Welcome to the eCommerce podcast with me, your host, Matt Edmundson, and

the eCommerce podcast is all about helping you deliver eCommerce wow.

And this week to help us do just that, I am chatting, uh, with Oliver Edholm

from Depict AI about how AI is changing shopping product recommendations.

That's right.

We are talking about all things AI, why it's such a big deal, why

you want to get involved in it.

You're not gonna wanna miss it.

But before we jump into that, let me suggest a few other eCommerce

podcast episodes to listen to that I think you will enjoy.

Oh, yes.

The first one, uh, is my conversation with Shanif Dhanani about why you

should be using AI in your eCommerce business, another AI conversation,

uh, and also check out my conversation with Tim Jordan about how to choose

a winning product every time.

Quite quickly becoming one of our most popular episodes.

So do check it out and you'll find out why.

And you can find them on our website for free at ecommercepodcast.net.

Now this episode is brought to you by the fantabulous eCommerce cohort,

which is gonna help you deliver eCommerce wow to your customers

in very real and practical ways.

If you are a regular to the show, you will know, for the past few weeks, we have been

waxing lyrical about the eCommerce cohort, uh, and there are many, many reasons as

to why, uh, if you're not sure what it is, it's like a, the best way to describe

it is like an online mastermind group.

It's a membership group, basically all to do with eCommerce where you and a

whole bunch of other folks, uh, are gonna build your eCommerce business.

Gonna learn what it takes.

You're gonna get some expert coaching, but fundamentally, you guys do the work.

So it's not just like an online course that you sit, watch the first half of and

then never do anything with, because what, they just don't work anymore, do they?

So it's very lightweight.

Uh, you can dip in, dip out.

It's not gonna be too onerous on your schedule.

But let me tell you, if you get in there on a regular basis, it's

gonna help you grow your eCommerce business like nothing else.

So if like me, you're a well established eCommerce, or even if you're just

starting out, if you're doing a startup, you're gonna want to check it out.

Strongly recommend that you do.

Honestly, it's gonna be great for your online business.

You can find more information at ecommercecohort.com.

Uh, do check it out.

Uh, or you can email me directly if you've got any questions, and I'll try

my level best to answer them for you.

Uh, you can reach me at matt@ecommercepodcast.net.

That's my email.

Yes it is.

Or like I say, eCommercecohort.com is the website.

Do check them out.

Now.

Without further ado, because you're not gonna want to miss it.

No, no, no.

Here is my

conversation with the incredible Oliver.

Check it out.

Well, I am here with Oliver, a 15, well, well, he is not 15 now, but when he

was 15 he dropped out of high school.

Uh, not quite sure what his parents would've made of that

actually, maybe one day we them.

Uh, but at 16 he seemed what he seems to have done right for himself.

He moved across the world and hustled his way into a machine

learning research position at the National University of Singapore.

Oh yes.

Uh, 17.

How many of you did that when you were 16?

Just raise your hands please.

Uh, at 17 he founded Depict, uh, which he basically wanted to revolutionize

the way we discover products online.

At 18, his company was valued at $15 million, and now at 19

he leads a team of 35 employees.

Uh, one of the most cutting edge, uh, e-commerce startups in Europe.

And it is all about AI.

It is all about shop recommendations.

It's all about that good stuff.

And I'm honestly, uh, Oliver, welcome to the show.

It's so great to have you here.

You are by far the youngest person uh, we've had as a guest on the show.

As I'm sure you, every podcast you have been on, you will be the youngest person.

So I'm really keen to talk to you actually, uh, because I,

I have to tell you the truth.

At 16, I was not hustling my way into machine learning programs

at the University of Singapore.

How did that come about?

Yeah.

Uh, thank you for the introduction.

So I'll, I guess I'll rewind a little bit from there.

Um, so always been this kind of person who likes technical things, computers,

computer games, all those things.

And then, when I was relatively young, even younger.

I played a lot of Minecraft, uh, the computer game.

Mm-hmm.

. And it turns out that if you get bored with Minecraft, they've been clever and

found a great way to make kids start programming by you being able to change

the programming code behind the game.

Okay.

And make modifications to it.

So that's actually how I got into coding.

And from there, you kinda realize more and more that the adult world is,

uh, more tangible than you thought.

Uh, okay.

Programming you can do more than just change the game of Minecraft.

Mm-hmm.

through programming, you can create smartphone apps.

And, you know, as a, like a 12 year old, it sounds very abstract.

Like, okay, you open this app and it adults do stuff.

And no, I couldn't make my own smartphone app and wow, people pay for it.

And then you get this, I guess, high and it kind of just spirals from there.

So, uh, when I was in, uh, middle school around 13, 14 years old, I came

across a book called Super Intelligence by a professor at OS Oxford called

Nick Bostrom, and I, I see that book as kind of a pivotal moment for me

since in that book he lays out the argument of why artificial intelligence

and its development, which is just exponentially growing year after year,

could be the most, or will be probably the most impactful thing ever for humanity.

Imagine what, what, what would happen if we created Einstein, but a thousand times

smarter, you know, , that kind of thing.

And as a kind of tech interested person, I felt well.

If you're going to have a meaning in life and all that.

I was very much looking for a meaning in life.

You know, when you just entered the teenage years.

Yeah, yeah.

You get into that crisis thing and the then, yeah, that's probably

a good meaning, kind of ensuring that AI and its development

happens in the best possible way.

Since we all know technology is a double edged sword.

It can go horribly wrong as well, so probably if you want

to be a part of it at least.

And uh, from there I just went all in on trying to learn as much

as possible around artificial intelligence and machine learning.

Um, and I skipped lectures in school already then, so

I was very all in one point.

on that.

And then, um, yeah, and then three is building things.

I never kind of went all in on the academic route necessarily.

I was building things constantly.

Mm-hmm.

. So I happened to want to build something which I thought useful

would be useful for myself.

And then through using technology, I learned more and more mm-hmm,

and through that, uh, when high school was approaching, I got in touch through

these projects and being able to show what I could do at a young age.

I got in touch with Klarna, uh, when I was 15, and they were very welcoming in

some sense, and lets me do like a summer internship in their AI research team

there.

Then I guess they got impressed, like, Oh, he does things fast and

he knows much more than his age, so I, I got to stay there after this.

And then from there, you know, that's kinda where the seed for

Depict started to come about.

Also, that's where it started having one foot in e-commerce, having one foot in

artificial intelligence and, um, yeah, then, from there, that's how kind of,

since I was working on things I loved and like I saw myself doing in the future

either way, that's how I came to the decision of dropping out of high school.

Mm-hmm.

, uh, I, I didn't really enjoy it as well.

Uh, the process of learning in the Swedish school system Sure.

And.

And then had like adventurous streak, uh, going to Singapore as you mentioned.

Uh, where that was where I was, I had, uh, some ideas on what I wanted to build to

kind of have positive impact on the world.

Mm-hmm.

through machine learning and AI.

Uh, that wasn't the Depict related.

Uh, so this is like a sidetrack of the story, but it wasn't Depict related.

It was the app for helping blind and vision impaired people browse

websites in a much more accessible manner since, you know, they can't

see the website, so it's much harder.

And through that I, I got this collaboration with the National University

of Singapore who helped out with that.

And, but it was during this, to keep it simple, it was during this

period exploring various ideas I, I, deep down you that I'm a builder.

I want to have, I am have, I have a lot of ambitions.

I want to have good, great impact through the world.

Was exploring various ideas and through this period where I had,

through Klarna, through consulting for various e-commerce sites, that's

where the seed Depict came about.

Where I had one foot in, in artificial intelligence, one foot in e-commerce.

And you could see, for instance, the amazing business cases and

the impact Amazon has applying

practically machine learning on the website.

You, you, if you, if you do research about Jeff Bezos, the founder, he has all

of these Jeffisms they're called, where he's constantly repeating himself with

things he loves and one of his Jeffisms is, uh, how much product recommendations

has impacted Amazon's business.

Mm-hmm.

, these others also who have predicted products with this, they've given

huge impact to their business, being able to help customers find what

they're looking for, upselling, cross-selling, all over the place.

And, uh, let's keep it short here, but, uh, If you look at the rest

of the industry in how they handle specifically product recommendations,

it's nowhere near the level of Amazon's.

Amazon has huge, the rest of the industry don't have the scale

to be able to get Amazon level AI and product recommendations.

So the thought of Depict was how could we create an organization

which democratizes this?

This basically by applying the latest research.

And I can go into what we specifically do to create recommendations

which require less data.

Mm-hmm.

since data is the new oil in artificial intelligence.

Yeah, it is.

And through that, have all this impact.

So now as you mentioned, we have a lot of clients.

We raised over 20 million US dollars, multiple founding

grounds, over 35 employees.

And, uh, yeah, it's, it is been, uh, journey to, to be.

Yeah.

It sounds like it's been one heck of a journey to get from, from

where you were to where you are.

Uh, such a, Yeah.

And I'm sure that people say this to you all the time, and I don't wanna

be patronizing at all, but it's such a young age to achieve such a lot

it's quite extraordinary, I think.

Um, Now I, I have a son who's not too dissimilar to you in age, and I'm

trying to, I'm sitting here Oliver going, How would I feel if my child

at 16 says, Dad, I'm dropping outta school and I'm, I'm going to Singapore.

Right.

Um, how are your parents with all of this?

Yeah, it's a good question.

So my parents are incredibly open minded and supportive.

Um, With that said, of course, it's not like you automatically say yes

when you hear something like that.

Uh, so it was a process, I would say.

Um, but also they've been very supportive that I should do what I kind

of have a passion for and so forth.

And, uh, Through also, like it was, I was, I'm in a thankful industry where there's

more of a, there's always a plan B.

If, let's say you don't, you want to start a company mm-hmm.

and, uh, well, you could always be a software engineer or something like that.

Right?

Yeah.

Whilst in some other industries, it's much less like if you want to

play an actor or somewhat, it's much.

Kind of I'm with you.

Yeah, I'm with you.

Yeah.

Well, your parents sound amazing, uh, and, um, to, to give you that freedom, um, at

such a young age, I, I hats off to them.

I'm, and all power to them.

So you, you are in this whole machine learning world and you, you

made reference to the fact that.

Obviously for Amazon, uh, we, we've all been on Amazon's website buying

something and there's a product recommendation and you kind of

go, Oh, I'll have a look at that.

And before, you know, yeah, you've purchased something that

you never set out to purchase.

Um, and occasionally I, you do sit there and go, How in the world did Amazon know?

That would be a good product to show me what is it?

Um, is it witchcraft?

Is it just, is it just luck?

uh, or was there something a bit more intentional behind it?

Yeah.

So this is where you say they've got these really clever

algorithms with machine learning.

Mm-hmm.

, right?

Yes.

So I can't ex, you know, I, I haven't worked at Amazon, I

haven't looked into intellectual property or anything like that.

But there are ways you, I've heard sources from various places and so

forth, and I kind know what the state of the art in terms of the algorithms

powering, let's say YouTube's recommendation engineering, so forth.

Uh, and I, I think so they have many variants and it's a huge company.

At its core, Amazon's recommendation engine is very simple.

Mm-hmm.

actually in sense, but what they are really good at utilizing is

their huge quantities of data.

Mm-hmm.

User behavior data specifically, and the fact that they have a lot of recurring

users, so they always come back.

So you can then start to see patterns where, okay, this kind of user.

Who bought this product tends to buy these products.

That's that kind of logic is the core of Amazon's recommendation engine.

And then they just extrapolated based on that with their insane

quantities of data they have.

If you're a say no typical merchant, well you don't have products across

every possible category, right?

They have like electronics, fashion furniture, blah, blah, blah, blah, blah.

You're probably a little bit niche.

Mm-hmm.

you probably don't have as recurring users constantly buying something

at least every month, right?

Mm-hmm.

. So, um, and then their, their biggest e-commerce.

Site in the world.

So that's kind of what you're standing up against.

So if you're a normal e-commerce merchant and you want to be on par or closer

to Amazon, you have to find other data sources than only this user behavior data.

People who bought this also bought that, uh, and that's what we've

been really focused on doing.

So what we also incorporate into our recommendation engines we

serve to our customers is for, uh, also the product information.

So, uh, the product information, uh, is of course incredible useful

when you recommend a product.

If you go to a physical retail store and ask someone for advice

there, that's like an essential part of a kind of salesperson.

They're giving you advice, but almost all existing recommender systems ignore.

And what, what we can do is we can apply incredibly smart image recognition

algorithms, understanding subtle patterns, which a few years ago was

impossible, but which is now possible.

So let's say it's, um, furniture.

Well, you can see these subtle patterns.

Oh, this correlates to Scandinavian design, or this is probably a little

bit upper end like this, subtle patterns, and then also understanding

all the text data behind the products.

And when you combine it with the user data they have, well it turns out you

have really good results and we, we are pretty confident in showing this.

So we've always had this approach where the first two months use

Depict, you can always kick us out, you can see that it works live on

the site and then decide from there.

So

you've got this system then that, um, doesn't need the quantities of,

because this has been the problem with machine learning for a long time.

And I think, uh, it's, we've had, we've talked about this a little bit on the

show in the past, that AI machine learning have been inaccessible for a lot of people

because we just don't have the data set.

And my understanding with, um, certainly in the early days was machine

learning needed insane levels of data.

I mean, you needed super computers just to process the,

the data.

Yeah.

There's still the case.

There's still the case is like if you Google, uh, uh, open AI, uh, image

generation for instance, there are this insane, uh, artificial intelligence

models which spit out like boththe realistic images where you can just write

a prompt of, let's say a teddy bear on the moon riding a horse like that

sounds weird and you can literally paint a foot realistic image of that.

So like it's really developing, but it still needs insane

quantities of data and it's like millions of dollars just to train.

I was interrupting you with definitely true still today, I would say

So

I mean, is this where you, um, I, you know, you kind of hear the stories

like say you raised 20 million in funding and you kind of go, where

do you spend that 20 million?

Does a lot of it go into then analyzing data?

It's to sheer just get data in and let's find some patterns in there.

Yeah.

Um, it still goes mostly to head count.

Right now.

We definitely have more server cost than the typical SaaS business

due to kind of having to deal with a lot of data, um, and so forth.

Uh, but, uh, and you're right about the fact that a lot of, a lot of companies

try to be on the forefront, Okay.

AI machine learning is really trending right now.

We, we need to get ourselves some AI, right?

Like in the early times we need to get sales on internet.

Mm-hmm.

I don't know what it does, but should probably get it and, uh, Usually

it doesn't go that well if you're not kind of explicit about it.

Since if you try to use some state of the art model, well, it requires huge

quantities of data, which you don't have.

Mm-hmm.

. Whilst probably for most eCommerce merchants, what's extreme still

extremely high leverage for you.

Being data driven and starting from there, having a core foundation

where you can measure things and have quick feedback loops where I have a

hypothesis, I can actually measure, change something, measure it, and then

go through the cycle pretty quickly.

Uh, that's probably where I would start.

Then there are third party services like Depict where you can like plug and

play and really get impact through that.

But that's, there's so many things you have to handle as e-commerce merchants.

So like I, I would start there.

And then, yeah, there, there are some AI things, machine learning things

which are probably not as complex, but you know, there's some, some simple

algorithms which require less data, which you can still, uh, gets use of.

But I would start with can just, how can we data driven and shorten

the feedback loops between having a hypothesis, measuring it and

make, making a change, and then measuring the impact, measuring the,

and so with, um, something like depict then, um, what I'm picking

up is actually now machine learning.

Um, AI is at a place where if you've got significant quantities of data, great, but

if you don't, we can still work with that.

Am I, am I, is that, am I understanding that right?

Exactly.

Exactly.

And that's especially where Depict comes in.

Uh, if you don't have, let's say, this is a great example.

Let's say you launch a totally new product collection, uh, and you, so

what's traditionally you would do from a product recommendation perspective

in the recommendation related product.

Bar is that you would look at the historical purchases of a product and

say, people bought this, bought that.

Well you, you just launched a new collection.

You have all these campaigns, all these kind marketing, getting a lot

of traffic to these products, but there's no historical data people

bought this also bought that.

Well, no one really bought it.

Yeah, before, so what you're gonna do well with Depict we, we understand

the product as well, so there's a lot, a lot of context that in the

same way a shopper would utilize or recommend, uh, in store clerk would

utilize when recommending products.

We, we, we can still do.

Um, so that's a clear example where kind of lack of data or cold start problem.

Big example.

Another example is let's say you have Black Friday.

Uh, you have a lot of campaigns all over the place, and, uh, the, the user

behavior on your site is drastically different than outside of Black Friday.

Uh, if your recommendation engine learned from that behavior and kind

of tries to copy it, oh, people buy this product and this product a lot.

but actually it's due to the fact that they have like a

90% discount or something.

Mm-hmm.

doing that after Black Friday is a really stupid idea.

. So it comes to this problem again, right?

So, yeah.

Um, yeah.

So how much, I dunno if you can answer this.

How much data do I need to have reasonably to get started?

Like if I was start.

Um, a new product range, that's fine, but mm-hmm.

, I you, there's an assumption there that I've got a website that's already

trading, that has already sold product.

Um, I'm, I'm starting from zero today.

I've got no, no track record.

At what point does AI start to make sense for

me?

Yeah.

So I would start with, Okay, you don't have any data then I would implicitly

assume you don't have that much traffic.

Mm-hmm.

on your site.

Well, if you don't have that much traffic, then you can't really work

as data driven as you would want to since you, you, you need to have

ways to measure your hypothesis and see how it impacts the customer.

You need to, have significant amount of data sufficiently the amount,

sufficient amount of data, so you can see statistically significant

correlations between different behaviors.

So there's some limit there where, where, And if you don't have any traffic, you

should probably solve that issue before.

Maybe there's a marketing solution which uses AI.

But I would look at what problem does this AI solution solve and does it

solve the problem I need to solve?

I wouldn't ask, Are you using AI?

Oh, you say it, then I should use it.

You should ask, what problem are you solving?

How well are you solving the problem?

Is it the problem I want to solve?

And.

And, uh, yeah, if it turns out that you want to increase, you have sufficient

amount of data to have a sense of that.

Well, we want to increase our conversion rate, average order value.

Let's say you have, uh, a lot of products in your warehouse, which aren't

selling, and they're just lying there.

Or you have some specific business objective you want to optimize for

is recommendation engines can help, uh, pushing certain products in

certain categories to, for extent.

Then I would look at, uh, look at applying a recommendation and then, and then see,

see what, what the recommendation engine costs and how much you get out of it.

See the ROI multiple.

Depict is a little bit more of a premium provider today, since we get so many

requests to work with us, we don't have time to work with with everyone.

So that's on the traffic part, it's a little bit of a ramble, but then there

another important aspect is also how many SKU or products you have on your site.

Mm-hmm.

. So, um, If you only have like 30 products, SKUsor whatever, then it's quite easy

as finding the products you want.

But let's say it's over 300, even a thousand, well, suddenly you really need

to aid the customer in getting a sense of what you have in your product collection.

So it's

a.

I guess if you wanna get started out with it, you are looking at both your traffic

and the number of skus that you have.

Exactly.

Yep.

Um, and so that's got to be at a reasonable level before

AI makes sense for you.

Exactly.

And so several hundred skus and probably what, a couple of thousand

people at least visiting your website.

I'm thinking, so, And is it, is it, is it fair to say, Oliver, that the more

data I have, the better my AI will be?

Um, that tends to be the rule of thumb, but Depict really works to ensure

that we can work with more sparse with sparse data sets, which have less data.

Uh, there's, for instance, we, for instance, work with some marketplaces,

they have millions of skus.

I think the marketplace with the most skus has 16.8 million.

Uh, they have a ton of traffic, I assure you that.

But a lot of SKUs have very few interactions.

So there's also the question like, uh, inter amount of interactions per

SKU, so we still in those cases, have to work with low quantities of data.

Uh, y

Yeah, no,

that's fair enough.

1.6 or 16.8 million skus.

I mean, that's gonna be a headache for somebody, right?

Geez.

Someone's gotta put those on the system.

Yeah.

So, Okay.

Uh, I tell you what we're gonna do, We just gonna take quick break, listen

from this week's show sponsors, and then I'm gonna be right back with Oliver to

carry on our conversation, uh, about AI and all things machine learning.

Don't go anywhere.

Hey there.

Are you a business owner ? Here at Aurion Digital we know firsthand

that running an eCommerce business can be really hard work.

As the online space gets more competitive, it is becoming even more

challenging to stay ahead of the curve.

We totally get it.

So we wanna help you succeed by offering a wide range of services.

From fulfillment, marketing, customer service, and even coaching and consulting,

just so that you can do what matters most.

Save yourself the time and the money, and let us handle the day to day tasks.

This way you can run your business without having to worry about the boring stuff.

So what do you say?

Are we a good fit for each other?

Come check us out at auriondigital.com and let us know what you.

So welcome back.

Uh, we are talking about AI machine learning, um, and how it can

make a difference to eCommerce.

So you've talked specifically about recommend what you call

the recommendation engine, which actually sounds like something outta

Star Wars, if I'm honest with you.

This is a recommendation engine.

It drives this spaceship over here.

Mm-hmm.

, Um, it's a, it's a great phrase.

Um, And so I understand that AI can be used to recommend products

based on the traffic, based on behavior and past behavior and the

data sets and so on and so forth.

Where else do you see AI having a big impact, uh, in eCommerce

that we should be thinking

about?

Yeah.

Uh, I would look, so I would look at three categories primarily.

And one is the space we are in, right?

Like pro, some sort of product discovery, personalization,

product discovery, uh, Depict.

Some people associate depict with only being on the PDP

product page for instance.

But no, we were across the whole buying experience, the landing page.

Uh, Product detail page check out.

When you add something to the basket, you helps slow cross off.

So there's a lot of areas we can, you can be there.

Um, and you can use product discovery for impacting not only like this

general conversion rate average shorter values, like talk metrics,

but you can go much more pinpointed as well with saying for instance,

oh, we have these higher margins products we want to push for this

kind of product customer segment.

Well, you can do that through great product discovery.

You can look at, at a lot of things like that.

Then I'm less, I'm, I have worked less within these areas, but I know for sure

that there's a lot of impact there.

Uh, first is logistics.

Okay?

Uh, yes, Google Logistics.

Artificial intelligence, Amazon, or, uh, then, then you, you

will find a lot of things.

It's everything from how you optimize various routes, how you place all

the products within the warehouses.

Um, there's a lot of, lot of optimization problems within logistics where, where

you can apply AI, so to say, uh, and.

Then there's also the aspect of marketing or more quantified parts of the marketing.

So where you have a lot of data where, well, there you

can apply AI to some extent.

Well, of course you have Facebook and Google ads.

Those are extremely sophisticated artificial intelligence engines.

But, um, I, there there's a lot of.

I, I, I'm less aware of the specific use cases there, but there's a lot

of cool stuff you can do there.

Okay.

For sure.

It's interesting, isn't it?

Uh, AI and marketing.

I've seen a lot of things coming up recently on my, uh, on my social media

feed, which is normally where you, you know, you see things in it first.

And, um, I've seen, uh, AI copywriters.

So, you know, the, the computer will write

copy for you.

Those are getting really good.

Yeah.

Yeah.

They all getting really good.

Scary, good.

Actually.

It's quite fascinating to me how that, how that all works.

Um, I was just, before we were on our recording, I was having lunch with

a friend of mine who's a barrister.

Um, so a lawyer here in the.

Contract barrister, but he, uh, like you, uh, has a really good understanding

of machine learning and he's, he's done all kinds of stuff at university in

it and quite how he ended up in law.

I have no idea.

But anyway, he, he's, uh, he's, he's a machine learning guy and he's

got some very expensive computers in one of the rooms in his house

where he's, he's got stuff being churned out and he's, he's doing AI

um, to process, uh, case history and legal case.

Case history.

Yeah.

Yeah.

To be able to fill in stuff, uh, and do a lot of the sort of the, the mundane

work that a barrister has to do.

Where he spends hours of time, this thing will just do it in

obviously fractions of a second.

And so he's, he's having a bit of fun experimenting with that.

Mm-hmm.

. Um, so I, I, I've seen AI in law, I've seen it in copywriting, I've

seen it in marketing, you know, where the claims are you don't even

have to do your own adverts anymore.

This thing will just do the whole thing for you.

subscribe to that service.

Some of it, if I'm honest with you, feels a bit gimmicky, a bit

like someone's just got an idea and they've thrown the phrase machine

learning or artificial intelligence.

It, uh, and really it's to stem with an Excel spreadsheet

somewhere in the background.

Mm-hmm.

Right.

Um, How do I, I guess, how do I avoid the gimmicks?

How do I avoid the hype and the bluster and the, just the sheerer

to nonsense and just focus on what is actually gonna help me in my

business?

That's a great question.

I, I agree.

There's a lot of buzzwords around AI, uh, ton of buzzwords

around ai and many exploit less.

But I'm not asking you to become an AI expert of any sort.

I think being an AI expert and being a great operator of an e-commerce site

doesn't necessarily correlate that much.

So I would actually become really good at asking yourself, what

problems do I need to solve?

Mm-hmm.

. And does this solution actually solve this problem.

You don't have to look at the inside of it necessarily.

You can just look at, can I run an AB test here?

Can I see some sort, you know, tricks around that.

Where does this, what's the track record of this solution?

Yeah.

And solving my actual problem.

And then don't care if it uses AI or not.

Uh, simple solutions tend to be.

Working surprisingly well a lot of times.

So that's, that's how I would think about it.

Then on a like more meta level, I really believe that in, as for

every year that comes, state of art within AI drastically develops

and it's developing exponentially.

So I think in 10 years or much less than.

The average e-commerce buying experience will be drastically evolved to the extent

that, let's say I'm buying, I dunno, groceries, uh, we have some of these 10

minute grocery delivery apps as customers.

Well, I, I believe that at some point I'm just opening my app.

And the basket is already filled in.

He knows what I want.

You know, , I buy, I'm so predictable in my, like grocery purchases.

Oh, it's scroll milk.

Yeah, avocados, toilet paper, whatever.

Looks good.

Buy right.

Then within fashion, let's say, uh, even though, uh, an AI engine

actually knows what you want.

Maybe you shouldn't have the same transactional experience as with

groceries, but in fashion, it'll be very interesting to see how

the user experience will look when the AI knows what you want, but

buy shoppers don't want to have the sense that it already knows what you want.

Right.

it's a real delicate balance.

Right?

Yeah.

Uh, some, some like illusion of choice and, and, and it's scary.

Scary.

You know, it's scary that kind negative side effects of

this on society and I think.

That's, I want to Depict to be a kind of thought leader in that a, as we

grow, I, I believe that just because it has bad people have bad effects

on society, you shouldn't run away.

Well, you should be part of it, ensure that it's done in the best possible

way, uh, since it's going to happen.

So, uh, that's on a more meta level.

But right now, if we want to grow your e-commerce business, I will

focus on the fundamentals and, um, a lot of AI solutions out there, uh,

are much more stupid than you think.

Yeah,

That's What do you mean by that?

Why?

Why would you say they're much more stupid than you think?

Well,

they, they, as you said, they have some Excel, some equivalent of

an Excel spreadsheet somewhere, and they do simple volume

correlations, et cetera, cetera, to.

That that wouldn't feel as much as ai.

It's kind of a subjective term when you call something AI or not.

Uh, yeah.

If it turns out that they use an Excel sheet and it really

solves your problem well, Well, I wouldn't be happy using that.

Yeah.

Yeah.

Yeah.

That's really interesting.

It's really interesting.

So, I mean, that's where you, I suppose you see the future of AI and eCommerce

going is it's gonna be much more I've, you feel like you've got a choice, but

really we know what the choices you're

gonna make.

We're not interesting where that evolves.

Yeah, yeah, for sure.

Yeah.

That's, that's

gonna be clever.

Where do you, where do you see eCommerce generally going in the future?

Uh, okay, so there's multiple timelines here, right?

So, uh, it's.

I would double down on what I previously said around kinda

on the longer term time scale.

I really believe that like at the end of the day, when you build a

core engine, which knows what the customer wants, and you can hopefully

take into account, you know, Okay.

How much does this customer care about sustainability?

How much does this customer care about you know, those things which

aren't necessarily as flagged in, in today's shopping experience,

uh, can be taken account.

I, I think that will be one of the most pivotal things,

the timeline for that time.

Less, less, less.

Sure.

About, um, something I see on a very, very short, short term basis, uh, is that, um,

e-commerce website, e-commerce merchants lack the IT resources for them to really,

uh, you know, move as fast as they should.

Mm-hmm.

. So a lot of, you know, when you want to shame, when you have a hypothesis on how

you want to shade in something, IT is usually involved in some way or another.

And if you don't have the foundation or the expertise in house, to make

those changes really quickly, then you're kind of locked and yeah, you,

you can't move as fast as you want.

So, uh, I hope that e-commerce is moving towards kind of building

a better foundation on, on, on that than in the future.

Yeah,

that's a really fair point.

I kind of see myself the, um, . I, I almost wonder, cuz let's say,

I mean Amazon by far is the biggest e-commerce platform.

Mm-hmm.

, right?

Uh, in terms of transactional, one of the other biggest platforms,

let's say, is Shopify for sure.

Yeah.

And so you've got a lot of sites who use Shopify.

Um, I'm curious to see if Shopify in their development are gonna build

in AI features into that platform.

Um, so that they, So that if I launch a website on Shopify, yeah, it starts

from day one, analyzing data and helping me build pictures as I go along.

Um, I've yet to see an eCommerce plat.

I, I, I see, you know, you've got depict, for example, which plugs into, say,

a Shopify site or to other websites.

Um, I just wonder whether at some point in the future that somebody's

gonna write a really clever uh, platform that understands eCommerce,

that understands ai, that understands, you know, all the different elements

that make up eCommerce and not just one aspect of it or one bit of it.

Um, so everything from shipping to, to whatever, and.

I'm, I'm, I'm really curious, will there become like this super AI platform that

really transforms how eCommerce is done?

I don't know.

It's in effect.

Like Amazon going here, here's my platform.

Set up your website with this,

right?

Yeah.

I, I, I really hope that, that it will happen, uh, doing like being

being great everything is hard , especially when you go into very like

technical niche things, which requires building up an organization and so forth.

So my impression based on the interactions I've had around Shopify

is that they really want to emphasize the ecosystem around Shopify and

the marketplace they have and

create an environment where if you are, if you are an AI researcher, an engineer,

and you have this great hypothesis for this new, new thing, which could work for

eCommerce merchants, instead of having to create all this integrations to all

these different platforms, etcetera, you could just in a very simple manner.

Create this algorithm or simple surveys and through Shopify could

be, could be this multiplier effect where other merchants can see the

track record of that app and so forth.

So, uh, I hope that it, it that will happen.

Uh, and, and my, my, my sense of Shopify strategy right now is they

want to welcome players like depict to kind of help on those things.

They necessarily don't have the focus to really nail down right.

Yeah,

that sounds fascinating.

I am watching and waiting with bated breath because I, I, I was there

when eCommerce was born and I'm, Yeah, it was very, very simple.

, and now it seems very, very, uh, it's, you know, uh, extremely smart young people

like yourself are taking over in ways that no one ever dreamed possible 10 years ago.

So, Um, it's brilliant to see and, and, and brilliant to chat to you Oliver, Thank

you so much for coming onto the podcast.

How do people reach you?

How do they connect with you if they, if they want to do so?

Yeah, thank

you.

So you can reach me at Oliver.Edholm@depict ai.

I think that's the easiest one.

So it's Oliver dot EDHOLM@depict.ai, uh, then I'm on LinkedIn if you

search for Oliver Edholm there uh, we can also chat through there.

I think that's wonderful.

The

easiest one.

No, that's great.

And we will of course put the links to Oliver, his LinkedIn and his email

in the show notes so you can get hold.

Uh, if you are subscribed to the show notes, uh, then we, they'll

all be there and, um, reach out to Oliver with your questions.

Uh, it's gonna, and watch, watch what, uh, Depict does, because I'm

really intrigued by how they're gonna motor forward on this.

Uh, Oliver, thank you so much for joining us.

Really appreciate you being here.

It's been honestly, a real treat.

Thank you.

So there you have it.

I'm still mesmerized by this conversation.

Uh, such an incredible story, isn't it?

Uh, huge.

Thanks again to Oliver for joining me today.

Uh, and also, let me give another big shout out to today's show

sponsor the eCommerce cohort.

Do head over to eCommercecohort.com for more information about this

new type of community that you can.

Be sure to also subscribe wherever you get your podcast from because we've got

some great conversations lined up and I don't want you to miss any of them.

And in case no one has told you today, you my friend are awesome, utterly, awesome.

Its a burden we all have to bear.

It's just the way it is.

Now.

The eCommerce podcast is produced by Orient Media.

You can find our entire archive of episodes on your favorite podcast app.

The team that makes this show possible is Sadaf Beynon, Josh

. Catchpole, Estella Robin and Tim Johnson.

Our theme song has been written by me and my son, Josh Edmundson,

uh, people ask me about this.

To be fair, I wrote a very basic melody and Josh did everything else, so I

should probably give him more credit.

Uh, if you would like to read the transcript, all show notes, head over to

the website, eCommerce podcast.net where you can also sign up for, our newsletter.

That's it for me.

Thank you so much for joining me, and I hope you have a fantastic week.

I will see you next time.