How AI Can Tell You WHERE to Use AI in Your Retail Operations With Tomáš Čupr | Spotlight Series

Foreign.

This Retail Technology Spotlight series podcast is brought to you by the Omnitalk Retail Podcast Network.

Ranked in the top 10% of all podcasts globally and currently ranked in the top 100 of all business podcasts on Apple Podcasts.

The Omnitalk Retail Podcast network is the network that we hope makes you feel a little smarter, but most importantly, a little happier each week too.

As and this podcast is just one of the many great podcasts you can find from us here at Omnitalk Retail alongside our Retail Daily Minute, which brings you a curated selection of the most important retail headlines every morning and our signature podcast, the Retail Fast Five, that breaks down each week.

The top five headlines making waves in the world of omnichannel retailing.

And that comes your way every Wednesday afternoon.

Hello everyone.

I am Chris Walton, your host for today's interview.

An interview in which we will explore together not only how AI can be used to improve retail operations, but but also how it can be leveraged to determine how one's operations should actually be improved.

Yes, yes, I will say that last sentence again.

Not only how AI can improve retail operations, but also how it can be used to determine where your operations stand to benefit from AI the most.

And the man you hear chuckling in the background as I said that last statement is none other than Tom Cooper, the founder and CEO of duvo.

He's going to share his expertise on that mind bending subject.

Tom, welcome to omnitalk.

It's great to be here.

Yeah.

I'm curious what made you chuckle?

You're the first person that's ever chuckled in the background of one of my intro reads.

So what made you do that?

I think you said it perfectly well.

And I think this idea of almost like self improving AI.

Right.

That's kind of what people, I think, have in mind when they talk about this artificial general intelligence.

Right.

So the way you said it, I thought, oh, we're already here.

Yeah, right.

It's kind of wild when you step back and think about it from that perspective.

Yeah, I mean, AI is writing code to improve AI.

Yeah, that's crazy.

And you know, retailers can mop their operations to build AI that improves those operations, maps the operations again and improves again.

I mean, what a time to be alive.

Right?

And what a way to blow my mind right at the start of this podcast.

All right, well, before we get to that subject, because I do want to get into that subject and exactly what we all meant by that in terms of, you know, all joking aside, it's a very important topic, but I want you to tell the audience about yourself because you've accomplished quite a lot in retail and perhaps you've accomplished more than anyone I've ever had on this podcast because to say that you're a former retail operator is kind of an understatement.

So why don't you tell everyone about your background and who you are and what brings you to this discussion today.

Sure.

Although you had some great people on the podcast, so very humbly, I built a Pan European e Grocer, about one and a half billion in revenue.

Started doing that in 2014.

So I always say this for everybody.

To be clear, this is not a Covid thing.

So this is way pre Covid, steady grind.

But the proposition is actually amazing.

So you get within 60 minutes, you get weekly basket, it's 20,000 SKUs.

So you get great selection priced at the market.

So it's you, you don't, you don't pay premium.

And we partner with amazing local brands.

So it's just not, it's, it's, it's not this gray supermarket selection only.

And through doing that, I obviously learned some pains and struggles that retailers have in general, not just grocery retailers.

And I mean, that's the start of my company, Duvo.

But it was this, I think it was 12 years I was doing Rohlik and I'm still a group CEO, so I still get a lot of insights from the battlefield, so to speak.

Right, right.

I'm glad you said at the end because I was going to say, man, Tom is really humble because, you know, he, he founded the Roll it group, which is, which is pretty, pretty impressive and, and still wears two hats.

So, so you mentioned Duvo, Tom.

So, so, so what is Duvo specifically?

And why did you say, okay, there's an idea here that I want to go out and pursue.

In addition to your day to day at Roelik.

I was so much into AI when things kind of, you know, blew up.

End of 24, beginning of 25, you know, probably, probably too late in a way, but still, still early enough to see this amazing boom of, you know, new models.

I have, I don't think I've slept between November 24th and May 25th.

You know, there was kind of like every, every day there was something new.

I, I started coding again because of the, you know, AI capabilities.

And I was like, how do I, how do I improve my company with that technology?

Right.

Like, you know, I saw rise of agents, so actually AI being able to execute stuff, you know, I mean, we had like great companies like lovable.

Yeah, you just typed in the prompt and the website would emerge.

So it's like I want to do that with retail operations, but you know, I hit a, I hit a, you know, I would say common roadblocks.

And one was retail operations are so messy and you know, if you want to try to get it into an IT brief, you will find very quickly.

This is why, well, automated retailer doesn't really exist.

Yeah.

On the operation side, yeah.

Like on the fulfillment and last mile, yes, there is a good technology, but you know, category management and supply chain, there's just so many exceptions that I would say standard way of handling things in a code didn't work.

And second, I would say even bigger roadblock is legacy systems.

There is a. Yeah, every part of some process interacts with some weird portal on a supplier side or, or even our side at Rohli, we had parts of the system without APIs.

And then every time I went to it, and this is my company, right.

So I thought this should be easy.

But I was looking at 6, 12, 18 months projects, I was like, there needs to be a technology that solves that.

Like if I want to automate category management process or some, you know, chasing in a supply chain, there needs to be a technology that makes it easy.

Right.

And that was kind of start of duo because that technology didn't, didn't exist at the time.

So that's really interesting to me on a number of fronts.

Number one, the first thing I think about is if you're passionate about work, it's not work, which sounds like that's the case for you since you haven't slept in like two years.

But the other interesting point, the other interesting point about this, to me and what you said is that it wasn't like you were on the AI curve ahead of, you know, everything happening.

You know, a few years ago, it sounds like you, you kind of got the bug, you know, around AI as it was happening in real time with everyone else.

Is that right?

No, absolutely, that's, that's correct.

And you know, obviously now I'm speaking to people at great companies, you know, like Walmart at Amazon and they were like, oh yeah, we, we've done AI five years ago, we didn't do anything like that.

But as it was happening, I became very early adopter of everything.

Every emerging model, every emerging technology, sort of grabbed onto it and tried to implement it into a retail company.

And again, that wasn't easy.

And I thought it needs to be easy.

Yeah, like if I can code today, like If I can do coding, which is hard.

Yeah.

Like I'm not a programmer now, I can code.

Why shouldn't some simple operation like supplier sends you an email with the price list and you need to do something.

Yeah.

That's actually much simpler than writing 11,000 lines of code.

And I was trying to build that system and that's basically what is at heart of Duo.

And I'm really passionate about this because I think future of retail can be autonomous in many ways.

And I don't mean without people.

I mean like operations that don't need people just happening.

You don't have to think about them.

And people focus on, I mean in the end, category management.

Right.

Like category managers should be in the field looking for amazing suppliers to serve customers better.

Right.

What do they do instead?

Some admin work like clicking downloading Excel, downloading second Excel, reconciling these two Excels.

Right.

Like thinking about it and then maybe uploading that Excel to some other system with where somebody downloads it.

Yeah, that, the modern retail work.

And I want to change that.

I mean, you, my friend, you're a needle in a stack of needles in a lot of ways.

As a startup entrepreneur in the technology space with the former operator chops, who understands the real pain points that you're describing here, which is, yeah, there's a lot of people inside a retail operation that are putting data into spreadsheets or hard keying information in from one system to another over and over again.

And it just becomes very mind numbing and as a result very mistake prone as well.

So, so I want to.

And the other thing about it too is, and we've been talking about this a lot on our weekly show is, you know, a lot of the retail organizations, you know, particularly in the US and really throughout the world are being tasked with, okay, how do I think about AI to improve the efficiency of the operations?

And the one thing you always hear is that, you know, the first thing you need to decide before you do anything with AI is you have to know what problem you need to solve.

But as we alluded to in the outset, you're also saying that AI can help you understand what problems actually need solving to begin with.

And that's the key here, right?

No, absolutely.

And that was a big learning for us.

Right.

I am a founder of a retail company, so I, I thought I knew every process.

So as I was trying to implement this great technology, you know, like, you know, Duo exists and you know, I'm like, hey, there's this great technology we should adopt it in the retail company as well.

I realized that not only I do not know the real processes, but people below me do not know the real processes and people below the people also don't know that.

The only people that know the real process, but sadly only part of it are the specialists doing the process.

Right.

And, and how do you reconcile that?

Because people on, on, I would say higher, in the higher levels of the organization, they have the decision making power to change something.

Yeah.

To maybe, you know, automate something, to, to buy software, but they have no idea how the promotion planning process actually looks like.

What are all the exceptions?

So I was like, well, you know, this is a problem because now I have two companies.

One, I'm the CEO of and I want to automate it.

Second, I'm a CEO of that provides an amazing automation and they cannot work together because nobody knows the real process.

So we built this agentic process mapping because, you know, I was like, I was asking and I was like, how do you, how do you solve this?

And we're like, well, you called consultants.

No, you know, McKinsey or Big Four.

Like they come in and they speak to everybody.

Right.

And they give you beautiful slides how your processes look like, like that's amazing.

Because you know what, agents can speak now as well, thanks to amazing companies like eleven Labs and others.

Right.

So why don't we give our agents a voice and they can actually speak to the specialists in their language, extracting every little part of their work.

We can overlay that with recording of the work as well, which is again what a very expensive consultant would do.

And in less than a week, we can map hundred or hundreds of people in the organization and we can tell the leadership, this is what the organization does.

These are the opportunities.

This is where you incur inconsistencies and cost.

And if you want to think about AI and automation, these are top three things you should be thinking about.

And these are the three processes you should absolutely automate.

And by the way, based on the benchmarks we see, these are the three processes you don't even have in the organization.

And you can absolutely, you can absolutely now have them with agents.

Yeah, you don't have to hire more people now to run these processes.

And then leadership looked at that and I'm like, wow, okay, check, check, check.

And then suddenly like this fog lifted and, and everybody had clarity.

So we called it dual clarity.

But clarity.

Nice.

I mean, yeah, we're very creative people, clearly, right?

Very black and white.

Yes, nice.

That's how it came about.

It was absolute necessity.

And, and everybody gets excited about this really, because they thought they don't have a hope.

Right.

You know, we speaking to a VP, I mean, especially in the U.S. right?

You see a lot of this eagerness now to adopt new technology, but what's at the heart is really an Excel file or Excel files, especially in supply chain and planning.

Yep, right.

And they are like, but, but like, how do we.

Yeah, like it's probably in two years we'll have some technology that can clean up this mess and it can actually be two weeks.

And I think that's, that's really eye opening for most.

Right.

And that's, and that's why I was specifically interested in bringing you on for this interview, because I think the approach is really novel here in what you're saying.

And you know, I, as you were talking, I remember thinking back to like a business school case that I had like 20, God, 20 plus years ago where we talked about, you know, identifying, you know, the process by which you, you run an operation.

And there's implicit and there's also explicit expectations of that process.

The explicit ones are the ones that, you know, like you said, the people at the lowest level, they know what those are.

But I think the things that people forget a lot of times is that there's implicit things that those people are doing that are, that are not actually codified anywhere.

And they just know how to do it because they've been doing the job all that all this time.

And so that's where the process mapping stage becomes so important and why the consultants come in and all that.

And I've been a part of those at multiple organizations and they, they take weeks and weeks of effort and coordination and interviews and then, you know, understanding, does this person do it the same way as this person?

And where are the consistencies?

Where are the gaps?

And so, and so what you're saying is that that's what you're trying to first and foremost eliminate and make simpler.

And the other thing I like about it too, when you deploy technology in it, the way you're describing Tom, it takes the biases out of who's.

Who's doing the explaining.

Right?

Like sometimes the process can be explained by the person that's the loudest voice or the most charismatic to the consultants.

But here you're listening to everyone, and the AI is basically, you know, taking what it's learning from everyone and then synthesizing it.

So explain that more.

Like if I wanted to use Duvo as a retail merchant leader, say, I was running planning and allocation at a, at a regular US retailer.

What would it look like?

What would it, what would I do to, what would be my first steps to get the most out of the technology?

I'll answer in a second.

But I think that the way to think about this is this really democratizes a process mapping.

Right.

Because the expensive part, expensive part is the consultant.

Right.

Like, you know, there's a scarcity.

They cannot sit with everybody because that would, you know, break the bank.

But agents are cheap.

You, you can now interview every single person in your organization.

So I'm ahead of planning and allocation in, in this retailer.

I'm just gonna send dual clarity interview.

It's really, they just log in and you know, agent starts asking question.

It's a very well trained agent, you know, on a, a lot of, of these interviews.

So it's probing into the exceptions, handoffs, systems, risks.

Right.

And you don't even say like what's implicit and explicit because if you interview 20 people in that organization or 10 people in that organization, you start seeing what's explicit and implicit.

Right.

Like, you know, there's maybe two people doing this differently.

Right.

But, but maybe it's not bad that they're doing it differently because they're handling different suppliers.

Yeah.

So I don't know, they are dealing with Procter and Gamble and they do need a different treatment.

Yeah.

And our SOPs don't really capture that, that relationship.

But now it's clear that actually any automation we built needs to capture that special PNG relationship.

So, so what you do as a leader, you just send this to everybody and we make sense and we'll give you process map of your team and department and, and, and the risk analysis and transformation plan based on these interviews.

So I understand the first part very well.

So now talk to me about the second part of this.

So it's also like you get the recommendation.

So, so basically Duvo makes it super easy to collect all the information and build a process map for your organization.

But then you're also helping me understand the problems that, that that map uncovers and where I should deploy new resources to get the most benefit.

Explain that more.

And then two, like what's a concrete example that you can share of, of that actually happening?

I think if you think what is an agent?

It's, it's, it's, it's a system or piece of software that when coupled with strong model, it can reason very well.

That's how I would think about it.

Yeah.

And the reasoning capabilities of current models are in the range of 150 to 160 IQ, right?

So what you do is you collect all the information and then you have a system with 150 to 160 IQ.

Say, make sense of this in a nutshell, create process maps and highlight some risk and look for these exceptions and divergencies.

That's I would say one pass complicated long.

But in the end person with that kind of IQ would do exactly that, right?

So you get the current state and then there's another person with 150 and 160 IQ and you're like, look at this, how could we make it better?

Maybe within some constraints that this organization have and maybe don't think about ripping and replacing the core erp.

Like we're not like let's be realistic, just be realistic.

But again, if you say to person with that iq, they will get it.

If you say to machine with that iq, it gets it as well, right?

So that's how you think.

Like this is, in a simple terms, agent is a piece of software with high level judgment with that level of IQ today that IQ will probably increase in the next six months to 180s and you'll get even better output.

But that's where we are today.

It's dialectical, right?

So like if I'm the merchant leader and I'm assessing these outputs, I can interact with it and say, you know, I like this one, I don't like this one.

Can you go into this realm a little bit more and you know, maybe we, you know, piece it apart a little bit more and take these actions versus, you know, X actions versus Y actions.

It works like that too, right?

Absolutely.

Like you can, you, you can take what you like, what you don't like, you can edit stuff, right?

But the point is really the first part, for most leaders, this is the first time they actually see how the org truly works, not how they think they work.

I run a multinational retailer, right?

So I'm judging by my home country, our promo process is ironclad.

That's amazing.

No drop of margin is left on the table.

But then in some other country it's not the same process.

And actually when we did this exercise, we were leaving so much money on the table just by not following the original process.

But maybe some other country making it slightly different for the good reason they fought.

The outcome was that we left so much margin on the table.

Or for example, you know, like there is a process that, you know, we thought was happening which is when a supplier misses the delivery Window to our fulfillment center, somebody follows up.

Because we are very obsessed with availability.

Right.

So we want to know, hey, what's happening?

Because if it's a small problem, we can wait.

If it's a big problem, we need to place an emergency order that maybe comes from a place nearby so we don't face out of stock.

When we did the process mapping, we actually realized nobody is following up.

Basically that part of the SOP isn't happening.

Because when we did the interviews, nobody was talking about following up.

They were just talking about we look at the dashboards and then do reports two days later to the management how poor the supplier's otif is.

But what we actually wanted to happen was place a phone call.

So when we did the transformation of that process, the phone call was a part of it.

But guess what?

Now the duvo agent is making the phone call.

Because why would humans do it if technology can do it these days?

But also there's so many.

I don't have enough people to follow up one minute after there is a delivery window missed.

Hey, what happened?

Are you coming in the next 60 minutes?

If not, what's the reason?

We don't have enough people to do that with 100% SLA.

And that's why.

That's where agents can help a lot.

Yeah.

Just basically fill in these gaps.

Another, another.

Maybe a very concrete example in buying what you want to do.

If you had unlimited amount of people, every time supplier sends you a price list, you want to look at that sku, right.

And you will say h. So that's the new proposed price.

Let me look what they are doing.

Okay.

They are proposing 12% increase.

Yeah.

Okay.

Let me do the analysis whether that increase is justified.

Let me look at the commodity pricing.

Let me look at the labor rates.

Let me look at logistic rates.

Let me look at packaging rates.

Let me look at their margin.

And then, oh, guess what?

Those commodities are 10% down.

Yeah.

So, dear Mr.

Supplier, your 12% increase should be in fact 5% decrease.

Thank you very much.

And you want to do that.

And that's how you negotiate.

Yeah.

Because otherwise what's the point?

Like you just accept as a merchant,.

You can't do that for every item either.

Like you just don't have the bandwidth to do it.

And you're saying that can all happen in the background for you.

Yeah, yeah.

And I think that's.

That's when you mop the merchandising process, you actually realize, oh, this is what I meant.

We can tell the leaders did you know, this process does exist in fact in many retailers and you can now have it maybe done with agents.

And this is the breakthrough because to be honest, I run a retailer for 12 years.

I didn't know.

It's apparently called should cost analysis.

I didn't know that term.

I didn't know that existed.

Right, right.

And now I can do it for every sku, and every retailer can do it for every sku.

And there are companies that sell these analysis for millions of dollars to retailers.

But it's not that hard to do because all the objective data is out there.

You just have to compile them since, you know, do the synthesis and mathematics.

And you say for the sku, this is the right range for this particular sku.

And, and, and that's what I, that's what I mean by autonomous retail.

That person is still going to be there, but, but they're going to get so much better information to do their job based on 100% of cases they need to deal with.

And I guarantee that for 95% retailers, this should, should cost analysis is not happening.

They may not even know it's a thing.

And if they do it, they do it maybe four or five biggest suppliers because that's their bandwidth.

Right.

The rest, generally what you do, the.

Rest they just accept.

Yeah, I'm not going to negotiate with this small supplier.

I don't have time.

But what if you did have time, you know, with agents?

That's kind of the, the logic, that's how you improve really the operations.

Well, the other thing you got me thinking about is like there's autonomous retail, but then there's also consultant free retail, which is kind of what you're getting at here too, which is, you know, to a degree, which is like.

Because the part of this is really interesting to me is as technologies like this take hold, whether it's you or anybody else, you're going to get an understanding and you're going to implicit and explicitly codify over time what you're seeing from all the different retailers with which you're, you know, working with to understand what are the best processes that are out there.

Whereas now from a consultant perspective, that's not, I mean, that's kind of, that's not as baked in, in terms of their understanding too, because that, that, that knowledge transfer gets lost, you know, over time as they work across clients and different things.

It's not necessarily codified systematically.

And so that's a key, key piece of this too, in that, in that you can understand, you know, or Say to the average retailer, like look, this is how you're doing it.

But God, there are a lot of companies that are doing it this way and you could be doing it better.

And this is why our system is telling you and recommending that.

No, absolutely.

This is what I mean by that self improving loop, right.

That we talked about at the beginning.

I think this is absolutely fundamental.

And you know this, I mean you called it consultant free retail.

I think the knowledge transfer is being lost as you said.

But there is another issue potentially that obviously consultants face, which is they, they come up with great ideas but the change management in the company kills those ideas.

Basically as a retailer you try to make a change and the organization resists.

I think if you have combined human and agentic organization that works in a symbiotic way, so to speak, change management becomes much easier because in many ways it's just telling the agents different instructions.

For example, this should cost analysis.

One thing is obviously understanding that the price should be lower.

That supplier is maybe trying to get too much.

But the other part is to negotiate that price.

That back and forth can also be agent by the way.

And you may realize that the tonality of the communication of that agent is too strict.

So suppliers release too much.

So the system could suggest, hey, why don't you play a nice person?

I'm just giving this example, right?

So if you say this to a human team, generally one third is going to do what you want, second third is kind of going to do it, but after some back and forth and teaching and the last third is just not going to do it.

They just going to carry on doing the same.

With agents, once you say change, they change because it's a computer program in the end.

Right.

And I think that's, that's why I would say this consultant free retail,.

Which.

I don't think is going to happen by the way.

But I think consultants will have easier job instilling the change because they will consult agents as well as ubens.

But that's not the case at the moment.

So we think the change that needs to happen is in, you know, creating this hybrid organization.

First we can do it without consultants and then consultants can come in again.

Say your agents are not behaving in the way they should.

Do you want to change them?

And then the change management becomes much easier.

Yeah, and I said that kind of flippantly, but yeah, I think you're, I think you're 100% right.

Like the consultants are probably still going to be there to help.

Help the leadership in the organization through the, the grappling with what you're discussing.

Right?

Like just taking this, taking this idea as a leader is one thing, but then getting it to work and being able to effectively manage it is another thing.

So like, what have you seen in working with, you know, this type of idea?

What does it require of the next generation of merchant leaders to harness it?

Like, how do, how do they need to think about their jobs differently?

What skill sets do they need to have?

Like, it sounds like, you know, you're pretty, you're pretty affluent in terms of coding and you know, your background there engineering wise and whatnot.

But not, not every merchant, you know, comes to the table with that same thing.

So like, so like, how, how, how should the average leader be thinking about this if they're, if they're, if they're buying into this idea, how should they be thinking about it and what should they be doing to prepare for this?

There needs to be a realization that this is inevitable.

Right.

Because I think most leaders are actually resisting, but this is happening.

And I had that conversation so many times even in my own organization.

Because you are speaking to people whose job didn't change for the last 15 years.

Yeah, and that's sad, but the job is essentially moving data from email to Excel, from Excel to some internal system, making phone calls in the process and maybe putting out some fires.

That's the job.

And they are very good at it.

And then what you're telling them is as a leader, you're not going to have to do that anymore because the fires will probably be put out before you even come to work.

Right.

And then you're going to orchestrate an agentic workforce.

So you're going to make decisions that maybe the agents were not able to make based on their instructions.

So you're, almost everybody becomes a coach or an agent operator as opposed to individual contributor trying to make sense of an Excel file.

And that looks like a lot more responsibility because now I have much wider blast radius, so to speak.

Right.

Because that Excel is going to get done in five minutes.

It used to take me 12 hours, now it's five minutes.

So I need to do a lot more in that 12 hours.

And, and I don't need 12 hours worth of five minute Excel files.

So what the company is going to give me is more radius to operate and more responsibility.

So people have to generalize a lot more.

So if you are this narrow specialist, that's probably going to get solved with AI very quickly.

Right.

What's not going to get solved is this general intelligence where you have much wider scope and you're thinking about things in a broader context.

And that's, I think, how leaders need to train the organizations, not just being these narrow specialists, but have a general understanding of the organization and of the world and in, in essence as well.

And I know it's, it sounds a bit vague, but to be honest, nobody knows like how this is going to develop in the next six to 12 months.

Yeah, like I have a agent that helps me massively, you know, in, in my day job very often.

It's because, you know, I mean, I'm, I'm, I'm just below that, that, that IQ range.

So that agent is actually suggesting smarter things very often than I would think of.

Yeah, I'm thinking, you know, will I be needed in 12 months?

Yeah, well, the crazy, yeah, the crazy thing to me is, I think you said it and this is a huge nugget of this podcast, which is like the average scope of what you're managing as an executive.

Like at the executive level, the average scope of what you're managing is just going to get wider.

You said the blast Radis gets bigger, which I thought was a great analogy.

Like, yeah, you're going to be managing more, but then the dynamics of what you're managing are going to be different too because you're not just going to be managing people people, you're going to be managing agents and how you adapt to that, which is going to take on a whole different skill set because it's going to, it's, it's probably, I'm, I'm just, I'm just totally thinking off the top of my head right now.

Tom is going to be, it's going to be less about the softer skills of management, you know, like making sure you understand, like, hey, what's going on in my, in my team members life, you know, what's, what's got them distracted?

What are they good at?

Are they good at math?

Are they good at quantitative thinking, qualitative thinking?

With AI, Those kind of questions, those softer questions are going to kind of be eviscerated and you're not going to have to worry about them, but you're going to have to be managing a lot more in terms of the breadth and scope of what those agents are asked to do.

And so you're going to have two sides of your workforce now, one of which you've never ever managed before in the past.

And that is, that's mind bending.

I'm actually, I would say Less pessimistic on this side.

Because what I think.

Yeah, me too.

Is going to happen is that agents will communicate in a very similar way that people do.

It will probably in the end be some form of voice interface.

So you're gonna talk to them.

And every sci fi movie that you've seen, AI talks.

Right.

Like it's a voice you don't see.

Like you just say, hey, I want a car.

And the car arrives.

Of course, it's a flying car.

Right.

But I don't think the interface will be much different than today.

Yeah.

So I think the only difference will be how intelligent creatures you are talking to.

Right.

Because today you're working with John and you know, John is kind of street smart, but he's not very good at math.

Yeah.

So.

So you kind of compensate him with maybe with another team member.

Yeah.

And then you have a team member that kind of, you know, doesn't have those limitations, also doesn't ask for many breaks.

So to utilize that team member is like working with somebody super smart today.

You have to give them clear and broad tasks that enable, you know, that use their raw intelligence in a way.

Yeah.

And to do that you have to understand the work deeply.

And I think now there's this disconnect for managements.

Right.

They don't necessarily understand the work.

Right.

They manage, they manage people.

Right.

They manage the egos, they manage the moods.

Right.

Yeah.

And, and that's not gonna be enough because there's, you know, they're gonna be still some egos and, and, and, and some moods, but 80% of the work, which again will be much broader than today.

Like everything I describe is the work that is not happening today.

So suddenly there is this well functioning retailer doing 10 times more things today that maybe they could have done it in the past, but 80% of that is done by something that has low ego, no moods, and just needs instructions that are clear and go deep.

Because the deeper and better instructions you can give, the better output you're going to get.

And that's going to be new for many leaders.

Right.

Like even in my organization, I'm speaking to a VP and I know more about the process and the issues in that organization than they do because they only speak to that layer below them.

Yeah.

The other implication of this too, which you've got my head thinking about right now too, is like, yeah, we've been talking about this from the perspective of a legacy retail operation, but there's going to be many retail operations that come online here over the next five to 10 years that are going to take this approach from the outset and that potentially gives them a lot of bullets in the chamber to be very successful.

When you look at things from a productivity and an ultimate profit, operating profit at the end of the day too.

That's absolutely correct.

Right.

Like you know, Rohlik has an infrastructure, we are ultimately very local business and our fulfillment centers are very automated.

That's a protection.

Right?

But if you're a, I don't know, e commerce operator that using a lot of, you know, third party shipping, right.

Like I'm pretty sure in the next 24 months what you do with 100 people and you have all these costs and all these issues, all these inconsistencies, there's going to be a founder that's going to do it just with agents and of course it has different set of issues, but certainly better reliability, better consistency and less cost.

Right.

And that person's going to start chipping away at your market share because they will have more profit to invest to customer acquisition.

Right.

And, and I think that's, that's the opportunity for new entrants and for startups.

But I think it's, it's, there's something about having scale already, right?

Because if you transform, you don't have to transform to one person operator at certain scale doesn't make sense.

But use that scale to dominate even more.

Right.

If you're doing a billion in revenue with this many people normally it would mean if you want to do 3 billion of revenue, you need to hire at least maybe double the headcount.

Right.

You know, there will be some, you know, some, some benefits of scale, but not that many.

But if you're doing a billion with headcount X, maybe now you can go to 10 billion and not hire every single, every, like not hire a single person on top of that.

Right.

But, but, but you already have a scale and that, that's a protection to an extent.

But if you don't evolve, you stay at 1 billion.

Yeah.

With that cost base and there is somebody hungry, they can get 2 billion maybe you know, very, very quickly and then they will scale with much, much, much lower cost base.

But I wouldn't, I wouldn't kind of preach doom and gloom on the big companies who, who already have people because I think that's a great asset.

They just have to use the people to manage agents as well as people and that's the unlock of the efficiency and value.

Right?

Yeah.

The existing, the existing infrastructure, if leveraged correctly, should be a competitive Mode.

Absolutely.

For those retailers.

Yeah.

All right, so let's get you out of here on this because, God, I could talk to you all day because you are definitely.

As I'm sitting here, the thing I was thinking in my head was like, man, you are a visionary in this space now on multiple levels.

But, you know, the question I want to get, get you out of here on this is, so if I'm a retail executive listening to this podcast, and I've been tasked with improving my team's efficiency with AI, as I'm as.

No doubt many, many retail executives are being tasked right now, how would you recommend that they approach that question?

In light of everything we've discussed up.

Until this point today, I think it would make total sense to do a version of Duo Clarity to really understand the organization.

I think that that's the base.

Yeah.

Start from the process, then what you do is, I would say there are many paths, but starting from the process, really understand what processes you miss and what processes maybe you should massively simplify so they are easier to automate or easier to manage by agents.

I think that's a definite start.

And then from there, you will have many options.

There will be many companies you can build, you can buy, you can use hyperscalers, you can maybe use more agile, smaller companies, you know, to work for you.

But, but every single company that will work with you on this will ask, so what's the process?

Right.

Start from the process and, and, and the common second piece I hear is data.

You know, like every single retailer says, oh, we don't have clean data, so we cannot leverage AI.

I think that's becoming less of a problem because AI can make sense of your poor data a lot better than maybe humans and the previous version of technology.

So I think I wouldn't worry too much about that.

I've seen organizations cleaning data for the last decade and still being six months away from being data clean.

And in six months, they will still be six months from the data clean.

Right.

So I wouldn't worry about that.

Start from a process and then the automation options open.

Yeah, we've heard that theme very consistently now about the data concerns that that's becoming less and less of a concern.

But you know what I take away from what you said first and foremost as we close it up here is if you're going to look to optimize and use the, and get the benefits of AI, you have to know your process and what you're actually trying to optimize.

First and foremost, it all begins and ends with.

With taking that step to then, you know, inform what you should do going forward.

So.

All right.

Well, Tom, thank you so much.

That was fabulous.

I love that conversation.

If people want to get in touch with you, learn more about Duvo, what's the best way for them to do that?

Yeah, I mean, you can shoot over an email Tomuvo AI or I'm super active on LinkedIn.

Thomas C U P R. Or you just put rolly group usually or duo AI.

I'll come up.

And I'm responding to all of the messages on LinkedIn as well.

Well, yes, I'm looking forward to continuing this conversation with you and continuing this relationship because, man, I just learned so much from you in this past, you know, 45 minutes to an hour that we've had together.

So.

So thank you again for those out there.

Today's podcast was produced, of course, with the help of the fabulous producer Ella Ella Sirjord.

And as always, on behalf of all of us at Omni Talk, as always, be careful out there.