Ask an Expert Ep.6 - A Deep Dive into AI with Steve Schlenker (DN Capital Founding Partner)

AI is revolutionising the world at such a rate that it could well be the most significant technological advancement since the inception of the internet. But how? And what can businesses do to harness its power?

We had the pleasure of welcoming DN Capital’s Founding Partner, Steve Schlenker, back to our Ask an Expert podcast to ask him those very questions. Due to his involvement in several AI-centric enterprises, Steve has a profound understanding of AI and how it can help businesses across almost every area of their operations. 

For him, there is no question that business leaders must integrate AI as soon as they possibly can or else submit to being left behind. Tune in now to see how AI can help your business become more efficient, effective, and ultimately, more successful.

Transcript

Steve: 0:00

the advancements that are taking place right now to use machines to solve problems is not being done in individual silos. It's being done collectively, and so the rate of growth is happening much faster AI makes it more affordable to customize your offering it can do more accurate testing of consumer response It can predict what will happen with different materials or different shapes, and they will do it in a way that's better we've got rising interest rates, we've got rising unemployment. How do you break that cycle? You break it with productivity. What's the number one productivity tool today, what we're calling ai,

Jamie: 0:55

Hello and welcome to Togethers Three Things podcast, and my third. Ask the expert with my friend Steve Lanker, founding partner of DN Capital, it's our third catch up, Steve. I'm almost tempted to create this as a new series called Ask Steve because I'm coming to you whenever I have, I. A question I don't know the answer to, and you are very generously giving us your time. So thank you and welcome back.

Steve: 1:23

It's a pleasure and I have to congratulate you on all the things going on in your life. So congratulations and thank you for having me back.

Jamie: 1:30

I mean, the congratulations is not really. Worth it, but thank you. What he, what, what Steve is referencing is that I have added on top of a, an already quite full portfolio, an interim c e o role for a few months back at my old haunt of Tom Dixon, where we are sat today. And we might use Tom Dixon as an example for this conversation because today I am following up the conversation that Steve and I had on his roof. Couple of months ago, three months ago, two months ago. And, and, and that was a conversation, for reasons that weren't obvious to me at the time, but had become blindingly obvious. One of the three things you said, of the pieces of advice you gave was that, For the scale ups and small businesses trying to navigate these difficult times because they still are difficult, it's time to leverage ai. And whether it's cost reduction, revenue generation, innovation, and creativity. You were kind of telling me that businesses can move the dial. I think we, you were, were implying from, from where they are today without adding heads per se, but leveraging ai and I said to you at the time, I think I want to do a follow up. That's what this is.

Steve: 2:46

Thank you. I'm pleased to be here. And let me just add that I'm not going to be talking about. The house not gonna be talking about how does one link to the APIs from a large language model or any of that.

Jamie: 3:00

No one can see my face, but my face is blank. I have no idea what you're talking about

Steve: 3:05

good, because we're not going to talk about

Jamie: 3:07

Phew.

Steve: 3:08

but we will talk about are some of the technologies that I'm seeing startups pitch to average everyday companies where those companies can use. Those applications that those startups are pitching to make their businesses run better. And as you say, help navigate these difficult times, make their businesses more profitable, make their people more productive, drive more revenue, drive better efficiency.

Jamie: 3:33

So where should we start, you know, to, to make this a, we don't want this to be a too long a conversation if we're gonna structure this. Where do you wanna start, do you think?

Steve: 3:40

let me start by giving an example of one of my companies that I have today that uses technology that would possibly be even relevant for this conversation.

Jamie: 3:49

Okay.

Steve: 3:50

So I have a company that's used by call centers in places like the Philippines and Its main use case is it uses artificial intelligence to predict what is the next syllable you are about to say, and in real time, sub 200 millisecond latency. So effectively faster than you'd be able to recognize. It removes accents from spoken speech, like my lovely American accent could be made to sound like a proper British accent But the reason I mentioned this is because what can it can also do, in addition to softening an accent, it can remove background noise so that you hear just my voice, or Jamie, just your voice. Not the hum of the machines in the

Jamie: 4:33

background. You're assuming that I don't already use an AI engine to remove all of that information anyway, which I very, I used to do it with them.

Steve: 4:40

Very possible. Although you're doing that for recording and this is doing it in real time as we're

Jamie: 4:44

Yeah, How should I be using this?

Steve: 4:47

So first, if you have already decided you want to outsource your customer service to save on costs, yes, this will increase your customer satisfaction while reducing your call duration, because there's less misunderstanding between the agent. The customer on the other end. In addition, the agent's job satisfaction goes up because they don't have people on the other end of the phone saying, please repeat that. Therefore, the training goes down because the employees stay on longer, which reduces the costs again for you on the call center. And the same technology can be used outside of call centers if your business is in, say, telemedicine, or if it's in online education. So that anywhere where there could be the possibility of less understanding. Over a digital interface or over a telephone, you can improve the understanding and therefore improve your ability to get business so in addition to technology that can be used to enhance the agent, there's technology to replace the agent, which would be called a chatbots. Yes. Or these days. They use the euphemism, cognitive ai.

Jamie: 5:52

hate chatbots,

Steve: 5:53

Yeah,

Jamie: 5:54

but are they better now because I'm still fine. The first, line of defense of customer service everywhere now is a rubbish chatbot.

Steve: 6:01

So for anything that's technically complex, I would rather talk to a human personally. However, there are ways now to use. Natural language processing to link together different systems in your business where it actually makes sense. So for example, if I call up a agent and I say, I need to change my appointment from Tuesday to Wednesday, how much additional money will that cost me? That actually is a quite complex question. You have to link together a payment system as well as a scheduling system. That's many ways handled better by. Natural language processing and an AI engine linking together two backend systems. Then human trying to search through a bunch of different spreadsheets

Jamie: 6:45

And, and I'll tell you, an internal customer service equivalent, which is, I interviewed the c e o of pizza pilgrim's. Fabulous, fabulous Pizzas. and they were, I dunno if he's rolled it out yet, but they were building. or they were feeding the data in any way an an internal, chatbot, AI power chatbot to answer all of their hundreds and hundreds, if not now, thousands of employees. Questions that's in an employee handbook and several other databases and, and locations of this process and that process. And they're just feeding everything into this, chat bot in the hope that the, the employee would just open up the chat and say, how do I book my holiday again? What's the process for, refunding a customer, you know, every piece of knowledge available through that, was his plan. And I just thought, well, that's pretty damn obvious.

Steve: 7:32

Well, again, the difference between AI technology and just putting in really large databases of, ways to do something is the AI engines have so much complexity that they're able to solve problems that can't easily be outlined in if then statements that one could just look at. In a f a Q list. Let's move off call centers for a second and off chatbots. Let's talk about if you are a, if you're a manufacturing retailer of let's say high-end luxury

Jamie: 8:02

goods. oh, I can't think of any, but, alright.

Steve: 8:05

And let's say you're trying to sort out logistics problems, whether it's inbound logistics or whether it's outbound logistics. you can use AI technology to try to predict. Which drivers are most efficient for which routes? Which routes are likely to have the most problems based on events going around, whether it's Wimbledon, whether it's weather patterns, whether it's traffic strikes or otherwise, so that you can better get products to your customers or make sure that products are in your warehouse on time.

Jamie: 8:36

Now, assuming said, business, outsource that part of their logistics to a third party. How my AI still. Leverage AI to improve the

Steve: 8:47

world. You could

Jamie: 8:48

run other than tell

Steve: 8:48

to do, You could run a digital twin, which is to say you could use a electronic copycat of your outsource partner and then confirm whether or not they're using best practice or whether they're cutting corners, whether they're drivers are perhaps taking extra time off, you know, maybe to see a significant other on their routes so that the route shows up late. All these things that would be hard for you to staff up to predict you could run

Jamie: 9:16

And, there are businesses that. I'll set up to provide this service today for me to use. I assume we didn't name your, portfolio, company that does the voice, tweaking, by the

Steve: 9:26

way. Oh, a company is called s

Jamie: 9:28

and, and, at every time I will be B b, C about it. there may or may not be other companies that can do that too, but we don't, we don't need to be. Fully listing. And in the world of logistics planning and, and this kind of analysis type of work, is this, is this a logistics specialty AI business, or are you making it up as you go

Steve: 9:48

on? there are some companies in the US that I'm familiar with. the one that I'm thinking of most is a company called Relink that was set up by the former head of logistics for

Jamie: 9:58

And this is, this is what's happening everywhere, I assume. Right? Which is experts in their verticals are who have witnessed. Early attempts to use AI in its rawest form in probably big corporates are building these tools. we're not gonna talk about the, uh, the, the investability of those businesses, But Let's just look at the categories then that, you know, you, you, you would put in your top three and we'll make this the equivalent of your three things, right? Maybe this is one way of doing. So. If you were to generalize across as many business as possible, the three areas that they should apply existing soft AI software to what would they be? Let's just list them.

Steve: 10:35

first would be around AB testing for customers.

Jamie: 10:39

Mm-hmm.

Steve: 10:40

So I'll give an example. I'm not gonna name the company because we're actually in due diligence to potentially invest it's shown Digital advertising that has video attached to it has a higher conversion rate than static

Jamie: 10:53

Okay.

Steve: 10:53

But the cost of doing video advertising, if you're a retailer of thousands of SKUs is very You can create through the equivalent of a lower end video for one 1000th of the cost. Of having an actual proper model record that same video for a product, so not for your broader audience, where you may still wanna spend a lot of money, but for AB testing in select audiences where you say, let's see whether or not this product plays better with a male actor or a female actor with an actor with a dog next to'em, an actor with the family next to them, an actor sitting on a beach, an actor in a suit and tie. All of those can be done for costs in the range of 10 to$20 a video as opposed to thousands of dollars per

Jamie: 11:47

And let's take this back to simplicity for a second, because step step one here is, if you're in fashion, let's say, and you have to have a photograph of every single skew, the more skews you get, the more expensive that is Now you premium end, you're gonna be investing in it. in a crunch time, maybe some of your, some of the skews are gonna get an AI model instead, which I have seen in many businesses already.'cause you can just about tell that they're not, they're not the human. so there's a basic thing you could do, which is don't have the shoot, just have AI generate the models. but you are now talking about.

Steve: 12:17

What

Jamie: 12:18

more that gives you, which is, if you start getting good at that and you know how to use it, you can start doing proper AB testing. And I, I think that's extremely powerful and makes

Steve: 12:25

makes it affordable to do multiple variant tests to lower your cost of customer acquisition. Not only to lower your cost of customer acquisition, you can use the same technology for retargeting your customers. So increases your lifetime value. And after all, what is the value of a business? L T V over cac? Increase your lifetime value of your customer compared to what costs to

Jamie: 12:45

So digital assets for a. Any kind, frankly, doesn't have to be a d c business, but probably a consumer business we're talking about here that usually involve models and photo shoots are very expensive. They can be done using ai and that leverages all sorts of options to double down your investment. Without much cost into AB testing and they're like, I love it. Number one, done.

Steve: 13:08

And, and number two, in a somewhat related space, what we didn't talk about last time. Using those similar types of avatars for internal training or compliance or education for your staff. So video education Tends to have a higher recall rates. As does compliance manuals done through video as opposed through text,

Jamie: 13:29

Oh, a hundred percent, right. No one likes to read of a

Steve: 13:32

manual. And similarly, if you are the head of HR or if you're the c E o and you want to be the one seen to be caring about your employees, so you want to teach them about safety or training or compliance, you really don't have time as a C E O or an interim c e o to record long. Education videos or compliance manuals, but you can have enough time to create an avatar of yourself. And have that avatar through natural language processing and facial movements and voice recognition. Basically record the entire video for you so you don't have to, including thousands of education videos. They're customized based on your employees. Personal receptivity. So you have certain types of employees who respond better to a very, very open and relaxed Jamie. And you have certain employees who might want a very domineering and very specific, Jamie, that's a nice

Jamie: 14:26

Listen, love it. But I still am gonna group avatar based AI solutions, whether it's for AB testing on your website or if it's internal training manuals, or, or any such thing. The power of a video, or even a good photograph is now pennies away. That's done. We can't do any more around that category. So where are we going next on our AI adventure, Steve?

Steve: 14:53

So, for pure digital businesses, the ability to inspect, rewrite, and augment your code base is done much better through AI than through

Jamie: 15:06

Inspect, rewrite and augment. Audit. Sorry, audit.

Steve: 15:10

for example, If you have a digital website, let's say you are a retailer that's mostly offline, but wants to have people purchase products occasionally off their that website might run faster. If it was written in a modern language, but it was written three, four years ago in a language that's not as performant, the AI can rewrite your entire code base for you in the new language.

Jamie: 15:36

I have a website. And I can get ai, I can't, but people who know what they're doing can get AI to rewrite the, almost the entire website into a, into a new language.

Steve: 15:46

almost the entire website, into a more performant language. They can also do test and automation when you make changes. So let's say the sales and marketing organization says, wouldn't it be great? If we change the way people do their checkouts, we have a new way and we think it's gonna be more efficient. But your main concern as a retailer is what if it breaks my payment flow and people can't actually pay money to me? So there are thousands of checks that you'd have to do. Can it accept visa? Can it accept MasterCard? Can it, be something that works off of a mobile phone as well as off of a desktop? You know, all those tests are really difficult to do, but an AI engine can build all those tests for you automated so that your people on the team don't have to do it, and the test can be run

Jamie: 16:30

if you are running a digital business, I'm hoping you are already playing with AI to support your software engineers because. It sounds like a, a must in businesses debate. I don't know if that's the truth or not. This is actually as relevant, I think for those, those businesses who are not digital businesses, but have a digital platform when looking for agency support and or external support. If there isn't a, a very strong AI elements to pitches today on, and, and as a result, I hope a cost reduction, walk away and go to someone else.

Steve: 17:00

and there are technologies that even three, four years ago were good enough without ai, but will be even better with ai. There's a company I sold a year and a half ago called YADA, and they work mostly with offline retailers who are trying to move

Jamie: 17:14

Sure.

Steve: 17:15

And in this particular case, If you have a website and you're trying with an outsource team to replicate the fact that you're not a digital first business, you start putting in a lot of bells and whistles. You put in things like click to call, you, put in place, video customer support. You put in place Google analytics. Some of these things are put in a way that it slows down the customer experience. As you know, if you take 6, 7, 8 seconds for a website to load the customer leaves, so our company, yada, that we sold it throttles all of the different applications, so it gives the best experience possible while you're loading. A photograph of, say a sofa in the background that's running your analytics or it's running whatever you want to do next. So that from the customer's experience, they don't see it, they don't see the extra delay. All of that now is being done with AI in a way that's more efficient. It could predict what are your likely slowdowns, not just based on how you structured website, but are there any congestions on the network? Are there any problems at your partner's

Jamie: 18:21

let me stop you for a second because I, I think. You started this by saying For digital business, and I want to take that away because if you're a digital business, you should be all over AI by now and understand how you're I want to talk about everyone who's not digital first, People who are know and are excited by ai, Let's talk about their website. Their website costs them a lot of money. Building it, amending it, changing it. Maybe they have invested in a small team, but they're probably outsourcing quite a lot of stuff when they Or maybe they're coming up to do a big rewrite and a big, a big new, refresh if you're looking for an outsourced agency type support. So you're gonna really come into my world, How am I gonna really know? Whether the agencies I'm talking to get this thing or not, because everybody's gonna be talking about it. Everybody's gonna say, oh, I do AI this and I do AI that, and how am I gonna test whether or not someone who's coming in to work for me other than the price they give me?

Steve: 19:14

you want them to be able to give you measurable tests afterwards to be able to demonstrate what did it look like before versus after. If you're looking for cost savings, then the output you're getting in terms of, for example, page load time should be at least as good as it was before at a lower price. Or the number of customers you can handle on any one time without needing more capacity from your cloud provider should be going up rather than down

Jamie: 19:42

So get to the, the key KPIs that matter in this. Piece of work. Yeah. And get them to show you what it would've been if they hadn't been using AI from a cost and benefit point of Avatars was number one. And everything we can do with it now. Software, coding and, and, and software engineering. Another area, this is the third area.

Steve: 20:00

area. So HR management inside of an organization, So predicting which employees work best together on different Predicting which employees will progress in the organization at the right level so that you can choose who to give a promotion to in a competitive predicting how to cut your cost in terms of HR benefits based on the likelihood of those benefits being used or not used. So different medical benefits or otherwise. All of those can be done

Jamie: 20:30

by but I'm, probably needs the size of data set that bigger companies are gonna have to be effective, right? If we're talking sort of smaller scale up businesses with employees in the tens to hundreds, is it really gonna do much

Steve: 20:45

for them? So here's the thing, because it's not core to your business, but it's core to the technology providers business. They gather data sets that include the large companies, and as a small company, you get to benefit

Jamie: 20:56

Okay, I can get that.

Steve: 20:57

so they can take very large data sets. For example, let's say one of the benefits that you have as an employer is that you have self-insurance for healthcare benefits, and you're trying to put in place wellness campaigns for your employees. If people go to the gym three times a week, then you give them an extra 50 bucks towards, Whatever it would be there. Health, food plans, the yoga plans, because it's gonna reduce your self-insurance costs by a hundred dollars. Well, the prediction of what you should give, what programs you could do. You have too small of a data set, but when it's aggregated with employers that have a hundred thousand, 200,000 employees, you could learn from their lessons and it'll apply to your employees. Maybe, I'm

Jamie: 21:42

I'm not sure. I'm gonna let this pass. So here's the thing, right? If I think AI is gonna be. and somehow responsible in, in businesses like this one for improving retention in my I'm missing the point, right? I got, I've got some basics to get right in my business as a leader. That's nothing to do with the technology that sits behind HR administration. I've gotta have a clarity of, of ambition and purpose in this business. I've gotta have. element of, of obvious care. now we get into the world of I've gotta have people empowered to do their jobs well and supported to do it. And, you know, I might get to benefits honestly, but frankly the benefits even the pay doesn't matter as much as the basics before that, how is AI gonna help me be a better c e o, a better leader when it comes to culture and is there something you've seen out there? Because that would really be interesting to me. Culture is so difficult, huh? But

Steve: 22:37

there are, there are a lot of tool sets that can be used. If you are a business with sales organization, like a, a B two B sales organization yeah, You can empower the salespeople using technology so that they don't have to come back to supervisors to make So for example, if I'm doing, pitches and I have like an R F P I'm responding

Jamie: 22:58

to Yep.

Steve: 22:58

I can use AI to mine all of the previous RFPs I've ever responded to in the past and find the best results, they're likely to be

Jamie: 23:05

yep. And I've seen some of this stuff. I get that. So we can improve our, our sales function in their RP writing.

Steve: 23:11

I can use AI to look at how I respond and look at how my customers response. So effectively I can audit the customer sentiment in my phone, conversations with the customer in my email conversations,

Jamie: 23:24

whoa, whoa, whoa, slow down here. I love a bit of sentiment analysis, but we've got now software in my phone listening to my conversations and reading my emails, coming back to me saying, your customer meant this when they said X, y, z. Is this, is this, are you making this stuff up? Or is it, it exists today?

Steve: 23:42

Your customers breathing pattern changed, which is consistent with them being more

Jamie: 23:48

this, this type of software. Is there any kind of, any brands you, any companies you want to, you

Steve: 23:52

want to

Jamie: 23:52

know? It's all right because, I mean, but I do, I'm gonna turn this somehow and we gotta turn this into practicality for some people at some point, but that's, you know, another piece of work. I really like that idea that, Because I've always been into sentiment analysis for brands and brand monitoring and, and I know natural language AI makes this a million times more viable as a, as a thing. So I love the idea of seeing that real time in, in my conversations.'cause how many people have the EQ to properly interpret what the person opposite them is

Steve: 24:23

More to the point. what AI's able to do is inside your organization, they could say, what type of pitch works best for your company? What type works best by specific types of customers in your company? What works best for specific types of salespeople in your company? And it can help craft the stories that are more customized. So you no longer are just doing sentiment analysis based off of, say, a high level theory where every company is treated the same way and every response sounds the same way, but it's very, very specialized to your So

Jamie: 24:56

I'm, gonna, I'm, I'm allowing this one in as your third area, because I think, I think enhancing your sales team, I know there are lots of tools out there. and if we are not. Even interested in exploring how AI could enhance our sales function and our business, we are going to struggle. and that's, that's true for, for, for everybody. I know it's called three things, but I feel there's a fourth and a fifth in you yet, what we've covered here is. Software engineering, coding. that's been amazing. Yes. We've been talking internal and external customer communication and management in terms of the A avatars. And he avatars. Why do I keep saying a avatars? Avatars

Steve: 25:33

and

Jamie: 25:33

kind of, saving one can make in creative, in

Steve: 25:37

creative so so you'll note, you'll note that I've been intentionally avoiding product design and product enhancements, and

Jamie: 25:43

didn't notice that because we've got lots of areas we haven't touched on yet, but okay. Let's talk about product design and enhancement.

Steve: 25:49

the reason I've been avoiding is because it's so bespoke by company and by industry,

Jamie: 25:54

Let's talk innovation and creativity generally. Can AI. So let, let's be clear, right, and let's use Tom Dixon as an example. We're a product business. There's lots of product businesses out there, or retailers of products or, even if it's services, businesses that create new products and, and services. We're often innovating. Everybody's innovating. How should AI fit into our innovation

Steve: 26:14

The first thing is where there are complex technical issues that you need to resolve in your product design. AI does a very, a very good

Jamie: 26:22

role Have you seen the worldwide regulations on lighting? There's a lot of them. It's complicated.

Steve: 26:29

I have not seen the worldwide

Jamie: 26:30

it wouldn't surprise me if you actually answered, yes, I have.

Steve: 26:33

but I, I have seen things around protein folding. I've seen things around battery design and choice of new materials to make batteries more efficient.

Jamie: 26:43

Okay, so AI's gonna do what for me here.

Steve: 26:45

in those cases, or particularly in

Jamie: 26:47

s you were generalized before, so let's sort of keep it general, but how is AI gonna help in the design element of those kind of

Steve: 26:54

things? So where it's too many variables for one human to keep in their head about what things interact with,

Jamie: 27:03

each.

Steve: 27:04

If you set a model of things you already know, in this case here is how lighting works with certain types of existing glass, certain types of existing shapes. It can predict what will happen with different materials or different shapes, and they will do it in a way that's better

Jamie: 27:25

So what complex data set

Steve: 27:27

Complex. Complex data set that you already have fed into an engine to predict what next product design you might want to test out.

Jamie: 27:36

it's a bit hard to put that into the world of consumer products, I think. which, which feel, it feels to me like there's a lot more of, what we would normal normally refer to as creativity in terms of someone just come up with something. We like it.

Steve: 27:48

it, Right? So, So, let me, try some other examples where it does fit with consumer design. But these are not startups. These are large companies using

Jamie: 27:57

Sure. Well, let's learn from them.

Steve: 27:58

all right, so there's a Chinese EV electric vehicle That the manufacturers realized that because you don't have to cool a, cataly converter based

Jamie: 28:13

Sure.

Steve: 28:13

you can change your front grill so that it doesn't have the same type of airflow. So they're able to create a different design that was more attractive in theory, but also more, gas efficient than the types of grills that you would have on a petrol Okay.

Jamie: 28:30

I mean, and that was done most, I thought most electric vehicles just got rid of the grill altogether,

Steve: 28:34

no, but it's not getting rid of the grill. It's how do you change the front of the car if you don't have a grill so that you have

Jamie: 28:40

what's airflow, what's the, what's the right shape and and design

Steve: 28:42

but so predicting the airflow that you would

Jamie: 28:45

have based on, that's a complex design

Steve: 28:47

engineering channel, complex design engineering. But it has to lead to a customer satisfaction because of course, A car at the end of the day is more about the sex and

Jamie: 28:56

sizzle you telling me. The AI not only ran the numbers to work out which designs will be most aerodynamic, but also then said, and this is what the consumer will love. I

Steve: 29:06

I really in

Jamie: 29:06

really

Steve: 29:06

want this In this particular case, that is not what happened. This particular case, the designers said, these are the designs that we think our customers would love based on our knowledge of our customers Now tell us whether or not this actually reduces or increases the airflow in a way that improves gas mileage in this case, not gas mileage, but

Jamie: 29:27

I, I, I'm, you know, the thing about three things and altogether is we do try and make things practical. And I, I'm, this is an area that's completely new to me, so I'm, I'm enjoying the intellectual discussion about some of these things. And then I should keep coming back to Yes, but how could I actually today use AI to. Innovate better in my business. I've got three categories of products in this business. I, have furniture, I have lighting, I have this. Large range of accessories, and I've got some sales data history, globally, I'm assuming some clever AI data tool will help me make a better decision of where to invest in, adding to products based on, I don't know, the families that they're in or the categories they're in, and all the rest of it. There's data. There must be data and analytics AI out there that's just gonna spit out answers for me on this kind of

Steve: 30:13

stuff. sadly, for the most part, it's reliant on your data set as opposed to third party data sets when it comes to that

Jamie: 30:20

level and the thing about AI is it does best when it can learn off more data sets than just the stuff I've got, unless I'm

Steve: 30:27

a big company. You can leverage other people's data, but ultimately it's your data that's going to matter in your design, and there are things that you can do. Again, I'm not sure if the goal is to increase sales, increase revenue per item, or reduce cost

Jamie: 30:44

That's a good question.

Steve: 30:46

If it's reducing costs, then AI engines can help you because they could say, based on the product output that you want, can we find you other suppliers that will get you a product of a raw material of similar quality or a part of similar quality with reduced logistics or where it's manufactured cheaper otherwise.

Jamie: 31:05

But we have no idea whether that service or product solution exists. But in theory, that would be something

Steve: 31:10

right now, if you're saying, can we get a customer to pay more for an existing product? Mm-hmm. Then that gets back

Jamie: 31:18

AB testing maybe,

Steve: 31:19

and marketing design as well. Mm-hmm. Can you show on your website specific combinations of, you know, sofa with lighting that haven't been tested before, and then when you test with small audiences, In ways that if you're reliant on a traditional agency, they would take too long or there'd be too many options. But where AI can effectively move things around dynamically, they could run different tests on different audiences. really interesting And then possibly get more revenue by just getting customers to say, that's a great combination. I think I'll buy the light and the tape and the, uh, and the

Jamie: 31:55

So, so, the interesting thing about this for me, and again, I'm thinking about those businesses, That I know and that I work with that are, that their e-commerce channel is a minority of their sales and always will. So, whether they're a physical retailer, whether they wholesale, whatever it is, there's plenty of businesses out there where their owning commerce site mm-hmm. Has a, a limited part to play in the business. we would not have considered leveraging that asset. To answer strategic questions about product pricing or, or, or anything else because

Steve: 32:26

we

Jamie: 32:27

not experts on. AB testing on websites and doing all this and the costs that would've been associated, it would've been, you know, we'd never have done it. We don't have the people out there to do it. But the more those tools become available to start using your own website to leverage, to get information that you can use to sell into your wholesale channels. That's

Steve: 32:43

fascinating. Now, in terms of the last part, which is on the product design in terms of what Tom Dixon's is doing, I think it's less likely in the short to medium term that AI will design the product for you. It may come up with multiple variants of designs that are going to be mostly horrendously garbage. But what it can do is it can do more accurate testing of consumer response if your designers come up with variations. So if you say, we know, roughly speaking that a sofa that has the falling types of curves is viewed as being modern, as opposed to being viewed as sort of old fashioned, you could test a bunch of different shapes. The AI engine will be able to do things around, you know, sentiment analysis from the customers who are looking at the product in the first place. As we were talking about before, gauging physical reactions from people's faces to see whether or not they're actually engaged with the product or whether they're not engaged. So the early warning signs, irrespective of what they might say about the product, what their body language is saying about how they respond to the product. So I don't think the AI engine today can take over from the human designer. I think it can augment the designer to basically narrow down which designs are most responsive to the existing customer

Jamie: 34:05

without wanting to rewrite your first one about avatars. I think the even broader category that was is, is how to use AI to create your digital assets that help in the sales of your business. So coming back to what you were saying on these, these products, where my mind also went is we sell a. Real products that look best in setting and photograph them in setting. And the more settings that you have of that product in different settings that could show someone what it might look like, whether it's in their room or not, the better. So creating those digital assets, whether it's with an avatar in, in fashion, or just different room. Setups or different styles of, of, of, of, of, of design, of other products and furniture around it.

Steve: 34:55

let me take it up. 10,000 feet from what you're What businesses really are in the business of doing selling self-actualization.

Jamie: 35:03

Sure.

Steve: 35:05

so What you're, trying to do is you're trying to drive to that part, that part of the brainstem. Yes. That causes an individual to feel additional self-worth by being a customer of

Jamie: 35:16

don't reduce it to these basic terms. It makes me sad, but yes, I am god dammit,

Steve: 35:20

And so, The more that you can customize to an individual consumer, yes, you are offering. So it to that consumer, it appeals to their self-actualization, the higher your probability is of driving revenue so AI makes it more affordable to customize your offering by

Jamie: 35:41

as long as you have a manufacturing supply chain that can deal with it. But yes,

Steve: 35:44

but even if you have the same It can do a better job than your entire sales and marketing team in terms of equating your existing product to how that customer defines self actualization.

Jamie: 35:59

that customer will need to say something for it to do that, right?

Steve: 36:02

Yeah. There was the example I gave you last time, which was how AI could be used for the good or for the bad, which is in the game space, the game theory

Jamie: 36:10

space. Mm-hmm. I

Steve: 36:12

this example, so you'll hear me use it probably in five other podcasts we

Jamie: 36:16

do. Sorry. You do podcasts with other people, or you just mean between you and me. Between you

Steve: 36:20

you and when we have future

Jamie: 36:20

jealous for a second there.

Steve: 36:23

just as the last generation of AI learned to play chess better than anyone else, last generation learned to play go better

Jamie: 36:30

It's what? It's what? Google, whatever they called before they were Google. Were famous for. you

Steve: 36:35

Deep Minds.

Jamie: 36:36

Deep Minds. Exactly. 24 hours later, beating everyone at chess.

Steve: 36:40

Right. So

Jamie: 36:42

I'll give you Oh,

Steve: 36:43

game. Oh, okay. The game is called Convince You have to convince me of what you're thinking before I convince you of what Now It's why shouldn't AI be the best at playing that game the way it's the best at playing chess the way it's the best at playing go So it can be used in many terrible ways, such as convincing, people to vote for Trump or vote for Brexit. It can be used for very good ways, like convincing a suicidal teenager to have a better outlook on life. It could also, that's a good one, but it could also be able to convince a potential customer why your product is the one that helps them self-actualize.

Jamie: 37:24

it makes me feel dirty okay. I, I, I get that and, and I'm worried that these AI engines were good enough back in the day of the Brexit vote to, and the Trump vote to have done that, but Trump's round round two coming. So they definitely are, they're definitely

Steve: 37:39

aiming to do it. Now the, the difference in the last election versus the next election, and I'm not sure whether 2024 or 2028 will be the first AI election, but in the 2016 and 2020 elections the early versions of AI were used to find likely

Jamie: 37:55

Yes.

Steve: 37:56

The new versions will be used to convert likely voters.

Jamie: 37:59

And, and the finding them was everything in the US system anyway to begin with because, A handful of us voters swing the election. let, let's take a step back if we can, to your 10,000 feet and just talk about AI and why for those who aren't, are less familiar with the, the, current state of ai. Why these sorts of tools we're talking about now, Why it's happened right now. What is it about? The moment that chat g b t ever hit the world that the dunks like me needs to recognize when people are selling ai because people have been selling AI on the back of their, I don't know, their logo generators for some time now, but this, this world has changed rapidly and we need to be waking up to it as

Steve: 38:39

So the way it's been described to me, and again, I'm no expert in the underlying sort of, you know, gradient descent, you know, vector analysis type modeling side

Jamie: 38:51

of things. Just saying all those things makes you wonderfully in the, in the, in that camp. But anyway,

Steve: 38:56

So the way it's been described to me is that people who did research in machine learning that is now ai. Used to work in silos. Yes. If you worked in language, you had your language silo, you had a bunch of experts who knew how to convert text to language. Yes. And if you worked in graphics, you had a bunch of people who would basically use a bunch of images of, dogs to say, this next image is a dog or is a cat. Completely different research,

Jamie: 39:25

different and tools that we sound fam we are familiar with that have been around for

Steve: 39:29

some same, same with video, same with protein folding. Same with all

Jamie: 39:34

these coding. Coding

Steve: 39:35

these individual About five years ago or so, researchers started realizing that you could describe anything as a language. So if you think about a The likely color. That's one pixel further to the right of the color that you're looking at right now. If you have Monet's you could probably predict that about the same as you could. What is the next likely syllable You're going to say if the first syllable was the start of a predictable and so a picture could be described very similarly as a language. A video could be described as a language because. What is the next image in that video, but something similar to the next sentence in a paragraph. Same thing about any other form of AI modeling. It could be translated to think of it as a language, which meant that all these researchers, instead of being in silos, their individual research could leverage each other. Therefore, the advancements that are taking place right now in terms of the ability to use machines to solve problems is not being done in individual silos. It's being done collectively, and so the rate of growth is happening much faster Now, you overlay that against the fact that G P U technologies, the chip sets are getting more and more powerful and some of that, I personally believe this may or may not be true, but I believe that this was driven by the cryptocurrency mining. You want it to be the most efficient miner. You got the latest version of, Nvidia chips, and then Nvidia had an arms race to keep making better and better chips, and as the chips got better, there were new applications outside of just mining Bitcoin. You now all a sudden had these great shifts that could process information very fast. You were able to build out farms that could run all these complex models that are behind the AI engines. So You put those two things

Jamie: 41:33

conflating, conflating of these has happened recently. Effectively,

Steve: 41:37

It's really accelerated in about the last five years, and you're seeing the output in the last 12, 18

Jamie: 41:44

and it's

Steve: 41:44

it's

Jamie: 41:45

part of the. Public conscious because chat G B T has become a consumerish device that people are playing with as have a lot of the image generators, right? So people have started playing with it and understanding it, and a lot of people, myself included, using it. To make ourselves personally productive in our day

Steve: 42:02

to day. It's partly that, but also you remember our investment years ago in Shazam. So people used to see Shazam and they would say, this is close to

Jamie: 42:12

magic. but how are you gonna make money? I said, I'm not investing and walked away and look at you.

Steve: 42:17

It did well, but the point being that your first experience with Chacha BT is that same sort of magical experience, and so it captures the emotion that technology is supposed to capture. I. You want to get behind Chachi PT because you want to be able to impress your friends. Look at this great new poem I wrote about my daughter and my

Jamie: 42:40

Everyone's got the Shakespeare sonnet that they they asked chat gt,

Steve: 42:43

right? Yeah. So that magical experience causes people to tell other people about it builds word of mouth and that word of mouth causes the companies. To invest more and it causes their competitors

Jamie: 42:56

to

Steve: 42:57

rate to

Jamie: 42:57

well, and it moves all the VCs to throw the money at it. But let's, let's avoid that part of the conversation. Listen, I, I, I am firmly of the view that the vast majority of, growth, growth businesses, you know, small and medium sized growth businesses if they are even using your, ai, they're, they're hardly focused on it. And I believe that, you know, there's this really amazing opportunity right now for businesses that, not just the ones that are struggling, the ones that are, are doing well to enhance. And improve their business through ai. But it's, it's a bit of work required for an organization to figure out how they want to do it and where they want to focus. But as much as it's an opportunity today, at some point in the near future, it's gonna be a huge disadvantage.

Steve: 43:44

eventually we will not be talking about ai. It will just be part of the normal, normal business tool If you think about one of the main purposes of technology, it's to continue to drive forward The reason we continue to drive forward productivity is because without driving forward productivity, you either have supply demand problems because consumers effectively start running out of income to buy things people want. They start running out of, leisurer time. To be able to enjoy things that are provided, we start running out of natural resources. By having additional productivity, then you're able to either increase leisure time or increase the ability to procure new items or increase resource, availability or the ability to stretch resources further. AI is the next evolution in that step. We're in a world right now that's at risk of stagflation un undoubtedly. We've got rising interest rates, we've got rising unemployment. How do you break that cycle? You break it with productivity. What's the number one productivity tool today, what we're calling ai,

Jamie: 44:56

there's so much better in there that I get worried about. Particular the way for those productivity gains where they end up. In your pocket as the VC backing these innovations, or in the employees who are getting more productive, or the owner of the one or two or three winning, ai, engines out there if such a thing exists, that aside for a political philosophy, debate another time. there is no doubt that, businesses need to get. The AI Acton and for that reason altogether is holding a summit in, uh, hmm, October, November date to be determined. where I would hope we get to take some of these things we've talked about today with CEOs who've got some practical examples, how they brought it to life, and, and, and have a real, a real roll up our sleeves. And, I will hopefully find a slot for us to have another conversation and have you involved if you're in town during that

Steve: 45:51

time? Wonderful.

Jamie: 45:52

it would be. you know that the idea of that summit came from our conversation three months ago, so we'll call it the Steve Lanker Summit or the Ask Steve Summit maybe, but no, in all seriousness, I, I, I think I'm beginning to learn. But I'm so far from it. and, there's a, there's a lot of practical steps that can be taken sooner rather than later. And I just want to get a, a, a, a platform out there for people to share their stories of how they've embedded AI in their businesses already. So we'll be, we'll be building that into our, into our little summit and, and hopefully that will be of value. Today has been an education. As always, Steve. I appreciate your time as always. You don't need to end on your three things. I'm gonna take the three categories of ai, tools that we talked about, and this food that's been sat here. You should have a little before you go and have supper with your son. Thank you so much for

Steve: 46:37

your time. Absolute pleasure.