promotions, so many raises of pay, and so many bonuses over the last 15 years. It says, “Oh, white males.”
TEMA: Exactly. [Laughter] Yeah.
JIM: Fine. Got it.
TEMA: Yeah. Right.
JIM: The bias is built in. Now, if a human looks at that and goes, “Yeah, no, all right. We have to figure out a way for it to not consider race and gender,” so you remove race and gender from the dataset, and it still creeps back in because it’s inferred, so it’s still going to pick people who are from this particular zip code, have this particular education, or drive these kinds of cars.
JIM: The data scientists are building things called adversarial layers in their neural network systems that say, “If you can determine race, then go back and make it so that I can’t,” so this particular AI is looking to see is there a way for me to infer race or gender? If so, oops, go back and change your model. That just cycles through until the model adjusts to where that adversarial layer says, “Oh, nope. I can’t tell race anymore.” Good, then you can go ahead and continue processing.
TEMA: Interesting. But, I can certainly see the temptation for the old white men to not want to tamper with it. It’s like, this is what we’ve been doing, and it’s been successful. [Laughter]
JIM: That’s part of it. The other part of it is just being lazy.
TEMA: Exactly, and not thinking it through.
JIM: The difficulties in marketing are rather modest. Two parts of this: The difficulty in healthcare, insurance, government spending, that is monstrous, and we have to be wildly careful of all that stuff. In marketing, yeah, it’s not such a big deal, but the problem for the marketer is the machine learning will help you get more of what you already have. So, if these are the attributes of your best customer lifetime value, it will go out and find people just like that but nobody else.
JIM: It’s actually limiting your scope.
TEMA: I can see where, if a company is in a fast growth phase, they’d probably be fine to just stick with what they’ve got. When things slow down, they might want to start broadening out.
JIM: Yes, exactly.
TEMA: Okay. All right, I think I’ve taken enough of your time, Jim. Is there anything that you wish I had asked you and I haven’t?
JIM: You have some great questions there. Thank you so much. I wish I had some brilliant question that would provide insight for everybody but, other than that, I think the proper answer is, yes, you can find my book on Amazon, and I’m available to do workshops.
TEMA: Okay. Thank you very much, Jim. It’s been an absolute pleasure and honor to speak with you.
TEMA: What’s my main takeaway from that? Yes, we all need to keep our eye on AI. Experiment with it to the extent that you can, but don’t expect massive immediate profits from it.
It’s like the Internet, though. Stay focused on what you really need it to do for you. What is an important problem that AI could solve for you and/or for your customers? Don’t expect it to be a magical solution to all your organization’s problems.
[0:35:47 end of audio file]