AI agents are changing how products are discovered and selected - but what does that actually mean for day-to-day marketplace operations? In this episode, Ingrid speaks with AI specialist Barbara Branca about why the logic is shifting from rankings to recommendation systems - and why “good enough” product data is no longer sufficient. Barbara argues that listings are turning into decision documents: AI systems no longer operate on keywords alone, but aim to understand intent and prepare purchase decisions based on that. This requires complete, structured, and consistent data - not just within a single marketplace, but across all channels. At the same time, the conversation gets very practical: which tasks in marketplace operations can already be meaningfully automated with AI, where real time savings can be achieved, and how to get started without overhauling everything. Barbara’s recommendation: don’t start with your entire catalogue, begin with your top products and assess whether your data is truly decision-ready.

Note from the sponsor eDesk:

As soon as sellers operate across multiple marketplaces, customer support quickly becomes highly complex. Irene Epp from Pertemba described this vividly: her team handles more than 1,000 customer inquiries per day - across over 130 marketplaces worldwide. Having spent many years in customer support herself, Irene calls it the “watchdog of the company.” Together with Gareth Cummings from eDesk, she explained in podcast episode 145 how this volume can actually be managed: all messages, orders, and customer data are brought into one central inbox, marketplace SLAs are tracked, and response times stay within required limits. Routine inquiries such as shipping status or returns can be automated - allowing the team to focus on more complex cases that require real judgment. If you missed the episode, it’s definitely worth a listen:

https://open.spotify.com/episode/1cKbdDr2ipbnGCwo0KmxKF

Note from the sponsor Kaufland Global Marketplace:

If you’re looking to expand your marketplace business internationally - for example to Poland or Italy - it’s worth taking a closer look at the Kaufland Global Marketplace. Through Kaufland, you can currently sell in seven European countries: Germany, Poland, the Czech Republic, Slovakia, Austria, Italy, and France. The key advantage: you can manage all marketplaces through a single account, without having to set up your products separately for each country.

We also have an interesting offer for you: With the promo code MP-UNI2026, new sellers can sell on all Kaufland marketplaces without a base fee for three months. The offer is valid until May 31, 2026.

And if you’d like to learn more about how brands can build visibility and branding on Kaufland marketplaces, make sure to tune in to our upcoming podcast episode 147. Our guests will be Markus Anstots (Kaufland), Jan Weiß and Marvin Thöne (both Sleepling).

Are all the relevant attributes filled in?

What are all the data you are giving to the system?

I don't know.

Amazon is asking for 20 fields to be completed.

Are you filling all the 20 fields?

If not, make sure you do from now.

Let's Talk Marketplace.

The Marketplace podcast with Ingrid Lohmer and Van Friedrichte.

Hello and welcome back to let's Talk Marketplace.

Glad you tuned in again, this time for episode 146, if you can actually believe it.

I'm Ingrid and today we will be talking about AI.

Yeah.

And before you all groan and go like, please, good lord, not again, just hear me out, you know, because I know agent E Commerce is everything everybody seems to be talking about this year, but you know what, I think it's always very hands off, very strategic, very half a glass of wine and philosophize about the end of E commerce as we know it.

But I'd like to get the topic a bit out of the ivory tower and down into the actual marketplace business.

You know, like, what does the shift to agent E commerce really mean for you as marketplace professionals for your daily work?

And that's exactly what I would like to talk about today.

And that's why I'm really happy that I can welcome today's guest to the show.

A very warm welcome to Barbara Branke.

Hi Barbara, it's really great to have you.

Thank you very much, Ingrid, thank you for the invite.

I'm really pleased to be here today.

Yeah, yeah, and me too, actually, because we met because a few months back I voiced over LinkedIn my frustration with, how can I put this?

The male dominance in the field of agentic or AI topics or agentic commerce topics, topics.

Because everything that I was seeing in my feed were male experts talking about it.

And so I went like, okay, where are the girls?

Where are the AI ladies that know how to deal with this new topic?

And that was when a good business friend of mine introduced you to me.

So that was really cool and we had a great talk.

So maybe, yeah, introduce yourself a little bit.

Barbara, what are you doing and why did Nico recommend you to me as an AI expert?

Yes, sure.

What I'm doing since a few years, I'm working in the E commerce and marketplace optimization.

Yeah, for more than five years actually.

And at some point I realized that one of the biggest problems in marketplaces today is the data quality.

I mean, data is not just something that became important now.

It is already important since some years but especially now with the upcoming of AI agents and AI systems is becoming even more important and relevant because, you know, AI systems can only understand something about your products if they understand the data behind these products.

I Co founded ICOM, an AI SaaS platform, and with this platform we automatically structure, enrich and optimize product data for marketplaces.

So this is now my connection to Nico, because he's a client of mine and we're working for his optimization.

And so this is how we came together.

Yeah, absolutely.

So, but this is a new topic, of course, so it's evolving all the time.

How are you keeping up with everything that is happening in that space at the moment?

Because I find that really difficult.

It is difficult indeed.

And honestly, I must admit that I'm trying to read daily, on a daily basis, the development happening really every day.

Right.

Because AI systems are continuously evolving and so also marketplaces, of course.

And yeah, it is very, very important to, to stay up to date.

It's not easy, but it is necessary to do so.

Yeah, absolutely.

But it's not an easy task.

So really good to have you today and dive a bit deeper into this, which we will do.

After a very short advertising break,.

I.

Wanted to come quickly back to last week of let's Talk Marketplace.

So if you missed that episode, let me tell you, we talked about a topic that many marketplace sellers underestimate, and that is customer support.

Because once you sell across marketplaces and across multiple marketplaces, support can become incredibly complex.

That was clear in last episode when Irene Ebb from the retailer Pertemba described a striking real her team handles more than 1,000 customer inquiries per day across more than 130 marketplaces worldwide.

Irene spent many years working in customer support herself, which is why she calls it the watchdog of the company.

Support is often the first place where problems become visible.

Whether it's a SKU with rising return rates or a car struggling with deliveries, or customers suddenly asking the same questions again and again.

Whatever it is, it always should make you perk up and think, okay, what is happening there?

So in that episode, together with Gareth Cummings, CEO of Edesk, she explains how technology makes this scale possible.

With Edesk, Irene and other sellers like her can bring messages, orders and customer data from all marketplaces together in one central inbox.

Keep track of different marketplace SLAs and make sure that response times stay within the required limits.

Automation helps handle routine questions like shipping status, you know, where is my parcel returns or tracking.

And while human agents can focus on the more Complex customer situations that actually need judgment and experience and the human voice.

So if you missed that episode, it's definitely worth going back and giving it a listen because, you know, customer support is something that affects all of us.

We've included the link to the episode in the show notes.

All right.

Okay, Barbara, so let's dive right into it.

If we are looking at marketplace development over the last 10, 15 years, we were mostly talking about Amazon SEO, for example, how to present your product data in a way that it can be easily found by the search engines on Amazon or some Google shopping.

Just the same thing.

And now we have a very different interface coming between our products and the consumer.

AI interfaces, ChatGPT, Gemini Publicity, you name it, they are recommending products.

So does this change the way that marketplace managers have to present their products now?

Yes, the topic you are mentioning is a very important one and this change you mentioned is indeed the most important shift happening today.

As you stated, mostly until just a few years ago was about optimization for keywords.

This was the main topic within a marketplace.

Now it is very important that sellers optimize their product data not only for humans or for classic engines or for the marketplace itself, but also for AI systems.

And what AI systems need, it is a very clear product structure so that they are able to interpret, compare and recommend any product with confidence.

So what I always say to the people I'm working with is that they need to start treating their listings or any product description.

It doesn't matter if it is within a marketplace or inside their online shop.

They need to treat their product description as a decision document.

Which, by the way, it has always been important, but today it is becoming even more relevant to do so.

It is not about keyword stuffing.

It is not about only an algorithm understanding what it is about, but also, you know, these external systems coming up very strongly nowadays.

I like that phrase decision document in a way.

So make your product listing into a decision document.

Let's linger on that a little bit.

So in what way does an AI come to a decision, to a shopping decision or to a recommendation decision?

More to say when it's looking at the product page, what it's looking for?

Well, the AI system is looking into the product data more broadly.

I will tell you very soon what I mean by that.

First of all, it needs to be confident that the product he's recommending or it is recommending is really for, or what is the consumer asking for?

It is answering the consumer need.

So as I mentioned earlier, a product listing or product data basically need to be very clear, complete and consistent in order for these AI systems to understand what you as a seller are writing about or you are writing about.

And in order for it to recommend your product.

This means that your product description or your product data basically needs to be as clear as possible.

It means that you need to explain exactly what the product is, what it does, and who is it for, and in which case, or what is the real use case for this product.

So this is what the AI system is looking for.

Also, it is very important that the product data are complete, so that all the attributes or any specification, technical specification about this product or any details are complete.

Consistence is one of the most important things.

So you need to tell a story across marketplaces.

Mostly sellers are not only selling on Amazon, for example, they are mostly selling also on Kaufland or on ebay, whatever.

So the structure might not be the same one as for Amazon.

Maybe in ebay you need another structure or there are other data in the back end you need to enter, but you need to make sure that you don't contradict yourself.

So keep on talking about the same thing, use the same language, and so be consistent in what you are describing.

That's an interesting point.

I'd just like to pick up on that because in the first part you were talking about, I was like.

When you were describing what the eye is looking for, I was like, in my head, yeah, but that is actually just what a human consumer would be looking for also.

So all the information that he needs to come to a shopping decision.

So my first assumption would have been if a product ranks well in a marketplace, then it should actually also rank well for an AI.

But now you're saying we have to look at the bigger picture and create an overall story over all sales channels that are consistent.

And if they are not, then we might confuse aii.

Is that true?

Yes, that is correct.

Of course, ranking plays a very big role.

It means that you have good chance to be mentioned or to be recommended by the AI if you're ranking well.

But it's not a guarantee for that because the way these systems are working, you know, within a marketplace, within a marketplace, there is an algorithm ecosystem working differently than the AI system is working.

You know, the algorithm ecosystem is looking for internal data, how often you are selling a product, how are the reviews within this marketplace, and so on.

But the AI system is looking more broadly, is also looking outside of the marketplace.

So it's looking from information about your product, from other sources, from YouTube, from Reddit, from Your own website, whatever.

So it is a different way of looking at the product data.

That's interesting.

Let's go a bit more into the detail of what that actually means for marketplace listings after a very short break.

If you are looking to expand your marketplace business internationally, for example to Poland or Italy, it might be worth taking a closer look at Kaufland Global marketplace.

Through Kaufland you can sell across now seven European countries.

That's Germany, of course, but also Poland, the Czech Republic, Slovakia, Austria, and since last year, Italy and France.

And the best part, you can manage all of these marketplaces through one single account, so you don't have to set up your products separately for each country.

This makes international expansion much, much easier.

You can test new markets, build, reach and scale your marketplace business step by step.

So you don't have to decide right now because our offer is valid until May 31, but still, you know, consider it quickly.

And if you'd like to learn more about how brands can build visibility and branding on the Kaufland marketplaces, make sure to listen to our upcoming podcast episode.

There we will have guests from there will be Marcus Anstant, team lead commercial development, retail media, Kaufland, and very interesting, Jan Weiss, senior e commerce manager at the Brand Sleeping.

And together they will tell us how a brand can.

Yeah.

Show itself on the Kaufland marketplaces.

So as always, you'll find the link to our bonus in the show notes and make sure to tune in next week.

Barry, you've been talking about how important it is to tell a story all over the sales channels if you want to convince an AI of your product.

But is that actually possible?

Because you also mentioned that as of now, most marketplaces ask sellers for different things.

So your listings on Amazon or on ebay or on auto or wherever might naturally differ from each other because you have been optimizing for the search engine of that particular marketplace.

Is that something that we can actually get around as marketplace managers or how do we deal with that if you still want to want our products to be picked up by the AI agents?

Well, you need to think about the purpose of the whole.

For example, it is very, very important that you are, as I mentioned earlier, it is very important the consistence of what you are writing about.

Like for example, let me Go a bit into the practical view.

A practical view so that I don't become too abstract.

When you are writing a product description, you need to be consistent.

For example, if a product has some weight, or, I don't know, colors, variants, whatever, you need to write the product data in the same way consistently.

Like for example, if you write somewhere of 500 dram like in G500G, you need to always keep on this same formatting, for example, so that the system does not see any conflicting inputs.

Also you need to go or be more specific when you are describing a product.

And also at the same time, it's very important that as you mentioned, these are basically points which are very, very important for humans as well.

So it should not be new that these points are important for a product description.

But this is indeed something that many Listers are missing today.

Mostly it is because of the quantity of all the data which need to be entered inside the system.

Like for example, my advice is there are a lot of fields to be completed in the back end and many Listers do complete only the necessary ones, for example, or the ones that the platform tells you you must fill them.

So a lot of other fields are remaining empty.

My advice is to fill, for example, all the fields, because even this is also the base for the algorith rhythm within a marketplace to recommend your product.

I'm not sure if I answered the question.

If this is the question you've asked.

Me, I think that's fine because that's where we're going for, isn't it?

Like, what does listing an AI ready product listing really have to look like?

And on the other hand, are the marketplaces already offering this possibility with their internal systems?

So are the marketplaces themselves already now AI ready in a way?

What would you say?

Partly, partly indeed.

Like for example, okay, let's speak about Amazon, right?

Amazon is trying to close this gap unevenly.

So like for example, it's using or there is the chatbot on Rufus within the platform Amazon and Rufus.

Is taking the information within the platform to give recommendations to the buyer asking for a certain product.

So it's giving a very personalized experience within the platform.

But my advice is not to wait for platforms to make this type of optimization or to try to solve this gap still existing today.

But that sellers really need to start enriching their product data, as I mentioned, beyond the minimum, what a marketplace requirement is, because first of all they will be better positioned within the platform, the marketplace itself, and at the same time AI driven systems will discover them Better.

So yeah, it is.

While these platforms are mostly are existing since many years, marketplaces like Amazon, they are existing since many years.

They are very good within the internal structure so that the customer experience within this, this platform goes very smoothly.

But these AI systems also need to look also many, many things on the front end.

So how you are describing a certain product, what context are you giving to the AI systems?

And at the same time they are also looking outside of the platform.

Let's, let's dive maybe a bit deeper into Rufus as you have now mentioned them and as they, this is from my knowledge, the only real functioning in marketplace AI at the moment though.

Functioning.

I mean, yeah, there have been complaints about Rufus a lot, but yeah, let's talk about them.

Are there any insights to be gained already on what Rufus is looking for?

Because I read a few reviews from sellers and complaints from sellers as well that show that sellers often don't seem to know what Rufus really wants of them.

So how does Rufus work at the moment?

What is it looking for and how can Amazon sellers in particular cater to that?

Well, first of all, Rufus, if we are talking, if we are talking now about Amazon sellers, right, Because Rufus takes the information only from Amazon.

Yes, not looking outside of Amazon.

So basically there is a very strong correlation, I would say, between the ranking within Amazon and the recommendation Rufus makes.

But most important for Rufus, it is that it is looking for descriptions that answer buyer questions.

So not just futures, but you know, it is looking for like for example, if a user enters, I want to buy hiking.

No, no, no.

A running shoe for the city, for example.

Rufus needs to be confident enough to find within the list of running shoes in Amazon the running shoes for running in the city, not for example, somewhere else.

So.

This is why I was enhancing earlier today that the informations you enter in Amazon or in ADR in any other marketplace should be as complete as possible.

The more complete and structured the information you give to this platform, the better the recommendation is.

So use all the, the fields you have available so that the system is able to interpret what your product is about and also use any enriched or yeah, any content field which you can enrich so that Rufus can get better insights like for example a content or any other comparison modules or videos, if you are able to upload any whatever.

So the more information you give, the more structure, the more clear, the more consistent, the better it is for Rufus.

And then we are back to storytelling in a way, isn't it?

Because we are not just talking about the figures that make up that running shoe, but also about the use case behind it, in this case running in the city.

So what do we do with products that have several use cases?

Do we create a storyline for each of them so that the AI can pick up on them?

It depends on the possibility you have.

But like for example, if you are now using for example Amazon and you have there the possibility to write use cases there, you can write all of them there.

Because it depends on, you know, it depends first of all on the type of shoe.

Right.

There are, for example, let's keep on talking about the shoe.

Okay?

A running shoe for the city.

This is a very specific use case.

You should avoid talking generic, like for example high performance running shoe, breath upper cushioned sole.

This is something you can write for any shoe, right?

Any shoe, any running shoe.

But if you know that the type of shoe you are selling it is one that it is very specific for running on a certain sole, I would say then describe, describe the details, what is the drop and under which conditions does it perform well so that the system can work with this information.

So basically there is always a way to find the main to any product serves mostly a main problem.

So try to focus on this main problem so that you keep on telling this very insane story or true.

Always try to match the intent.

So buying whom it is both your shoe and for which situation, which type of problem solves your shoe and then match it to the intent of your buyers and describe according to this.

Because an AI is basically an intent machine and not a search engine, isn't it like they're looking to fulfill the intent of their user and then providing them with the product that might fix that intent while a search engine is looking for a product.

So that's one step ahead.

Yeah, maybe enough remaining time.

Let's turn this a bit around because we have now been talking about how to present your product for the AI agent to pick it up.

But of course to AI, there's for marketplace managers the other perspective as well.

And that is how they can use it for their own marketplace operations.

And I know you're helping customers with that as well.

So what is possible today if you use AI in your marketplace operations?

How much can you actually automate with it?

Well,.

The specific tasks you can best automate within a marketplace are for sure tasks which are very repetitive and which needs to be structured perfectly and are of high volume.

So at scale, let me be more specific, there are a lot of sellers or listers listing products for their clients who are listing even millions of products.

Within a platform, within just one platform, with just one language for one country, mostly the same ones are listed also in parallel for other marketplaces which follow other rules, or even for other countries which have then different, different needs.

And so this becomes mostly, I would say impossible for a human to manage this in a timely manner.

Of course it is always possible.

It has been done during the past years manually.

But you can imagine the time to be online, how long it passes before you have even just hundred or, or thousands of products listed live in a platform.

So the time it passes is very, very long and is mostly today not really affordable.

So if you want to keep all this huge amount, high volume of products, of listings as specific as possible, as consistent as possible across of hundreds or thousands of products.

And realistically, I'm not saying manually, but realistically, you really need a very systematic way to structure and enrich everything across different sources, across marketplaces, across languages, across countries at scale.

Sorry.

So this is something you can automate very, very well within just a few clicks.

You don't need any more hours to work on that.

You just need to tell a tool for what platform are you optimizing, for which country you optimizing, and that's all.

And then the rest the system does it itself.

So within a few clicks you really have a very good product data within a system, within an algorithm ecosystem like a marketplace.

Also the data you need for this is not, it's just a part of the work.

You also need an automation to create very good listing at scale because one is the time you need to enter all the fields in the back end and the second one is the time you need to create a listing, which is for the search engine within the platform, which is for human, which is for AES system, which considers any internal keywords, whatever.

It's just, well, I would say mostly impossible to manage.

So you can reduce your manual work dramatically if you use an AI tool.

Also we are talking about translation, about market localization.

These are all things cases coming on top if you want to make a really a listing at scale on this platform.

So you always have to consider that the data you need for these AI systems or these AI tools to work correctly, they need to be qualitative, very, very high because the quality is the result of the data you have within this platform.

So all these are tasks you can automate within the platforms and which saves you a huge time, up to 90% of time when talking about these repetitive structured tasks.

That's a lot of time that you can then put into enhancing your product listings as you said before, up to the ninth degree.

So that AI is going to find it so slightly might just turn out yeah even in the end.

Okay Barbara, I always like to finish our podcast with what I call the next Monday question because let's again try to get this down to business, to really everyday business.

So imagine a marketplace seller has been listening to this episode today.

What is one actual concrete thing that they could start next Monday to make the product catalog more AI ready?

Well my advice is first of all look at your best selling products.

Are all the relevant attributes filled in?

What are all the data you're giving to the system?

If I don't know, Amazon is asking for 20 fields to be completed.

Are you filling all the 20 fields?

If not, make sure you do from now even the optional fields.

I'm also talking about the optional fields.

Secondly, make sure you give to the platform, to the marketplace consistent information.

For example, if you use units, formats, whatever, be consistent across all audio channels.

Also very important is make sure that all your products are not just described from a technical way, but are described also realistically for the buyer to understand in which situation this product can be used or which problems it is solving this type of product.

And also make sure that if you are selling across different platforms, across different countries or whatever, or different channels, make sure that all the information matches to each other, right?

That you are always telling the same story.

So data are very important nowadays and they have always been very, very important, but now even more so.

All the agentic systems from now can only rely on data and the quality needs to be really perfect for them to really help you selling more.

I'll take that as the final word.

Barbara, thank you very much for your insights today and for bringing this whole topic a little bit out of the ivory tower and into the engine rooms of the marketplace.

People, thank you very much for joining me.

You're very welcome.

Thank you.

Where can people meet you if they want to talk to you about AI?

Well, I'm on LinkedIn so feel free to contact me on my LinkedIn page.

But I will also start soon a podcast in regards of AI within the E commerce and I will give some more information in my LinkedIn page very, very soon.

And of course we have a www.icom.com and of course you have my contact details there and I will be very, very happy if you contact me, of course.

Okay.

Right.

Thank you very much and thank you to everyone who was listening to this.

I hope that was helpful and you now know what to do next Monday to get AI ready.

Thanks for joining us.

And if you liked the episode, do like subscribe and comment on LinkedIn.

We always happy to hear your in.

So thank you very much and bye.

Bye.

Thank you.

Bye.

You listened to let's Talk Marketplace, the Marketplace podcast with Ingrid Lommer and Valerie Dichtel.