Speaker 0 | 00:00.080
Right. Welcome everyone back to Dissecting Popular IT Nerds Today. Doom and gloom, fear, how AI is taking over the world, all those horrible things that are probably, well, we don’t know yet, but we’re hoping it’s not going to happen. Welcome Edward Chouinard to the show, Artificial Intelligence, AI Strategy, Data and Analytics Leadership. Very happy to have you on the show. And so welcome.
Speaker 1 | 00:27.147
Glad to be here. Thank you.
Speaker 0 | 00:28.327
And, um. So yes, why AI, first of all? If it’s something to be feared and hated, why are we talking about it so much, and why is it taking over? And has it taken over yet, or is it kind of just in the joke stage?
Speaker 1 | 00:42.756
I don’t think it’s taken over yet. And I think calling it intelligence is still being a bit favorable towards it.
Speaker 0 | 00:51.478
I have an AI guy. I tried to have, and he’s really good with like… like, you know, knowing all the prompts, he’s been playing with it since the inception. He’s pretty much had to, he was able to quit his day job and do just AI all the time. And he got really good at having it mimic my, somewhat of my writing style because I was a creative writing major. That’s how I got into technology, of course, being a creative writing major. And he was able to get it to mimic my writing style somewhat, but as it moved on it just became too it was too animated i think you could just tell it was kind of cheesy and not hitting on point anymore so i mean i think there’s a lot of great potential with ai but it’s the
Speaker 1 | 01:39.264
it’s it’s just still a tool if we’re going to have some issues with it it really comes down to the fact that there’s an idiot behind the keyboard so uh as far as it leadership goes you
Speaker 0 | 01:54.072
What should we be concerned about? Or is there anything that, one of the topics on the show that comes up a lot is technology leaders stepping into the business seat and no longer use, just keeping the technology running, so to speak, right? Managing the silos, being stuck in a cost center. What does, what can we use or what do you foresee in the future AI being used for?
Speaker 1 | 02:22.048
as a business force multiplier to help the business do what a business is supposed to do primarily which is make money well i myself actually came more from like the business analytics background not necessarily like a technical background so i’ve always had sort of this interesting position that i’ve been in uh i’ve always been very focused on the customer experience now with ai i think a lot of conversations that i see in the media they’re talking about like all these jobs being lost and i’m like Yeah, some jobs are going to get lost, but some jobs are going to get created. So how do I position my team to be well-versed? So, you know, like I mentioned before we started recording back last December, I’m sitting in my house with COVID and reading all these articles about OpenAI, which had just really come out. Started playing around, chat GPT, and I’m sitting there telling everybody at work saying, hey, you got to get on this thing. This thing’s going to take off. Of course, they’re all like, okay, you probably got fever hallucination or something. There was two people that actually listened to me. Then around April, they’re like, hey, everyone’s talking about this stuff. We should get on it. It’s like, yeah. Did you ever check out that prompt engineering course I sent you all? No. Okay. I see it as like, yeah, we’re at that early stage. There’s a lot of gimmicky things going out there. companies are struggling to really come up with like really good business cases for ai that i’m that i really haven’t seen up there that i’m like hey this is like really innovative a lot of it’s just like like this morning i saw somebody saying like hey you can use chat gpt to find stuff on a map i’m like so basically it’s google map with chat gpt what’s the newness here what what are you really solving that i don’t have a solution for already okay A lot of companies are still struggling with that. We got to get over this gimmicky stage to that real, like, what pain point did I just solve for someone?
Speaker 0 | 04:21.294
So since you’re in logistics, let’s, maybe we can use that as an example. Do you have any ideas? So for people out there listening, logistics, we’re talking about transportation, we’re talking about those guys that drive those big trucks down the street, independent truckers, truckers that work for a business, and it’s kind of a… If you have a bunch of trucker friends like I do, which I just happen to have a bunch of trucker friends, then you know that they’re constantly complaining about load prices and gas prices going up. And even though COVID created a lot of demand, the gas prices went up, but the load prices didn’t go up. And now they’re not making any money anymore. And they’re pretty much looking for another job or a way to get out of the business. Or they’ve figured out, or if you talk to the right guy, he’s figured out some little niche. in the trucking business that makes more money than the others. And you talk with the Home Depot guy that drops off your load at the house and he’s like, I love it. I just work local from Monday through Friday. So this is kind of the, that’s on the trucker side of the business. Then on the actual other side of the business in the trucking industry, you have dispatch, you’ve got, what do you call the guys that book all the loads? What’s wrong with me? Why can’t I think of the name? I need you. Yeah. Yes. Thank you. Brokers. You’ve got the brokers and you know, putting all these loads on what I imagine is like a, let’s call it a single pane of glass, since salespeople like to use single pane of glass all the time. You know, going to, yeah, like a broker and bidding on loads, et cetera. Am I describing this correctly? Yeah. From someone, from a layman that’s kind of, and kind of understands logistics. So how can AI make a difference in that world? Well,
Speaker 1 | 06:03.367
I was building out like a logistics-based large language model. Now, A lot of people might be like, well, why does logistics need one? Well, logistics has a lot of like interesting terms that most people would be like, well, what’s a reefer? What’s a dry van?
Speaker 0 | 06:19.170
What is a reefer? I do want to know what a reefer is.
Speaker 1 | 06:21.672
Refrigerated trailer.
Speaker 0 | 06:23.333
Ah, that’s a reefer. Reefer madness.
Speaker 1 | 06:28.456
And you know, I hear these guys talking about dry vans and I’m sitting here picturing like a minivan and I’m like, what the hell does that have to do with moving a bunch of pallets? but no that’s a typical 40-foot tractor uh trailer and so for like that’s called a dry what’s that called again right drive-in is there a word for a flatbed well flatbed okay i don’t know maybe there’s you know like what what is like you know optimization well depends what you’re talking about or full distribution uh you know what is drayage you know all these terms that somebody who’s not who’s new to the industry is going to have to go out and learn it. It’s like, hey, if you had like a large language model that you could just use for like training, that would be great. So that’s like an easy one. But now it’s like…
Speaker 0 | 07:12.119
You’ve got me thinking here, but hold on, hold on. Don’t go anywhere. I guess you got me thinking because I’ve been in, obviously, telecom and technology and, you know, for 20 plus years. And I’m building a training program for entry-level people right now. And we’re on the tail end of the entry-level program. And believe it or not, my AI guy has been building most of it. But the one thing that we’re trying to come up with is a… huge glossary of terms that these people need to know what would be your suggestion there and it’s not reefer it’s like you know it’s mpls it’s loa it’s fock date foc date whatever there’s all these different you know terms and acronyms and insanity there so i mean you know let’s see what else do we have demarcation demarc point we’ve got 66 block numerous things like you know so what do we do there well
Speaker 1 | 07:59.271
I think a lot of companies…
Speaker 0 | 08:00.191
Selfishly, I’m selfishly asking because I’m going to go have them do this after.
Speaker 1 | 08:04.093
A lot of companies, they probably have their own terminology library. And if not, well, create a bot to go scrape the web. Or, hey, hire somebody cheap to just pull all these terms from the internet. Well,
Speaker 0 | 08:16.980
if we hire someone cheap, isn’t that defeating the purpose of AI?
Speaker 1 | 08:21.242
Well, that’s the funny thing. That’s what OpenAI does for their training is hire a bunch of cheap people to help train the model.
Speaker 0 | 08:29.171
You mean to start putting prompts in, try different things, that type of thing?
Speaker 1 | 08:32.232
Oh, yeah. That’s like tuning, training. So you want to have a good database of the terminology. And then you want to actually have a list of questions. This is what we did was just went to the company and just said, hey, what are some questions that are kind of difficult the layman would not know that you would want to ask a tool? So they gave us a bunch of questions. And we’re just going in, feeding the questions. Like, are we getting the right answer here? Yes, no. If not, how do we tune it? How do we get there?
Speaker 0 | 08:59.127
to get that then it was just like how do we stop the hallucination so it does does it make stuff up so then we put in like a graph database to add some rigidity to it two of those things together work how did you do that how did you add a database or yeah how did you so so well first of all the people person sounds interesting but can you walk me through that like the first section which is have people feed it you know uh prompts uh number two what database and how did you feed that to it so it
Speaker 1 | 09:27.038
the so the first thing is like uh selecting one of the the open source uh lms don’t don’t try and like build this completely by yourself that’s like a waste of time right now when you can go get like Dolly, one of the Falcons, Lama 2, and use one of those to start out with. So first go in there, put it on your system, start asking it the questions you want, and see how good is this.
Speaker 0 | 09:54.376
Give me an example from logistics.
Speaker 1 | 09:57.957
What is a quote? Simple question. So you should be able to understand a quote is a price that you’re asking for moving a truck, say, from Chicago to Dallas. If it’s saying like, well, it’s an insurance quote, you know, it’s needs some work. So those are like where we started out with just asking those simple questions. Then the questions get more complex, like, hey, what is what is full distribution? Why would I use full distribution? And again, checking to see how good does the answer come across? Now, it’s more qualitative at this point because it’s not like you’re looking at a score chart. It’s more like, hey, if I actually know what I’m talking about. Does this make sense?
Speaker 0 | 10:41.077
Okay. Yes, exactly. Exactly. Or is this just a, or is this like a sales rep that’s been in the office day one and you ask and you start talking about, I don’t know, IP addressing and adding and BGP and stuff like that. And he’s like, yeah, yeah, yeah. Yeah. It sounds like you need some networking and it doesn’t make any sense or it’s real simple. So do you have to feed those questions? Like when you say, what is a quote? You have to say like, Hey.
Speaker 1 | 11:06.865
pretend you’re like in the transportation industry or whatever what is a quote yeah so i well personally i like telling the the model to i like giving it a personality and saying like hey you are xyz type person and you should know this this this and this great i always you know whatever i’m doing i always find that pretty good even you know if i’m using like chat gbt bard or quad too i do the same thing so i just find it helps
Speaker 0 | 11:35.445
we’re thinking of doing one that’s uh i’m probably i shouldn’t give this away because it’s kind of giving away but we want to be like hi chad gpt pretend that you’re like my old uh granny in um england or whatever and you know nothing about technology right like explain uh the whatever to explain whatever to my my 90 year old granny explain uh mpls to my 90 year old granny or something like that you want some entertainment you can tell it to kind of it’s a reddit troll you Wow. That’s great. That’s actually outstanding. Troll every hater on my LinkedIn feed, please. If you could do that, that’d be great. Every now and then I get some haters. Okay. So, all right. So we were feeding a bunch of questions and how is it responding?
Speaker 1 | 12:27.681
you know, again, it’s more qualitative. So it’s like, you know, there’s a bunch of us being like, does it sound good? Yes. No. If you get the majority saying, yeah, okay. If it’s not, it’s like, well, what data would we need to train it? And, you know, luckily for us, we have some libraries internally that helped us out. And if we didn’t have it, my first stop was to go to the Snowflake marketplace because we’re using Snowflake and there’s like a lot of free data sets there. So I was like, I will just go there. We were already set up with it.
Speaker 0 | 12:55.209
So how do you train it? How do you train it and feed it? Because this is beyond my level now. So how are we feeding this?
Speaker 1 | 13:01.092
My principal data scientist really takes over from me. He’s there. He’s in there working with my head engineer, data engineer, pulling data in from Snowflake, then feeding it into the model. Exactly all the steps they do. Not quite sure on all that, just because I was just like, I’m looking.
Speaker 0 | 13:19.202
Once they feed it, is it now global or is it proprietary? In other words, are we feeding an open AI model data that it might not have already? So you’re basically helping everybody else. And that’s okay. I’m not saying that’s bad. I’m just, you know, I want to know.
Speaker 1 | 13:32.953
Well, this is an intern. So we’re not going out there like scraping data to have permission to use. So, you know, before we grab the data set, we’re looking at terms of services. Like, can we actually use this or not? Yep. That’s one of the things I don’t like about open AI. It’s like, hey, I don’t know where you frame this model on what data. Like, hey, if I’m a content holder with copyrights, I’d probably be suing them, seeing what I can get out of them. And I don’t want to be part of that lawsuit. So we’re quite cognizant about what data we’re pulling in there and do we have permissions to do it. So we’re not going out scraping the Internet saying like, oh, hey, this looks good. Let’s just take it. No,
Speaker 0 | 14:13.463
I’m saying once you train it. So, you know, you said you’re training, you’re feeding it data. whatever is proprietary you own you got it you have rights to it whatever now once you train this model that’s not escaping the environment is it no it’s not it’s all in our own systems nothing leaves our systems uh everything’s there are you running this in the cloud or some little server sitting in a closet well our first iteration was basically principal
Speaker 1 | 14:36.931
data scientist laptop but then that got a little too big So yeah, then we moved it into, uh,
Speaker 0 | 14:45.063
let’s see, where are we going? Oh,
Speaker 1 | 14:47.365
first we did data bricks and then it’s more of them. Then they decided they wanted to use SageMaker and AWS.
Speaker 0 | 14:54.911
Hmm. Okay. See, this is great. Cause now, now, now it grows. Now, now we’re seeing this, this is actually, this is awesome. Okay. So we fed it some data. What was the second step? I already forgot. Um,
Speaker 1 | 15:10.028
Well, you got to go back and once you feed it the data, then you got to go back and ask it the questions again. Right. If there was any improvement to it. And? And, you know, if there is, it’s like, great. If not, it’s like, okay, so what else might we want to put in here to improve it? And that’s really where it’s like, this is where you are going to probably get stuck because it’s like, hey, there isn’t any manual out there to say, hey, to train it, go get this data set. So you have to really. put your creativity hat on and figure out where can i get it and do i have a right to that data now if you’re trying to be ethical in what you’re doing you want to make sure you have a right to that data if you don’t well you can deal with the consequences that probably will come with that because i think we’re just at the beginning of the uh all the legalities that are going to come up so like when i started uh 10 years ago well 10 plus years ago doing big data and data science at best buy started that group there we were getting sued on a regular basis i feel like i got a uh just from that experience people just yeah we’re you know we think you’re using it we don’t know if you are and good documentation say this we never lost well we never went to to court because in discovery we could show we never violated anyone’s patent only once did we actually pay and that’s because legal said it’s cheaper for us to pay them the 25 grand than to actually litigate it even and i was like
Speaker 0 | 16:35.608
no we didn’t do anything wrong let’s do it no no no we don’t care bean counting it’s the well that’s a rabbit hole that i i we we could go down maybe we’ll save is like how does someone accuse you of stealing data anyways like
Speaker 1 | 16:50.616
in this model and in this model in this ai model how would anyone ever know anyways are they just guessing they accuse you yeah they basically just say hey we think you did it and then you go through a discovery process where it’s like hey open your books show us how you came up with this And it’s the same thing I was dealing with 10 years ago. It was like, hey, we think you violated our patent. Show us how you came up with this. And yeah, we were doing patent checks. And anytime I actually ran into a patent that I was like, hey, this looks very similar to what we’re building. All of a sudden, I had to pull myself out of that project. And I couldn’t actually tell the team about it. So whenever I got pulled out, they knew I probably ran into something that if I was now in the mix, could have ran us afoul of something. Ah. Yeah. like what we’re seeing with like uh edward’s out keep going yeah and now what we’re seeing with like chat gpt uh you know these uh content uh creators that are accusing them of using their work and not crediting them or paying them and i’m like hey if you did you should because i’ve written some stuff and when somebody you know uses me as a source in like their book or paper i get like a small check i mean it’s not much but i’m like okay you know 25 bucks is 25 bucks. I’ll take it. And I’m like, if they’re making money off it, which they are, then yeah, they really should be compensating those content creators in some way.
Speaker 0 | 18:15.542
So how would someone be taking their content? Would they be feeding it just into the engine and just saying like, Hey, we want to do, I don’t know, copywriting for this marketing program for blah, blah, blah, blah, blah. And here’s the top five best copywriters in the business right now. feed all that information to the AI and pop out something that’s original, but not really original. Is that what you’re talking about?
Speaker 1 | 18:40.441
Yeah. And you know, the thing was like how these large language models work. I don’t think they’re really truly original. So let’s say you’ve got somebody like, I don’t know, we’ll just take like Noam Chomsky, write me something in Noam Chomsky style. Well, it’s going to be truly 100% original. Probably not. I mean, I’m, I would probably assume some of those sections are basically just plagiarism i want to double check to see it’s like is this truly original or is this just basically and we took a paragraph from this book a paragraph from another book and kind of stuck them together yeah okay good point why is this plagiarism right
Speaker 0 | 19:20.792
okay so to to come back to making money while avoiding plagiarism um getting maybe fake sued or real sued or just paying the bill anyways We fed the AI engine glossaries of terms from the trucking industry and freights and how to answer quotes. And now we feel that it’s somewhat like a.
Speaker 1 | 19:44.232
real bot that what do we do now what was the next step that’s where it you have to start looking beyond the the large language model that’s where i was like generative agents so i ran across this interesting paper from stanford where they were like hey look you can like create video games where you inject data and the you know these npc players aren’t so npc anymore and i was like Yeah, well, I’m thinking about how do I use that in a business environment? So I was like, imagine if I gave you an app with an actual generative agent that you could give it data, it can remember the data and could give you some alerts. So not just like for logistics themselves, but people that are interested in logistics. So there’s like a lot of retail, CPG, manufacturing executives. that do want to know, hey, what’s the spot rate today? What are the loads like? This agent could literally tell you what’s happening and give you some prescriptive analytics that go along with it, like, hey, based off what we’re seeing here, LTL rates are probably still going to stay low for the next two weeks. That’s the kind of stuff that could be really helpful for a lot of people.
Speaker 0 | 21:01.138
The other aspect of it is just-What’s the current gold spot price? Should I buy gold or not? These types of things. Right now, the financial guys are going crazy. They could never do what I do.
Speaker 1 | 21:13.965
Or, you know, you could look at fitness. Like, you know, I want, like, give me a generative agent that talks to me like the snake diet guy on YouTube.
Speaker 0 | 21:24.854
Snake diet guy. Let’s take this. This comes to this break in the show where we talk about random people that we follow on YouTube that are really cool and awesome. Today’s episode focuses on the snake diet guy.
Speaker 1 | 21:36.764
okay what is this guy i’ve never heard this guy because i’m gonna google this it’s this canadian guy that basically talks about fasting and he like starts out his videos usually yelling hey fatty just starts like horse fat shame that works yeah i mean it’s kind of funny watching him but i i also hear people saying like actually it’s kind of motivating i’m like okay if it works for you great snake diet lose 35 pounds in 15 days okay that’s it that’s above the first team that’s
Speaker 0 | 22:06.036
above the 15 pounds. Okay. This diet fam claims eating like a snake can help you lose weight. Ah, I get it. That’s the snake diet guy. So, uh, but applying that to AI, what would be cool is if we could feed the AI engine, here’s my DNA test results. Here’s my blood work. Here’s where I’ve lived in these zip codes for the last X, Y, Z number of years. I’m crediting Mark Isaacson, Mark Isaacson credit to this. He owns Village Green Apothecary and he did my DNA testing, my blood testing, and he has an API into the, oh, what is that US government industry where they map out? It’s the EPA. So he does something really, really cool. He has, do your DNA testing, find out what your genetic genome markers are and all this other stuff, do your blood work, and then map, then take what your genetic code is and map it to where you’ve lived in a zip code and all the toxins that could be in the environment where you’ve lived over the last years and then customize a diet to you. which I think is where we’re going because there is no snake diet, liver King, whatever carnivore, keto, vegetarian, whatever it is, because everyone’s body is different. Right. So it’s really not necessarily. And I learned that kind of the hard way last month being in Morocco. Cause I was always kind of like, yeah, vegetarianism, like, you know, um, you know, vegetables. No, no, it’s keto. It’s a carnivore. And like, so all I did was eat fruits and vegetables for a month. And I didn’t even really think about it and walk. And I came home and I thought the scale was broken. I lost 14 pounds. Didn’t even think about it. Anyways.
Speaker 1 | 23:32.455
But you know, that’s the thing with AI. So that’s why I’m very positive on it. I’m like, hey, look, we could go down this dystopian path where it’s like monitors every little thing we do to squeeze every ounce of productivity out of us. And everyone’s going to be like, life sucks. Or we can actually use this stuff to actually benefit people. This is where my personalization background comes in. Because I was always like, I’m not selling products. I’m buying. selling experiences that people want and the same thing with ai applies it’s like i’m not here to sell you another twinkie or ipod or whatever i’m here to sell you an experience what do you want to do i mean i worked at best buy selling electronics i was like nobody comes to our store to buy a an xbox just to have it sit in their living room and say hey look i got an xbox Some guy in his 20s is thinking about playing Call of Duty or Madden against his friends. Mom on a snow day is thinking about keeping her kids contained so they don’t tear the house apart. Buying experience. With AI, you’re buying experience as well. So if I’m a logistics manager, what am I trying to do? I want something that’s going to solve my pain points, make me look good to my boss. I’m not necessarily buying quotes. And if I’m a consumer, well, what am I doing? Am I trying to lose weight? Am I trying to have an ideal vacation? Am I looking for somewhere new to go eat or some cool new song? Or am I just looking to have like a little digital chat buddy that I text with when I’m on the bus going home? Those are all experiences. If you focus on the experience and the pain points.
Speaker 0 | 25:14.182
that you saw for people this stuff works really well but if you want to just track everybody for productivity everyone’s going to hate let’s finish the um let’s finish the logistics example so where are we going with that once we fed this off how do we feed this ex so i guess you told me like you say so we got executives they can use a um like a chat bot application to give you spot prices and stuff like that how can we generate more business if we’re a broker? How can we take it to the next level? How can we do that?
Speaker 1 | 25:45.557
It’s really about optimizing. So you’re using the agent to communicate with the customer and it’s like, hey, maybe you ought to be using this trucking company because they seem to fit what your needs are better, better delivery time, better dock etiquette, things like that. Or, hey, you’re usually shipping on Tuesday, but you could probably get better rates if you ship on Tuesday or, hey, stop pulling your loads from… the long long beach port at 3 p.m you’re going to get hosed in trucking fees because no trucker is going to want to pick up a port in la at la and have to drive out to arizona at three o’clock in the afternoon they’re going to be like yeah i’m going to be stuck in traffic for the next four hours just in la i’m going to charge you more for that so a tool to help agents possibly even sell better yeah um
Speaker 0 | 26:34.884
or grab new customers that might be with somebody else
Speaker 1 | 26:38.886
Or, you know, you go back to the logistics manager, like, you know, I’m based in the Twin Cities and we have this customer that’s also in the Twin Cities and their primary retail that they sell into is Target. Okay. Also in the Twin Cities. The thing is…
Speaker 0 | 26:54.437
How many Targets are in the Twin Cities?
Speaker 1 | 26:56.178
Oh, probably like easily 50. Okay. I think every suburb has at least one. Okay. So if I had an agent for them that helped them figure out how to communicate better with Target… because they’re they’re a small fry uh in terms of like targets uh you know who they deal with right so they need to figure out how do they get targets attention what what’s the right information to talk about when they talk to target so they can actually get targeted to pick up the phone that that’s useful information so i can get data about how how other companies interact with target and how and tell them like hey i’m a small local business trying to sell to target what’s the best way that i can interact with Oh, this is how, because if you’re like General Mills, yeah, that’s a very different relationship. You’ve probably got like a direct line connection to the merchant.
Speaker 0 | 27:44.339
You’ve just got me thinking, I’m literally, yes. What about return on investment in all of this AI stuff? Are people not just seeing that, is there just no numbers to crunch yet or where’s there, just not seeing it or what?
Speaker 1 | 27:58.749
Yeah, I mean, that’s the issue I ran into. It was like, oh, well, I read it takes like hundreds of thousands to set up an LLM and I’m like. No, it took us two grand.
Speaker 0 | 28:08.116
I’ll pay you two grand plus an extra 4,000 to bill it for. We just got to figure out something. What do we want to do for dissecting popular IT nerds? Let’s just do this as a test for dissecting popular IT nerds. What tool could we do for IT directors? What can dissecting popular IT nerds invest into the marketplace as a tool for IT directors? What could we do for them? Leadership. How to speak to my… How do I talk to my CEO and sell him this forklift that’s, I don’t know, $100,000 over budget? Let’s see what else we would do. What top security products should I be using right now? Or who’s getting hacked the most? I mean, the possibilities are probably endless.
Speaker 1 | 28:46.607
Yeah, I mean, like, if I take, like, engineering groups that I’ve dealt with, you know, they usually come with metrics like, hey, our uptime is this. And this is how…
Speaker 0 | 28:55.516
KPIs, yeah. This is how…
Speaker 1 | 28:57.538
how many story points we dedicated to new projects. Here’s what we dedicated to bugs. And it’s like, okay, that’s somewhat interesting, but can you bottom line it for me? I don’t really understand, like, what’s the connection here to revenue and profits. And that’s, I don’t know. I got this from Datadog. This is what it told me. And, you know, a great example, like when I was working at Target, so I was… I was working for this guy who liked to be very technical. So we’re going in and talk to the SVP of marketing. And he’s like, I’m going to handle it. We walk in there, the guy’s like, You got 15 minutes. My boss is like, yeah, so we’re building this with Scala and we’re building this other thing with Python and we’re using Agile on this project. And the guy’s just sitting there like, yeah, yeah, looking at his watch and I just go. So for this project, we’re going to take the conversion rate, which is typically a 2% and bring it up to 5%. And for this other one, that’s typically a 3%, we’re taking it up to 7%. He goes, hold on. Calls in a bunch of other guys. That 15-minute meeting went for an hour and a half. Now I’m talking what he wants to hear. Conversion, average order basket, time on the site, how much of our marketing spend is going towards this. He didn’t care what programming language you were using. It was like nothing to him. He wanted to see dollars and cents.
Speaker 0 | 30:20.000
Golden. Golden advice for everyone out there listening. Absolutely golden advice. How do you make someone that speaks like the data nerd that’s really excited about himself and how smart he is? talk about all these programs and python everything speak like what you just did oh that’s like that’s like uh yeah please pull that book off the shelf it’s uh and this is an audio show so what are we looking at here alchemy to me by rory sutherland who’s the uh
Speaker 1 | 30:49.414
vice chair of ogilvy okay he’s got a bunch of videos on youtube but essentially this guy is like hey people make emotional decisions and then look for a rational excuse and i’ve always been like yeah that’s old school that’s real old school that’s even back to zig zagler and all that yeah go ahead or yeah you know and it’s true and he’s talking about like hey you know uber didn’t didn’t kill taxis because it was digital you could order a taxi before uber came along on but what uber did was they removed uncertainty so i could see the car the driver i got an eta and i see the little map just literally walk out of my hotel and there i am the uncertainty removal is really the thing and that’s that’s the same thing that sometimes you have to remove the uncertainty for people and just speak their language uh but that doesn’t always work because again people are irrational and you know sometimes they’ll just be like no i need a short bet it’s like well then you’re in the wrong business go work in government i guess go work at the dmv remove the uncertainty remove the uncertainty that’s uh
Speaker 0 | 31:52.824
that’s the highlight that’s the highlight of this entire conversation for any IT director out there listening. How do we remove that? Then go ahead, remove the uncertainty for chat GPT. That’s what we have to do then. I want to throw that back in you. The irony of this whole thing is remove the uncertainty for chat GPT.
Speaker 1 | 32:08.551
It’s the uncertainty in the customer experience, but as you’re building these things, like as I was just going through and I was just like, hey, you know, they’re like, this has risk. I’m like, hey, getting up in the morning has risk. I’m sorry, but that’s life. I don’t mind risk. I’m comfortable with risk. because I understand that’s how innovation happens. And I just look at this space and I’m saying like, I go back and I look at the big data, data science time. So I was building out the data science team at Best Buy back in 2011. And I can tell you everybody, and I was in what was called emerging technologies. Everybody in IT told me I was nuts. Data science is a fad. A year from now, no one will be talking about it. I’m stupid for putting money into it. I was like, no, you guys don’t get what’s coming. There are certain trends you can see where it’s like everything’s data. Yeah. And I was just like, they weren’t collecting any unstructured data. And I’m like, I’m going after all that. And everyone was just like, that’s worthless. Nobody, nobody uses that. Everything’s relational databases. I was like, nah, you don’t get what’s coming guys. The world’s about to change. And yeah, sure enough, we turned that into like, so this was all around personalization, which we were using a vendor at the time and making 25 million. uh, turn that into a billion in revenue.
Speaker 0 | 33:23.658
How’d you do that?
Speaker 1 | 33:24.618
Well, the nice thing was the timing. Uh, I worked there during the year of the three CEOs. So nobody was really paying attention to what I was doing. Just like you’re nuts. You’re stupid. It’ll never work. Okay. Then leave me alone. And they did. So we, all these algorithms for online in store, best buy for business, the warehouses, uh, you know, anything we’d get our hands on. And we just started even the call centers. And we just started figuring out like, hey, how do we measure the revenue impact? And so we were doing like average order volumes were going up by like $25 using just like basic collaborative filtering recommenders.
Speaker 0 | 34:07.283
Ah, recommendations.
Speaker 1 | 34:10.524
Yeah. True one-to-one personalization. You were going from a 1% conversion rate to a 15% to 17% conversion.
Speaker 0 | 34:18.289
Now I get it. That is so powerful. That is so powerful.
Speaker 1 | 34:23.842
Yeah.
Speaker 0 | 34:24.344
Other people like you. who bought this also bought this. It’s like, what?
Speaker 1 | 34:31.998
I want it. The thing was the time on site metric was insane. So you had like people doing like five, 10 minutes without personalization. Then you just throw in these collaborative filtering recommenders, which really I don’t even call them like personalization because they’re technically not, but people would spend like 40 minutes on the site. Just keep clicking on all these different like recommended products.
Speaker 0 | 34:54.536
Yo, that is, that’s.
Speaker 1 | 34:56.458
Then the call center, this was mind-blowing. So I’m sitting in the call center. We’re testing this recommender for the call center. So somebody’s like, hey, I saw this laptop like six months ago, but now it’s not there. You got something similar? They give all the specs, and we recommend another product for them. Now, we actually took this from Geek Squad. So we created one for Geek Squad where like a part, we’d recommend a like-for-like part. And then it’s like, hey, wait a minute. Let’s just take that same thing, tweak it, put it in the call center. So that normal call without the recommender, 40 minutes. Okay. With the recommender, 15, 20 minutes. And half of that was just like chit-chatting. So that person went from having to sit on one call for 40 minutes, and now they could do two calls in the same amount of time.
Speaker 0 | 35:41.980
I don’t get it. You got to explain that again. How did we cut the time? How did we cut the call in half because of a recommendation? Is that because they didn’t know what to recommend the customer? Is that what it was? And they’re searching for them?
Speaker 1 | 35:53.888
They’re constantly like, well,
Speaker 0 | 35:55.369
is this people? Like.
Speaker 1 | 35:56.149
I get what you want. Yeah. So the recommender we gave, it was spot on. And they were just like, yeah, this saves us so much time. Now these guys are hourly. So the fact that they can now churn through more customers.
Speaker 0 | 36:08.278
You just smoked it. That was, that’s a home. That’s a, that’s a, that’s a grand slam.
Speaker 1 | 36:14.623
Yeah. And then best buy for business. You got like, you know, business accounts. And again, you got like your traditional B2B salespeople. They’re calling up like, Hey, you’d like, you need any supplies and stuff. It’s look at their purchasing history look at like a grassroots thing we could do like a grassroots thing here like how do we how do we support the small business owner to like be super nimble and around and jump around all these other big guys yeah and the thing is like the cost has come down dramatically i mean back when i built it so my first year operational run that includes like labor as well 3.3 million which for a fortune 500 that was actually cheap now i mean you could do that for like you know a couple thousand
Speaker 0 | 36:53.574
Oh, man. Everyone listening, Edward is up for grabs. I’m pretty sure his base salary is like $500,000 plus 20% of business growth. Huge. Just the tip of the iceberg of why AI is exciting. Forget about the fact that the, you know, deep fakes and all that other stuff. Go ahead. Deep fake Phil Howard. I love it. You know what I mean? I already told my wife, honey, you’re going to see a video of me with another woman very soon. And it’s not me. I just want to let you know ahead of time. It’s going to look fake now, but later on, it’s going to look real. Yeah,
Speaker 1 | 37:38.307
I tried that with Mid Journey. That didn’t go over well with the wife.
Speaker 0 | 37:43.932
Yeah. The. There’s going to have to be some security aspects around that. That’s a whole other thing that’s going to have to… I was thinking, why don’t we have some kind of weird digital token there? You have to provide a blood sample to your computer to verify that this was created by you or something like that. Probably going to get that far.
Speaker 1 | 38:02.458
The legal and ethical side of things is going to be very interesting. But that’s where I’m still like, hey, I focus on the positive aspects. And I think, how do we actually use this to help people? you know better experiences uh and this is something that it’s interesting when i talk to like 20-somethings like you know some of the people on my team like i’ve got this great analyst alana she’s just like yeah you know the the older generations this grind it out at work for like 80 hours which goes people my age don’t want to do that no they’re ruined they’re
Speaker 0 | 38:35.535
all ruined lazy bums should they be doing that you know we know i’m not
Speaker 1 | 38:43.886
We have these philosophical, on my team with my engineers, data scientists, analysts, we always have these philosophical conversations. And it’s great because even like 10 years ago, we were having these conversations like, hey, we’re building these recommenders. Like, we’re building the right thing for the right reasons. And it’s like, hey, would you be okay with this thing being used on you? If the answer is no, then we probably should stop.
Speaker 0 | 39:07.667
It’s just a recommendation. Would you like to supersize it? I mean, no.
Speaker 1 | 39:14.671
At one point, I had another team ask me, it’s like, hey, can you grab like the Mac address and everything associated with it? I can, but I’m not going to.
Speaker 0 | 39:22.115
Yeah, yeah, yeah, yeah. It’s like that article that came out of on Zoom communications the other day when they updated their terms of service. It was like, by clicking here or something like you agree that we can use like, you know, camera feeds, data, everything about you and feed it into an AI engine. And, you know, it was like, you know.
Speaker 1 | 39:40.846
very very wrong well and that’s where i think it’s going to get interesting because i and they’re recording this right now i think people really need to uh push back and say like hey it’s actually my data i’m letting you use to help me get something out of it and i’ll pay you if you give me a good service but it’s still my data that’s
Speaker 0 | 40:00.531
kind of cool yeah how can we personalize our data how can we take back the power and put a price tag on every human being you um and attach it to theirs how can we put a price tag on every human being and attach it to their social security number and can we go down that uh um uh what we usually have this like you know uh conspiracy theory part of the show and i have a friend that’s like you know what they use your your social security you’re really owned by the united states and you know all this type of stuff doc
Speaker 1 | 40:28.274
searles uh he was one of the clue train manifesto authors he came out with a book uh called the intention economy and there’s a more technical book, but written by a professor from University of Utah called The Live Web. So the two of them are really kind of companion books. But in there, Doc was talking about a personal cloud concept. And like around 2013, it got a little bit of traction, but really fizzled out. And I was just like, hey, this might be a really good idea in retail, but I don’t see any of the retailers would actually build this because they all don’t trust each other to share the data so it probably has to be somebody outside of retail essentially you have your own data locked up in this personal cloud and you can say hey i want to lose some weight what do i need to lose some weight like here’s like what i what i’m thinking of doing it’s like oh you probably need some shoes you’re gonna need some gym clothes water bottle all right go to go out yeah go here yeah go out there get me some bids on this stuff isn’t that what amazon is well it would go to like amazon target Walmart and like see like hey who’s got the best price and i was like it’s a good idea but i just i always saw like the retails will probably really push back on this well it’s a race as they like to say it’s a race it would become a race to the bottom but that’s kind of already going on yeah the thing that i was i liked about it was it’s going to protect the the data of the individual a bit more because i do think companies really do sometimes take a little bit too much advantage of that
Speaker 0 | 41:58.962
This has been very inspiring for me. I’m definitely thinking of how to selfishly use this to my own advantage and take advantage of all the data out there and then make many suggestions, many suggested buys to people somehow, somewhere. This is great.
Speaker 1 | 42:13.271
Well, there’s lots to do. Like I said, if we look at it from a positive perspective, I think AI, especially merged with personalization, would be absolutely amazing. The funny thing is in the B2B world, they do very little personalization. But it’s actually, it’s easier because you often know the intent of the customer. Whereas in the consumer space, somebody walks into a target and they walk out. You have no idea why they even got, went in there in the first place.
Speaker 0 | 42:41.341
It’s so sad. It’s so sad. It’s true. The, and I kind of just throw my hands up now. I was like, what do you expect? I mean, these people are like hourly workers. What do they care? Like, you know, I, I actually expect people to be, if I run into an employee. any retail established now that’s actually enthusiastic and personal and everything. I’m like, this person’s going somewhere.
Speaker 1 | 43:01.395
Well, when I worked at Target, the interesting thing was they didn’t actually know who was in the store. They only registered you as a customer when you actually went to the register. So there was no light beam tracking how many people were in the store.
Speaker 0 | 43:16.549
Yeah.
Speaker 1 | 43:17.469
You go to a bed, you’ll see the employees walk through the exit because there isn’t a light bar there. There is on the entrance. They’re tracking like how many people come in.
Speaker 0 | 43:26.613
Yeah. It’s so in the B2B space, it is very kind of impersonal. Like, so there is a lot to be there’s a lot of low hanging fruit is what you’re saying. There’s a lot to be done. And you are the guy to get that done to get that done. Edward Shannard, thank you very much for being on the show today. It has been a pleasure. and it’s certainly got my mind thinking about many different things that probably we should be doing here on dissecting popular IT nerds and it’s it’s exciting I think it’s not something to be feared but then again that’s coming from you know I guess an ex marketing guy data science guy versus a you know the guy that’s building the machine out there and probably worried that it’s gonna
Speaker 1 | 44:12.544
kill everybody in the future well probably the uh like i i’ve i’ve because i’ve heard people say well ai is going to be more efficient i was like well if you yeah i mean but if you tell it to do something stupid it’s going to do it efficiently but it’s still going to do something stupid yeah yeah yeah well thank you sir um any final words of wisdom uh yeah just stay positive on it i mean i think we’re at the start of something that’s really exciting and it’s the time to shape it the way you want it to be don’t let somebody else tell you how it’s going to be
Speaker 0 | 44:39.900
Yeah. And with that being said, look forward to, I am going to have my data science PhD friend on the show in the next couple of weeks. And he is a fully 110% doom and gloom. We need to do everything we can to protect the world on this. So look forward to that showing that show coming up. Thank you everyone for listening to Dissecting Popular IT Nerds.
