If so, we'd love to hear it!

Simply click the link below and submit any ideas that you have for us to talk about on future Wheel of Risk episodes!

EPISODE TRANSCRIPT:

00:00:02

Alix McCabe: How do you feel about the AI Revolution? Is your organization ready? Will A. I. actually improve productivity? How can it enhance the customer experience?

00:00:11

Clip: They're just questions, Leon. It's a test designed to provoke an emotional response. Shall we continue?

00:00:17

Alix McCabe: See what we did there? That is a clip from Ridley Scott's dystopian sci-fi masterpiece, Blade Runner, in which synthetic humans have become nearly indistinguishable from the real deal.

00:00:27

Clip: What is real? How do you define real? If you're talking about what you can feel, what you can smell, what you can taste and see, then real is simply electrical signals interpreted by your brain.

00:00:39

Alix McCabe: And yes, that's Morpheus from The Matrix, another classic that blurs the line between human and machine. Despite how often A. I. Has been featured on film, it's still not clear where the technology will take us.

00:00:50

Clip: What is this? Where's your body?

00:00:52

Clip:I am a programmer.

00:00:54

Clip: This feels weird.

00:00:57

Alix McCabe: Let's embrace the age of A. I., and make some space for it on-

00:01:00

Clip: Wheel of Risk.

00:01:04

Alix McCabe: Welcome to Wheel of Risk, proudly presented by Allianz Trade. I'm your host, Alix McCabe. On every episode, we spin the wheel to choose a topic, and then provide expert insights and actionable advice to help you keep your business solvent, secure and well ahead of the competition. Today, I'm super excited to introduce a longtime fan but first time contestant to the Wheel of Risk Studio. He's been diligently sharing the podcast on LinkedIn since way back in season one. Aaron Lindstrom is the regional head of e-Commerce for Allianz Trade Americas. Aaron, thank you so much for joining us.

00:01:37

Aaron Lindstrom: No problem, Alix. I really appreciate the invitation. Excited to be here today.

00:01:41

Alix McCabe: We are glad to have you. Since you're such a loyal listener, I assume you know what we do first.

00:01:46

Aaron Lindstrom: Yeah, I spin the wheel and you tell me how much money I won.

00:01:49

Alix McCabe: Basically, that's it.

00:01:51

Aaron Lindstrom: Perfect.

00:01:51

Alix McCabe: Should we do it?

00:01:52

Aaron Lindstrom: I actually brought my little robot assistant to spin for me today.

00:01:55

Alix McCabe: Okay, this is a first.

00:01:56

Aaron Lindstrom: Well, I'm the digital and technology guy, so here we go.

00:02:00

Alix McCabe: All right, bring him out.

 

Wow, that was an impressive effort. The robot has landed on an incredibly timely topic, Aaron, artificial intelligence, natural opportunity. What do you think?

00:02:17

Aaron Lindstrom: I think it's almost like the robot picked it for us.

00:02:19

Alix McCabe: I, for one, welcome our new robot overlords, so let the revolution begin.

 

Despite all the attention it's been getting lately, artificial intelligence has actually been around in one form or another for a while. It was first founded as an academic discipline way back in 1956, but it wasn't until the advent of deep learning in 2012 that interest and funding really took off. Today, thanks to natural language processing and generative AI, the technology has gone fully mainstream. A host of AI-powered chatbots, assistants and apps has come online with implications for a huge variety of industries, from logistics to healthcare. Before we dive into the impact on the business world, let's make sure we're on the same page with our terminology.

 

Okay, so Aaron, what is generative AI, natural language processing, machine learning, and what's the difference between them?

00:03:09

Aaron Lindstrom: We get a lot of people that get intimidated by AI technology, and the terminology that's around there. These are all different fields of AI. Within the discipline of artificial intelligence, which is really underpinned by getting a machine to solve problems with human-like qualities, perception learning, decision-making, reasoning, three of the big ones, as you've mentioned, are that natural language processing. That's really getting a machine to understand the difference in dialects, the difference in lexicon. Think about, " Wow, it's great to be here today," versus, "Wow, it's great to be here today." Getting the computers to understand the difference between those two phrases, that's really your natural language processing large language models.

 

Degenerative AI is where we've seen a lot of press, where artists are creating songs or art pieces, those things, so it's creative. The GPT models you hear about really combine those. It's a generative, pre-trained transformer, that's what GPT stands for, so it synthesizes both of those. And then really, underpinning almost every form of AI is machine learning and deep learning models, where you're using large sets of data to train an AI, to make decisions, and really get feedback on its own without human intervention. The machine teaches itself what is right, what is wrong in any given decision matrix.

00:04:37

Alix McCabe: Okay, got it. Of those technologies that you just described, which would you say is the most useful for businesses in the context of E-commerce?

00:04:45

Aaron Lindstrom: I think right now, machine learning is definitely the most important, especially since it underpins so many other areas of artificial intelligence. As we evolve, you're really seeing more, though, of the generative language models, your chatbots, those things are AI 101 for a lot of E-commerce, but you also have to look at the back end of E-commerce. The deep learning models, the natural language models are being used to analyze everything from supply chain to the ESG risks of different supply chains, supply chain snags, how much inventory to have, what's likely to sell when, a ton of uses that we're starting to see in the other areas of artificial intelligence.

00:05:27

Alix McCabe: You touched on some of this, but how can AI help businesses understand what customers or clients are actually looking for?

00:05:33

Aaron Lindstrom: I think a lot of that starts, Alix, with just the whole search function. There are some recent stats out there that say that more product searches now start in Amazon than they do Google. Everybody's really racing to get these language processing models to understand what you as a consumer want. If you put in something like a purse, is that a racing prize or is that a handbag, and understanding what the different context is behind those product searches. People trying to describe a product that they saw on a social media post, or on a TV commercial that they don't actually know the name for. Using those language-processing, those fuzzy logic disciplines to try and get the consumer to the right product, I think that's a huge mainstay use. And then obviously, automating customer service functions through things like chatbots, interactive shopping processes, I think that's another really critical area for businesses to look at.

00:06:28

Alix McCabe: Just FYI, for me, a purse is always going to be a handbag.

00:06:31

Aaron Lindstrom: Well aware.

00:06:32

Alix McCabe: You mentioned this a little bit, chatbots, what's the liability of a chatbot if it provides people with inaccurate information?

00:06:39

Aaron Lindstrom: That's one of the big fears that industry has, obviously. There's an article that was published, I believe it was February or March of this year, where an Air Canada chatbot gave a consumer the wrong information, which normally, you wouldn't think would be a big deal, but it was all centered around a bereavement fare, so you have a heightened emotional state. Air Canada actually tried to state that the chatbot was its own individual answering on its own, and not for the company. They lost that, but that's really where you start to get some of the challenges around chatbots, and the fears. Are they going to give misinformation? Are they going to leave a consumer more frustrated? I think that's just part of the curve.

00:07:22

Alix McCabe: Right.

00:07:22

Aaron Lindstrom: It's really not that much different than hiring a new employee to be on a chat line. They could make a mistake early on in the process, they need to learn, but it does really open up doors as more and more consumers are purchasing and interacting with companies during non-traditional business hours.

00:07:37

Alix McCabe: Okay, fair enough. How can AI tools help companies improve the targeting for, let's say, their marketing and advertising efforts?

00:07:45

Aaron Lindstrom: I think this is where the natural language processing plays a big role. Looking at sentiment analysis of people's interactions with companies, what they're looking for, building better personas of your buyers, knowing that for you, a purse is always a handbag, that's critical. If you like handbags, maybe you like other accessories, or you like certain brands. Trying to make those connections and present you with the most applicable choice possible, I think is critical in any customer-facing business, whether that's an end-using customer or another business that's looking for inputs for their products, really guiding that user journey and getting people to the right product in the fewest clicks is paramount right now.

00:08:29

Alix McCabe: Yeah, and like you said, it's applicable not just for business to consumer, but business to business marketing and advertising as well.

00:08:35

Aaron Lindstrom: Yep.

00:08:36

Alix McCabe: Maybe let's switch gears a little bit, and talk about how AI can improve cybersecurity, which is another huge opportunity. What are you seeing in that space?

00:08:44

Aaron Lindstrom: Yeah, that's a real fun one because it's a race between the good guys and the threat actors of who can get the most systems integrated initially. There are certain things for E-commerce specifically, and in our Allianz Trade pay product, we're really looking at buyer identification fraud, making sure that we have the right company, the right individual associated with that company. That's been a critical one that machine learning, natural language processing, those things can play a role in for all businesses. Identifying fraudulent bank accounts, identifying fraudulent transactions, these are all big areas, as well as just security around access to your site as well, to pick up not who is definitely a threat actor, but who could be a threat actor, and then flagging that in many cases for human interaction.

00:09:32

Alix McCabe: I know there've been some dramatic improvements in this area. Anything else to add about how AI is changing fraud detection algorithms?

00:09:39

Aaron Lindstrom: Oh, yeah. It's amazing what these fraud-detection algos can pick up on, Alix. I get really excited about this stuff. We all know the CAPTCHA, right? That's where you click which box has a motorcycle. Well, if you've noticed recently, more and more of the CAPTCHAs, you're not even required to click a button other than to say that you are not a robot. The reason for that is, the algorithms are now actually reading the movement of your mouse. If your mouse jumps from point A to point B in a perfectly straight line, you are definitely a robot.

00:10:07

Alix McCabe: Wow.

00:10:07

Aaron Lindstrom: If you type at a certain speed, you are definitely a robot. The typing one is pretty fun. They figured out robots can fill out these forms really quick, so now, there's regulators. If the form is filled too quickly, you've got a fraudster or a robot on your site. The fraudsters and the robots started putting pauses between characters. Well, now they can detect if the pause is not the right length, or all the pauses are the same, then you are definitely a robot. They're so good at this, they can even tell if you're primarily left-handed or right-handed, if you're on a phone or a keyboard, just based on the different typing speed and the lays where the keyboards are laid out. They can even sometimes tell what language keyboard you are using, and tie that to your IP address. If your IP address says You're in the U. S. but you're using a French keyboard, that might be something you want to investigate.

00:10:56

Alix McCabe: That might be a flag.

00:10:57

Aaron Lindstrom: Yeah.

00:10:58

Alix McCabe: Interesting. Let's talk a little bit about Allianz Trade. What is Allianz Trade doing to take advantage of the possibilities presented by AI?

00:11:05

Aaron Lindstrom: There's a number of initiatives that Allianz Trade is undergoing. The first is really in our risk scoring, and that's really heavily dependent on machine learning, taking large data sets and making accurate projections, predictions that you can then back test once that decision has been made. Fraud detection, for example, in our Allianz Trade pay product, in our claims process to identify both fraudulent buyers, potentially fraudulent claims. A lot of that, again, is based on language analysis, sentiment analysis, making sure that the data is accurate. And then, as a company as a whole, we've launched a series of learning weeks really around how we can use AI, with each of our business units or regions being responsible to do a ground up AI project that they feel will implement successfully in the business, so a lot of different initiatives. The concern you had about bots earlier that you mentioned, we call it hallucination, where the chatbots will make things up, that obviously is tricky in a regulated industry. We do want to make sure we go slow, and we do it the right way, but definitely exploring those various initiatives within the company.

00:12:12

Alix McCabe: Yeah, we don't need hallucinations. Let's avoid those.

 

You mentioned Allianz Trade pay. Can you share with us what that is?

00:12:27

Aaron Lindstrom: Yeah, Allianz Trade pay is really our entry into the B2B payment space in E-commerce. It's centered around these high-volume, low value transactions, where a merchant selling online now has a new buyer potentially coming in that they've never interacted with before. Buyers always want credit terms when they can get them, so we're using machine learning to make real-time grading decisions about those buyers, whether they should be approved for credit, how much credit to be approved.

Also, really another critical piece in E-commerce that Allianz Trade pay is tackling is making sure that you have the right buyer and the right individual within that company. If somebody registers as alix@gmail. com but says they work for Allianz Trade, that's a great red flag that, "Okay, maybe you are Alix, and maybe you do work for Allianz Trade, but maybe this isn't a sanctioned purchase."

 

Identifying those type of areas where we can help merchants really transact safely, quickly, efficiently, without that risk that they're dealing with a threat actor or a fraudulent buyer down the road. And then, handling the whole payment flow, which is another critical piece to business to business transactions. The days of salespeople walking into a customer's office to pitch their widget still exist in some areas, but more and more of that is online obviously, as a response to the pandemic. E-commerce had a huge boost, so we really want to make sure that we can provide our clients with an omni-channel solution, not only for their larger recurring transactions, but also these high-volume, low value transactions where, if they don't have the systems, it takes a little bit more work, but if they do have the systems, it's typically a lot more profit margin for them.

00:14:06

Alix McCabe: It's an exciting space.

00:14:08

Aaron Lindstrom: Definitely.

00:14:09

Alix McCabe: It's worth acknowledging, though, albeit exciting, that there are some business owners and leaders out there who are maybe a bit nervous about AI, or don't fully understand the opportunities it presents. What would you say to them?

00:14:21

Aaron Lindstrom: I would say good luck and Godspeed to you.

00:14:23

Alix McCabe: Good luck, my friend.

00:14:25

Aaron Lindstrom: There were a lot of people that didn't like those newfangled personal computers when they came out, or them fax machines. Whatever it is, whatever piece of new technology, there's always reluctance in some folks to adopt that technology.

00:14:37

Alix McCabe: Sure.

00:14:38

Aaron Lindstrom: In the case of AI specifically, whether you want to adopt it or not, it's going to happen. Your supply chain is going to be using AI tools. Your clients are going to be using AI tools. If you have a website, it is being rated by AI tools. Traffic is being directed by AI tools. You can either be part of that, which could have great positive impacts on your business, or you could just sit on the sidelines, but that tide has turned. You're going to be part of the AI system whether you want to be or not, because everything around you in the business world is going to be AI-driven to some extent.

00:15:14

Alix McCabe: When we talk about the speed, because that's one of the very interesting things, the speed with which generative AI is rolling out and evolving, and the potential benefits as well as the risks for business owners, what are you most excited about? I know you get pumped on this topic, so what are you most excited about?

00:15:30

Aaron Lindstrom: I really believe that this is an opportunity for us to become more and more efficient in our business processes and our daily lives. I do believe that AI is not going to cost people their jobs. It's going to enhance their living standards. We have a lot of employees, and I say we, that's the world, not specifically Allianz Trade, that are in these really repetitive, low-meaning functions, and all of those could be automated. You hear about automation in robotics and manufacturing, in other industries where the actual production process was costing human lives. Now, we can have robots do those jobs, but someone needs to train the robots. Someone needs to fix the robots, someone needs to dream up the new robots, the new processes.

 

I think it is more of a shift into a digital economy, but I think we will see more job satisfaction coming out of that as we can take away a lot of those low-value tasks. Again, I think from a global connectivity standpoint, if you have a website, you're potentially exporting products anywhere in the world, so the ability to use AI functions to allow you to serve any customer in any geography at any time of day is just a game changer for a lot of these businesses. I see a ton of upside on this.

00:16:49

Alix McCabe: What about people who are worried about AI taking over the world? That certainly is a concern, right?

00:16:55

Aaron Lindstrom: Yeah, I think that's more the science fiction, Alix, than the science fact. OpenAI gets cited in a lot of these articles, both as the good guy and as the bad guy, depending on who is writing and what the literature is around. I think they did research paper recently, it's a 98-page report that they released, trying to figure out if GPT-4 had what they called any dominance behaviors. If, given access to all the algorithms and actually some funds, could it replicate? Could it expand its power? Could it become more influential? There's a funny story that came out of that where GPT-4 ran up against a CAPTCHA model that it couldn't solve.

00:17:36

Alix McCabe: Really?

00:17:37

Aaron Lindstrom: It actually went out and hired a gig worker who asked it, "Hey, just to make sure you're not a robot in GPT-4," which was instructed not to disclose that it was a robot, said no, it had a vision impairment.

00:17:53

Alix McCabe: Oh, wow.

00:17:55

Aaron Lindstrom: It hired this worker to get through a CAPTCHA to beat a website security that's designed to keep robots out. While that was a little frightening to some people, a lot of the facts were left out. It cited the wrong type of CAPTCHA. It referenced a company that you can subscribe to automatically fill CAPTCHAs as what it was trying to access.

00:18:15

Alix McCabe: Okay.

00:18:15

Aaron Lindstrom: There were a couple of other linguistic errors that another AI would have caught on in NLP, and flagged it as a fraudster. There were no behaviors of it trying to self-replicate, trying keep itself from being turned off in the wild, as they call it, or find power in the wild, or any of the really scary behaviors that we associate with things like The Terminator, or Blade Runner or those things. Definitely growing and pushing boundaries, but so far, there's no evidence that AI either wants to or is capable of taking over the world.

00:18:45

Alix McCabe: Well, that's comforting at least.

00:18:47

Aaron Lindstrom: Yeah, it's a fun story, but again, I think it's just one of those situations where if you use the same tool that other people are using, you can catch those things pretty easily, and if you're not, it's one of those areas where you might be exposing yourself to some unnecessary risk.

00:19:02

Alix McCabe: Yeah, makes sense. Well, I hope you're good at separating the real from the artificial, Aaron, because that music that we're hearing means it's time for a special quiz. We are going to give another nod to Blade Runner by calling it our very own Voight-Kampff test.

00:19:17

Aaron Lindstrom: Ah, nice. I am not a replicant. I am a real boy.

00:19:20

Alix McCabe: I know this about you.

00:19:21

Aaron Lindstrom: Yes.

00:19:22

Alix McCabe: Here's the deal. Our producer has provided us with a series of people dispensing dubious financial advice. I'm going to play them, and you and I have to guess which ones are actual humans, and which ones were generated by AI. Does that work?

00:19:35

Aaron Lindstrom: Is this where I win the prize?

00:19:36

Alix McCabe: Well, we'll see. I'm going to compete as well, and I don't have the answer.

00:19:40

Aaron Lindstrom: All right, we'll see here.

00:19:41

Alix McCabe: All right, it'll be fun or it'll be terrifying. Either way, let's roll the first clip.

00:19:46

Clip: For instance, following influencers on TikTok and Instagram is a great way to learn about investment opportunities that you won't hear from an advisor.

00:19:55

Alix McCabe: What do you think, Aaron?

00:19:56

Aaron Lindstrom: That is dubious financial advice, but I'm going to say that's a real person.

00:20:00

Alix McCabe: I don't know. All right, I'm going to go with AI, just to be different.

00:20:02

Aaron Lindstrom: Okay.

00:20:03

Alix McCabe: Wait, I have to write this down, because I am going to tally the scores at the end. All right, let's do the second clip.

00:20:10

Clip: I would say it's best to rack up as much debt as possible on a credit card in order to demonstrate that you deserve a higher credit limit.

00:20:19

Aaron Lindstrom: You were right, this is spot-on dubious advice. I'm going to go with an AI on that one.

00:20:24

Alix McCabe: I am too, actually. I'm going to go with AI. This is pretty tricky, actually. All right, let's hear clip number three.

00:20:31

Clip: Every investor should have at least 10% of their retirement portfolio in cryptocurrency as a hedge against inflation.

00:20:39

Alix McCabe: Okay, what do you think Aaron?

00:20:41

Aaron Lindstrom: I'm going to revert back to real person on that one.

00:20:44

Alix McCabe: I'm going with human on that one, too.

00:20:46

Aaron Lindstrom: Okay.

00:20:47

Alix McCabe: All right, I think we have one more to go. Let's hear it.

00:20:51

Clip: Another thing you can try is, if you start an S corporation and claim everything you buy as a business expense, you don't need to pay taxes.

00:21:00

Alix McCabe: That one had some inflection in it, too.

00:21:02

Aaron Lindstrom: Yeah, the pauses were a little over dramatic to me. I'm going to go with AI on that one.

00:21:07

Alix McCabe: Okay, I'll say human just to be different.

00:21:09

Aaron Lindstrom: All right.

00:21:10

Alix McCabe: All right, so I actually don't have the responses. I'm going to have to go to our producer to tell us. For clip number one, was it human or ai?

00:21:18

Luke: Hey Alix, it's Luke here. This might surprise you, but those were actually all AI.

00:21:23

Alix McCabe: Stop it.

00:21:24

Aaron Lindstrom: Oh, that's not fair.

00:21:28

Alix McCabe: That was a trick.

00:21:28

Aaron Lindstrom: That is so not fair.

00:21:29

Alix McCabe: That was a trick. They were all AI, so it's a tie.

00:21:33

Aaron Lindstrom: Oh, wow.

00:21:33

Alix McCabe: It's a tie. But you know what, Aaron? Since you're the guest, I will give this one to you, and I have a major prize for you. Let's put 30 seconds on the clock, and you can tell folks at home what you'd most like them to take away from our conversations today.

00:21:47

Aaron Lindstrom: That sounds like a great prize.

00:21:49

Alix McCabe: All right, go for it.

00:21:50

Aaron Lindstrom: I would say be open, and look optimistic at AI. We're not looking at Skynet coming online anytime soon, or anything that's super scary like that. It's another tool. If we don't take advantage of the benefits of AI, there's no way that we can be involved enough to worry about any of the risks. I would just say, welcome the technology, get familiar with the technology, get comfortable with the technology, and help shape the future rather than react to it.

00:22:19

Alix McCabe: That's really good advice.

00:22:20

Aaron Lindstrom: Thank you.

00:22:21

Alix McCabe: If I want to get more familiar, if I'm a business owner, just a general person, if I want to become more familiar with AI, do you have any resources that you'd recommend?

00:22:29

Aaron Lindstrom: It's funny, because all the search engines now are integrating AI. You can just ask the search engines, but there's a couple of different organizations. The Allen Institute for Artificial Intelligence here in my home city of Seattle, Washington is a great one. There's the National Science Foundation Artificial Intelligence Research Institute, that's a U. S. government website that folks can look at for resources. Just about every major university in the world right now that has any type of computer science program has resources available for AI. I would really encourage folks to start there with the facts, both to protect themselves from just a groovy sales pitch that somebody throws out there, but also to help alleviate the fears. Even OpenAI puts out a ton of resources. They've published all kinds of research papers where they've given GPT-4 and other models to nonprofits to test, to see if they show any threat behaviors, and all of that is public record information. I think it's like anything else, it's just self-education and your degree of comfort.

00:23:30

Alix McCabe: Yeah, that's good advice. Thank you so much for sharing your thoughts with us, and for being part of our podcast today.

00:23:34

Aaron Lindstrom: It was a pleasure, even if I didn't win a prize. Thank you, Alix.

00:23:38

Alix McCabe: I'm so glad.

 

My guest today has been Aaron Lindstrom, the regional head of E-commerce at Allianz Trade. I'm Alix McCabe, and this is Wheel of Risk. Thanks for listening.

 

If you learned something today, we'd really appreciate if you could leave us a nice review wherever you get your podcasts, or share this episode on your social networks.

Talk to you soon.