The AI Employee You Hire on Day One | 5IM

Foreigning us now for five insightful minutes is Arbor Sethi, the CEO of Zenline AI.

And Arbor is here to discuss with us how merchandising teams should think about deploying agentic AI within their organization.

Arbor, let's start with this.

You said something to me at Shop Talk, something that quite frankly, I'm never going to forget.

And I've talked to everyone I can about it.

You said an agent is like your new employee on day one.

What exactly did you mean by that statement?

Bruce?

First of all, thanks for having me.

The concept is quite easy, right?

Your Harvard educated graduate, they come to your company on day one and what you'll see is that on day one they aren't productive.

And the question is, why is someone who has 13 years of education, five years of graduate and postgraduate education, why aren't they productive on day one?

They are highly intelligent, highly capable, and the answer is context.

We'll get to the concept of context engineering, but the answer is context.

They don't know the terms of your company, they don't know the abbreviations, they don't know which products you already offer, they don't know what the processes are, what is the workflow to actually come to an assortment recommendation.

They have to learn.

And then this is basically why you have junior positions, then mid positions and then senior positions.

And then at some point you'll be ahead of merchandiser planning.

And what will happen is that over time you'll accumulate context, or how we call it now, learnings.

And these learnings are what agents need.

So an agent basically starts on day one.

They don't know nothing about your organization.

And what you need to do is you basically need to understand what are the learnings that a head of category management or merchandise planning makes across these four, five, six, seven years throughout the promotions and try to build basically a company brain.

And we call this context engineering around so that whenever the agent gets the question, can answer the question by looking at, okay, embedded in the knowledge and in the context of the company.

And then they will be as productive as a senior merchandise planner would be after a few years.

That is such a great analogy because I can tell you, as a, as a former Harvard mba, as much as I wanted to think I knew everything going into my job day one at I had no idea what I was doing, absolutely zero.

And so I'm curious, like context engineering, like what is your background that helped you arrive at that being kind of how to think about the problem at hand.

My background is basically I'm right in between engineer and business guy, I studied engineering at ETH in Zurich.

I did half a year of research on AI, but then I wanted to basically dive into the business world and did the three years of consulting at the Boston Consulting Group, where basically the entire work that we did was, how do you bring technology, AI into business processes, into the strategy?

How can the company make faster decisions and basically, yeah, become more digital?

And what you learn when you're an engineer, the way you think as an engineer, is very technical.

How do I solve this problem technically?

And BCG always had this concept that basically every business transformation, every Digital transformation is 80% processes and people and only 20% tech.

And I think this is the main advantage.

When we then started zenline that I understood that, hey, as good as your tech can be, you look at Claude, you look at the GPT models, as good as these models are, if you don't have the proper context of how the organization works, there's no way to actually improving the organization.

Yeah, that's a great proving ground too, because a lot of ways a consultant is the same thing as an employee on day one.

Like the consultants coming in, they have no contextual knowledge of how the organization works.

The consultant has to learn that as well.

So it makes sense as your proving ground.

So you mentioned zenline AI.

So what is it and what are you trying to do in regards to it in terms of contextual engineering and improving retail?

My last few years at bcg, I was working on commercial excellence programs for retailers.

And we always say the two most important questions are, what do my shoppers want?

Which products do they want to purchase and what are they willing to buy?

But these questions are answered and solved in a very, very, very unanalytical way for most of the retailers that we work with and also the ones that we are having discussions.

And we said, hey, for the first time, you can use analytical, you can basically use numerical data, tabular data, but also TikTok posts, transcriptions of TikToks social media.

Everything that is out there in the web can be gathered by agents, be structured and then analyzed to find out what your shoppers actually want and be very, very quick in order to make these decisions.

And yeah, when we started zenline, we said we want to answer and help retailers with the core and the two most important decisions.

And that's why we built ZenLine.

And what it basically is, is we sell outcomes, we give, we show you the, not only an analysis, but basically we give you the recommendation of what, what you should act upon in Pricing and in merchandising.

So, so you're basically trying to get into merchandise planning.

Like of all places, like I. Oh, man, fair play to you, my friend, because I've seen a lot of companies come and go over the last 30 years that have tried to do what you're trying to do.

So why do you, so why do you think now is the right time to tackle merchandise planning?

Why of all places, do you want to start there?

Like, are you, are you nuts?

Are you insane?

Like, why do you think this is the right place to go first?

Actually, you're right, it is hard.

And as I mentioned, these are the two most important questions that a retailer can solve.

And these are the ones that actually improve 2, 3, 4 percentage points in margin if you get them right.

But when we started Zenline, the initial thought was if we start a company, if we bother to start a company, then we should do it at the hardest problem that a retailer has.

And these are the ones.

So that's why we're doing it.

And then simply because for the first time ever, it's possible.

Before you could, with, let's say machine learning, you were able to do numerical analysis or numerical data, you had to have everything pre processed.

A lot of human work in data cleaning, data pre processing pipelines, which took weeks to months, sometimes even half a year to just get the data right.

And by now you can work with much, much more unstructured data.

And this was when we realized this from going through the discussions with CEOs, CCOs of the largest retailers in Europe.

Everyone was saying, basically telling us the same, hey, these are the most important questions, but we're still solving them very in a, let's say, more cumbersome and manual way.

So if there would be an AI company that would be able to solve these questions, we would be willing to pay.

And that's what we're seeing.

Yeah, yeah.

No, I mean, I can think back to like when I was a buyer too, like back in 2005, if I had the tools or was equipped with some of the, the abilities that generative AI now affords me to do it.

God, I could have done merchandising so much better and so much faster.

So.

All right, well, let's get you out of here on this end.

My last question I have for you then is like, you know, if I buy in a thesis, like, hard problems are meant to be solved, but they're hard for a reason.

Like, what is the lowest hanging fruit when it comes to deploying AI within assortment planning like, where am I as the average retailer going to see ROI First?

Within the first hour, when we speak with the retailer, we show them, hey, these are the gaps in your assortment.

This is what shoppers are actually looking for.

Here's the combination of what's on the products that you have on your website, the products that competitors have internationally, right?

We scrape companies in South Korea, marketplaces in the US And Europe, and we find where are trends coming up earliest, understanding why shoppers are asking for it.

So what is the shopper need that they're trying to fulfill?

What are brands, very young, recent brands that just recently came up and are offering this?

And how can you, as a retailer solve this gap and give the shoppers what they actually want?

Very, very, very quickly?

And we see that basically within the first two hours, we can have the first discussion on, hey, these are actual gaps in your assortment.

And then within the.

Within weeks after that, some retailers that have quicker processes have these on their shelves.

Wow.

And so, so from gap for gap standpoint, gaps can be like one of two things, right?

It could be like holes in the assortment either online or in store, but it could also be like, hey, maybe you're not even buying enough of this, or this is going to trend up and you need to get more inventory into it.

Is that right, Arbor?

Absolutely.

The first one is more strategic, right?

You want to understand, what do my shoppers want?

Am I serving the needs that they have, that they have, my shoppers?

But the second one is more operational.

And the operational questions can be very, very, very complex.

Because after deciding that you want this one product, as you said, inventory is one question, but the facings that you use in your shelf.

Second question, also, where you put it on your shelf is another question.

And all these operational questions take very often, even longer than the strategic question, whether you should offer this one product or not.

And the first one, I would say agents are already very, very good at answering.

So we're working already with a lot of customers on that.

On the second question, we are teaching agents on how to think about planograms, how to think about merchandising in the operational way, and we're making good progress to also answer that question in a productized way in the near future, which.

Is a good point to end on, too, because you're bringing up the point of why speed matters within the context of merchandise planning and assortment planning, because you've got to get into the products quickly because there are all the operational decisions that are difficult to implement down the line.

And the slower you are on those, the longer it's going to take for you to get to take action on them and get them into market.

So.

Well, thank you, Aubrey.

That was really great.

Thanks for having me.

Chris.