Latest / Deep Tech Germany – AI, Robotics & Frontier Innovation by Startuprad.io™ / Agentic AI for Retail: Zenline's Assortment Systems & Margins

Transcript

Agentic AI for Retail: Zenline's Assortment Systems & Margins · Deep Tech Germany – AI, Robotics & Frontier Innovation by Startuprad.io™

0:00
European retail is in a margin war. It didn't choose.
0:04
Assortments move faster than teams can analyze data
0:07
quality collapses on the scale and global platforms like
0:11
Amazon and Temu operate on a level of speed that
0:15
exposes every structural weakness.
0:18
Abba SETI built Zenline AI
0:22
to close that gap. Agents that act not
0:25
observe systems that detect margin leaks,
0:29
resolve clutter and surface opportunities
0:33
humans never see. Today we examine why
0:36
autonomous decision agents are becoming to operate the system
0:40
of modern retail. And what happens when a retailer
0:43
shifts from dashboard analyzers to product level action
0:47
in seconds. By the end, you'll understand why the next
0:50
competitive edge isn't scale its velocity.
1:00
Welcome to StartUpLead IO,
1:03
your podcast and YouTube blog covering the German
1:07
startup scene with news interviews and
1:11
live events.
1:14
Hello and welcome everybody to Startup Rio. Today my guest is
1:18
Arba Seti, the CEO and co founder of
1:21
Zenline AI from Switzerland, Goodsy
1:25
Goody, a company that is redefining how European
1:29
retailers decide what to sell, what to move and where
1:32
the real margin sits. With experience at BCG,
1:36
Procter Gamble, ETH Zurich and Research work at
1:40
Cambridge, he brings both the strategic clarity and the
1:44
technical depth required to rebuild assortment logic.
1:47
From first principles. Zenline agents
1:51
clean messy catalog data, detect cannibalization,
1:55
map substitutes and reveal trends before they
1:59
become mainstream. They already process more than a million
2:03
SKUs and deliver recommendations with measurable
2:06
financial impact enable retailers to operate with the
2:10
speed of global platforms without the overhead.
2:14
Aber is here because his work signals a structural change.
2:18
Decisions that took once weeks now happen in
2:22
minutes and the retailers who adapt will redefine
2:25
competitiveness in Europe's Euro trillion
2:29
market. That was an intro, don't you
2:33
think? Welcome to the show and let us
2:37
get right into it. You said the real power
2:41
center of retail isn't price or promotion, but
2:44
but the assortment itself. When did that
2:48
insight crystallize for you? In my previous
2:51
job with bcg, we were advising retailers with assortment optimization.
2:55
And what becomes very obvious when you
2:59
work with retailers on the assortment is that the assortment touches the entire
3:03
organization. So it's not only the top line or the bottom line as is
3:06
for pricing and for promotion, but
3:10
certainly it also touches the top line. So if a customer, for example,
3:13
doesn't find a product that it wants in the stor stores or on the
3:17
shelves, that's lost revenue. However. Moreover, it
3:21
really defines your brand. So the brand is typically
3:24
not only shaped by the marketing, but
3:28
it's mostly shaped by the products you offer. So you can imagine when
3:32
a consumer has a demand, what do they first Think of what's the
3:35
brand, what's the retailer that they think of to go and
3:39
buy the product from? And that's so important to be top of mind for the
3:43
consumers. And, and that only works with an assortment. That consistency
3:46
delivers what the consumers actually want. And then the
3:50
last one is the assortment. And the depth of the, and the
3:54
breadth of the assortment influences the
3:57
complexity of your operations, the logistics, the supply chain, the
4:01
inventory management. So the broader the assortment,
4:05
the higher the complexity and the more difficult it is to maintain
4:09
lean operations. When you talk about
4:12
assortment, you have something in mind. For example, when I was
4:16
thinking about this interview, what I had in mind when I was reading this,
4:20
at first I was thinking, okay,
4:23
option one is I look for the
4:27
basics. That is a retailer with a yellow logo. Then I look
4:30
for some more fancy stuff. Then that comes in a
4:34
retailer with the orange logo or a red logo.
4:38
Is that something that you're thinking about here? Yes,
4:41
exactly, that's exactly it. What comes to your mind when you want to
4:45
buy a product, what is it? Is it someone who's super, super trendy
4:50
and always knows what the consumers want and can offer it, offers it
4:54
fast, or is it typically 6 to 12 months behind
4:57
the schedule? Your early
5:01
discovery work exposed how little transparency
5:05
retailers had into their own margin drivers, what
5:09
revealed the debt of that blind spot.
5:13
So the importance of that blind spot comes basically
5:16
what you said in your intro. It's a super, super, super competitive market
5:20
with Temu and Shein entering the European market not only in
5:24
fashion, which a lot of people think, but it's across all categories from electronics to
5:28
fashion to pet supplies. And retailers
5:31
are too slow because of the analysis, just take too
5:35
long, which should be done automatically and systematically with
5:39
AI. So basically we're entering
5:42
this competitive, basically war
5:46
in retail coming from different countries from different angles.
5:50
And if retailers don't move fast
5:54
enough, don't filter out products that have bad
5:57
margins and focus on the ones, focus on the entire assortment. Instead
6:01
of managing just 10 to 20%, European retailers will
6:05
not be able to cope with the speed and with the
6:08
velocity of these big platform players.
6:13
So that you're mostly targeting
6:16
this towards retailers that would
6:20
compete with Amazon, Temu, Shine,
6:24
Alibaba Express and so on and so forth.
6:27
Yes, we're trying to basically help them to have the
6:31
same data, science and AI capabilities as
6:35
these platforms have, who have actually their own internal assortment
6:39
teams to make these decisions. And you're
6:42
now entering a market that is basically shaped
6:46
by Amazon level speed. You order one day, you get it,
6:50
the Next day or a little bit later. What convinced you that
6:54
Europe's response had to be agentic and not
6:57
incremental? So what's happening right now is a perfect storm.
7:01
We're seeing years of inflation, declining consumer confidence,
7:05
increased saving rates and what's making it now
7:08
very hard for retailers to stay ahead the curve and stay
7:12
profitable. And this does not only apply to
7:16
retail, right? We're speaking about all industries that are
7:20
basically undergoing this wave of the industrial
7:23
revolution or let's say revolution led by general purpose
7:27
technologies. And in these times of general purpose
7:31
technology is changing not only one but multitude of industries,
7:35
the answer has never been incremental innovation. And US and China will
7:39
move ahead and they will basically equip millions of white
7:43
collar workers with agentic AI. And speed will be the
7:46
differentiating factor, launching new products, understanding
7:50
the assortment, understanding relationships in the assortments much, much, much
7:54
deeper, much earlier than all competitors will.
7:59
And again just to to basically highlight that if we
8:02
still think we can do the same, the
8:06
same thing that we did for the past 20 years, which is incrementally
8:11
increase efficiency, these companies will
8:14
basically frog leap the European retail industry.
8:21
In the entrepreneur's world, we'll be talked
8:24
about the moment that data forced a decision
8:28
you didn't expect. Stay tuned and learn
8:32
more down here in the link.
8:36
We've been talking category teams already that some of our large
8:40
competitors have. They focus on top
8:43
sellers and ignore the long tail. What breaks inside
8:47
an organization when the long tail suddenly becomes
8:51
visible through good data, through AI, real time
8:55
observations? I would say the first reaction is
8:58
inconvenience. There's a lot of things that get exposed, like
9:02
data flaws, wrong images linked to wrong products which
9:06
then lead to very high returns because consumers got the
9:10
wrong product which they actually didn't want to order. But the product
9:13
image on the website was leading to the wrong product
9:17
attributes that are misleading or misconfigured or inconsistent
9:20
pricing and all of these errors, or
9:24
let's say inconsistencies,
9:27
misconfigurations in the assortment become
9:31
very obvious when you start focusing not only on the
9:34
1015 to 20% of the assortment that is typically
9:37
managed very well, but then
9:41
surfaces and this then creates, let's say
9:45
inconvenience at the beginning. But
9:49
organizations then learn to accept the fact that it was
9:52
very, very hard to manually be able to manage
9:56
hundreds of thousands, if not millions of SKUs. But now with the
10:00
power of AI, you can actually do what retail
10:05
wasn't able to do for the past. Basically for the past decades.
10:11
You'Re the agents you provide for the companies, they do
10:14
challenge internal narratives. Which narrative
10:19
collapses the fastest once real data
10:22
pressure was applied? You said first the flaws are exposed and
10:26
then I think the next step is their internal narratives breaking
10:30
down. Yeah, very, very, very good question. And I would
10:33
say the most common narrative is the intern.
10:37
The most common internal narrative is that the consumers know
10:41
exactly what they want to buy. And this consequentially leads to
10:45
an increased breadth of the assortment. And we as read retailers,
10:48
we don't have power to because the consumers have all the power
10:52
and they will either way just buy what they want. If you don't have
10:56
it, they will not buy. However, data that very, very
11:00
early on will show that curated
11:04
assortments is actually what consumers want. They don't want
11:07
decision paralysis, but what they want is a curated
11:10
assortment. Exactly. Curated to their
11:14
demands and their basically the consumer behavior. And
11:18
what you typically see immediately is that revenue and margin
11:21
goes up significantly. We're speaking 1, 2, 3 percentage points
11:25
of margin when the
11:28
retailer makes the decision of what should be in the store, in the stores
11:32
instead of basically giving up and saying it's always, it's the
11:36
consumer who's always right and we need to offer everything that they
11:40
can possibly imagine.
11:46
Part of your thesis or your main
11:49
part is speed. What's the cost of
11:53
operating a full year behind the real markets decision
11:56
Tempo. I just mentioned it. We're speaking 1, 2 percentage
12:00
points in margin improvement and we're speaking 1 to 3% in
12:03
revenue increase. Very often we're not even
12:07
one year behind the real market, especially in the private
12:10
label segment of the businesses. And in Switzerland, I think for the
12:14
first time, private label revenues surpassed 50% of the market.
12:18
So private label is now making more revenue than the branded market
12:21
for retailers. And in Germany, in the rest of Europe,
12:25
I think it's close to between 30 to 50%.
12:29
And there we're speaking very often about 18 to 24 months
12:34
of let's say distance to real consumer,
12:37
consumer changes. I do have
12:41
in mind that basically we talked about here the
12:45
less fancy products, especially talked about food here
12:48
and clothing. There's a big online retailer who offers
12:52
essentials and, and that's kind of stuff we talked about here,
12:56
that that's retail brands, that's the brands they come up with themselves
13:00
and for some larger specialty food retailer, they are meant by, they
13:04
are produced by household names. So that' so
13:07
pretty interesting area to watch. But let us go a little bit further
13:11
and talk about
13:15
what impact the data had on teams because you've Watched
13:18
messy data, paralyzed teams. What's the silent
13:22
limiter they underestimate most?
13:29
So I would like to answer slightly
13:33
different question which is, and I'll get back to yours, which is
13:37
companies very often think that data overall
13:41
is the limiter of, for the generation of the value
13:46
margin, improvement, revenue increases, making faster
13:50
decisions, and then they very often focus on the 5
13:54
to 10% of the data that is imperfect, where the master data
13:57
is not clean, where the attribute data is not clean. And then
14:01
organizations often stagnate and move very, very, very slow
14:06
because they think of, oh, data cleaning being this one
14:09
incredibly big IT exercise that they need to perform.
14:13
However, what should happen is that
14:17
while doing these use cases, you clean the data as
14:21
you need them because you will figure out where the data leads to. Wrong
14:24
suggestions, wrong recommendations, and then it
14:28
makes sense. And then there's also a business case to clean up
14:32
the data and make it worth this exercise.
14:36
And what a lot of companies and retailers have not realized yet
14:39
is that agentic AI is making data cleaning and insight generation
14:44
so feasible. The cost to or the cost and the time
14:48
to get cleaner data is reduced on a daily basis
14:51
and retailers can already save millions in margin
14:55
without having the perfect data. So instead of focusing too much on the
14:59
limiter, I think companies do that either way all the time.
15:03
The message should be although data can be messy
15:07
and too much data can paralyze, the
15:11
limiter is very often the mindset towards the data.
15:19
When we talk about agents, many people
15:23
have absolute neutral
15:27
AI in mind. But every agent encodes
15:31
a judgment, either by the person who decides what to
15:35
encode or the person who does the encoding. Which
15:39
judgment was the most difficult to translate into
15:43
an autonomous logic?
15:47
Very good question. And I have to say
15:51
the entire space is very, very, very early and it will change in a
15:55
year from now. But
15:59
what is happening right now is that agents are very, very, very
16:03
sensitive to a few sentences in the prompt in how they
16:06
make the final decision. So because of
16:10
that, as you were saying, rightfully, the
16:13
design team of the agent, of the system prompt of
16:17
the context engineering that allows the agent to make a decision
16:21
of which products to recommend. One very
16:25
simple example might be that one of our customers
16:28
doesn't want to be the most trendy firm. So they don't want to just launch
16:32
products that are, let's say, that
16:36
are just coming to the market, but they want some validation, some proof points,
16:40
and to know that this product is actually performing and will be here two, three,
16:43
four years from now. And now the question is, what is A trendy product,
16:47
when does the trend turn into something more sustainable?
16:51
And in the end, the makers, which we are
16:55
here with Zenline, need to be very, very, very thoughtful
16:58
of how do we encode that? And I think
17:03
the most difficult part of this is the memory layer. So
17:06
you do one recommendation and then the question is next week,
17:10
next month, next quarter, how do you encode all the recommendations
17:14
that you made? How do you encode, let's say, the
17:17
feedback from the real world? Did the product launchers
17:21
succeed or didn't they succeed? How do you bring that in and how
17:25
do you feedback that into something that the agent
17:29
can use to then be able to make better
17:32
recommendations next time? Because this is something that is incredibly natural to
17:36
us as humans. We do something, we, we observe what happens
17:40
and we have an initial gut feeling of whether we should do that again or
17:43
what we need to adjust. And this is something that's very hard to, to encode
17:46
into. Into autonomous systems. Yes,
17:51
there's a lot of stuff to really hard to encode in
17:54
autonomous systems. And I think the more systems
17:58
we'll have, the more it will be
18:02
of importance. But let us
18:06
go to the next question, because I was wondering,
18:11
your launch validator runs on social signals,
18:14
competitive shifts and emerging demand. When
18:18
did you trust its signals over the human
18:22
instinct? So what's
18:25
important to distinguish is whenever you launch a product,
18:29
which is what we aim to support with the launch validator
18:33
is we want to strengthen the human
18:37
instincts by giving it more data. However,
18:41
as every single human being, category managers, for example, in the
18:45
fashion world, are biased by what they see on their
18:48
channels, their social media channels, by walking through the city, watching
18:52
colleagues, friends and so on, what they wear, and
18:56
then using hundreds of thousands of data points
19:00
actually allows them to double check your gut feelings
19:04
and use these data points from social media, from Google shopping,
19:08
from competitors, bestseller data, to see whether
19:11
your initial gut feeling was right
19:15
or whether you were missing something. And it can be much, much, much more
19:19
nuanced, especially in the fashion world, where new designs
19:22
from quarter to quarter, from year to year very often change
19:27
in subtle ways which are very hard to grasp. So it's not
19:31
trusting the signals over human instinct, but it is challenging
19:35
the human instinct, which can be biased by their own world, which is very often
19:39
not representative of, let's say, the
19:43
consumers, what they want in certain categories. Always,
19:47
always when it's working in consulting and
19:51
always if you do have strategy project at one point, there's always
19:56
the question of internal cannibalization. So
19:59
if the, if I'M adding product
20:03
A, does it take away from B, does it take away from C?
20:07
And your consolidation agent exposes
20:10
internal cannibalization. What's the most
20:14
counterintuitive callization call that
20:17
surfaced? There's always revenue
20:20
cannibalization and there's margin cannibalization. And the
20:24
revenue cannibalization is basically where a lower priced product
20:28
is purchased by the consumer because
20:31
although there is a duplicate which is maybe higher priced and
20:36
is the one product that you as a retailer would want the consumer to
20:40
to purchase. And the issue here is that
20:44
revenue cannibalization is very often not counterintuitive.
20:48
It's very intuitive. However, it's never in the data. I haven't seen a
20:52
single retailer that has good relationship model across
20:55
tens of thousands, hundreds of thousands of products to know which products
20:59
are very similar because you would need a very good
21:03
relationship mapping of all products, similarity scores,
21:06
quite good attribute data to figure out which products are
21:10
closely related to each other and what is the impact of one to the other.
21:13
However, margin cannibalization is very often
21:16
counterintuitive. You take for example two
21:21
examples that we had are drinks where
21:24
branded products that are higher priced
21:28
had significantly worse margins than the product than the
21:32
private label brands that are cheaper and have higher margins.
21:37
So in these situations what happens very
21:40
often is that companies
21:44
do not understand sufficiently the
21:48
impact that let's say price differences and then the relative
21:51
margin of these products has on in the end the
21:55
purchase decision of the consumer and then incentivize
21:59
the consumer to purchase the wrong product
22:03
by for example, reducing prices by watching
22:06
the market too closely or trying to increase
22:10
velocity of the product. We'll be back after a
22:14
short ad break. And after the break we step into the
22:17
strategic misread that holds retailers back and
22:21
where the next advantage isn't data but decisiveness.
22:31
Hey guys, welcome back to our interview with Arbor from
22:35
Zenline. Helping retailers to increase revenue
22:39
and margins. Working with AI here and I was
22:42
wondering, your weekly decision cycles replace
22:46
seasonal ones. What operational structure
22:49
fails first when a retailer attempts that
22:53
transition? And also a little bit tongue in
22:57
cheek question, when are we at daily
23:00
decision cycles? Very
23:04
good question. I have to say we
23:08
have spoken to many retailers and we're working with many retailers where the decision
23:12
cycles can sometimes take more than 12 months. We were
23:16
seeing very good examples and we're seeing bad ones.
23:20
The operational structure that fails very often is the
23:24
relationship, especially in the private label business to the
23:27
suppliers where it can take up to 12 to 18
23:31
months to actually supply the product. So from
23:34
let's say initial ideation to concept development
23:38
to iteration to receiving the stock that's required
23:42
to sell the product, sometimes it takes between 12 to 18 months, which
23:46
in the end makes AI driven quick decisions
23:49
completely obsolete because you don't need it if your supply chain process in the end
23:53
can't deliver what you're trying to do organizationally in
23:57
terms of commercial output, delivering the assortment that consumers want.
24:01
However, with
24:05
many of our retail customers, we're also seeing very good examples where the
24:09
operational architecture actually is already in place. But the
24:12
bottleneck was never the supply chain delivery but, but it was the
24:16
analysis and the inside gathering because it's bottlenecked at the category
24:20
managers. And if the category managers simply can output
24:23
more decisions, the let's say supply chain
24:27
behind it is already supportive enough, is already, let's say
24:31
established so that they can quickly start to
24:34
see, see these products being in the
24:38
shelves within the weeks. On the question when are we on daily decisions? I
24:42
think daily decisions will not be the bottleneck,
24:46
but daily supply chains will be very hard to achieve for
24:49
retailers in most industries. If we wouldn't be on a time constraint,
24:53
I would go into
24:57
deviate a little bit and go into some thoughts,
25:00
experiments, what we could do with Jones here. But let's take
25:04
our main line of interview questions here.
25:08
Rejected recommendations reveal a lot about the
25:11
organizational psychology. What patterns do you see in
25:15
teams that aren't ready for agentic systems? I would
25:19
say organizations that focus too much on the errors that happen,
25:23
or let's say the parts of the
25:26
recommendations that are misleading, that are error prone, that
25:30
are just simply not useful at the beginning. And
25:35
organizations that focus too much on what's not working
25:39
and where the data is not sufficiently good are
25:42
the ones that will become the laggards
25:46
of the, of the decade. So however, on the other
25:49
side, the teams that are utilizing agentic
25:53
systems most effectively basically treat them as junior
25:57
analysts, I would say. So you have a junior analyst. What happens if
26:01
you receive a lot of input that will be good, some will be exceptionally good
26:05
and some will be bad simply because it misses context, it doesn't know
26:08
exactly what has happened in the last 12 months of the organization.
26:12
Sometimes recommendations are bad because they have been done,
26:15
they have been tried in the past, maybe 12 months ago, 24 months ago.
26:19
And certain products, certain product types or categories just don't
26:23
work with the retailer. And a
26:27
junior analyst also wouldn't know them. So what very
26:31
good, what patterns of very good teams are
26:34
is the willingness to give feedback, give
26:38
nuanced, and detailed feedback to advance the memory of the system and
26:42
give more context into the system, define the strategy, the company
26:45
strategy, the category strategy, and thus be able then
26:49
to improve it over time again, as you would do with the junior analyst. And
26:53
I think the more organizations start to think of
26:57
agentic systems of colleagues, of basically just another part of
27:00
the workforce, the faster they will be the ones that actually
27:04
gain margin benefits from it.
27:09
Your pricing agents benchmarks
27:13
internal logic against the market reality.
27:17
What's the structural pricing error that retailers report?
27:23
Okay, so I haven't seen a single retailer yet who
27:27
has been able to be very strong
27:30
in utilizing internal relationships
27:34
in their pricing. So what happens very often is that many retailers
27:38
are very much focused on hey, how is this one single product
27:43
priced compared to the market to other other
27:46
retailers selling this one product? What every
27:50
retailer does is they compare their own assortment
27:53
very much with the assortment of other retailers.
27:57
And what they neglect is that a lot of the pricing,
28:01
a lot of the pricing power that they have is not
28:04
comparing themselves to competitors,
28:08
but to the internal assortment. Because very often when a consumer is
28:12
already in the stores, they will make a decision
28:15
not based on what is the price of the competitor, but they will make a
28:18
price based on what are the pro, the other products, the
28:22
alternatives, the substitutes that are already in the shelf.
28:26
And especially with regards to decisions that impede the margins,
28:30
customers just retailers typically don't have a, let's
28:34
say, sophisticated enough system to
28:37
understand the relationships. As you mentioned before, the
28:40
cannibalization impact of not only delisting a product, but also
28:44
changing the price of one product or what does it do to all the other
28:48
products in the assortment. And this relationship model is something that
28:52
retailers should think about. However, they're doing it
28:56
way too little because it's very complex technically.
29:01
So the classical substitute products
29:04
substitute variant. Pricing as soon as you get to
29:08
10,000, let's say 10,000 products. It's
29:11
very hard to manage a very clean relationship mapping of your
29:15
products so that for example, when you change one product, you also change
29:19
either in, let's say proportionally
29:23
or at the exact same level. The price of variance can also
29:27
be the same for substitutes to just understand what does the change
29:31
of one product's price due to the rest of the, let's say,
29:34
very similar, very related products.
29:39
We've been talking about margins a lot and leaders often
29:43
freeze under margin pressure. What differentiates
29:46
those who accelerate versus those who stall?
29:52
I think a lot is in the tools. Some
29:56
retailers still consider power bi as state of the art
29:59
and A lot of them still stick with Excels, which is
30:03
very convenient because they don't have to learn new tools. But the
30:07
spreadsheet logic is failing. It's failing as soon as you have
30:10
hundreds of products, not even mentioning
30:14
thousands or millions of products. And I think retailers
30:18
just need to accept the fact that let's say human
30:21
driven and spreadsheets driven
30:25
working modes are not optimized
30:29
to support retailers where decisions are
30:32
incredibly complex. And yeah, in the end,
30:36
competition will not wait. Competition from other markets,
30:40
new more digital versions of retailers popping up. And
30:44
the margin winners will be the ones that rewire how
30:48
fast product decisions can be made because they have the data
30:51
ready, readily available, and then they have the systems that
30:55
output recommendations and then review the
30:59
recommendations. So basically the shift from going I need to do
31:02
the spreadsheet analysis myself to I
31:06
only review the output of systems and I can then
31:10
fine tune for a few hours instead of being the one who has
31:14
done the analysis for weeks.
31:18
I know how this feels.
31:23
You gotta be creative here. You guys
31:27
scaled at a pace even few teams
31:30
attempt which moment
31:34
demanded your highest level of personal, personal resilience.
31:39
Honestly, there, there are many. As a founder, you constantly
31:42
are on the edge, but when you have a strong team,
31:46
you, you can go through very, very, very long hours
31:50
together. And I think that's the happiest and the
31:54
best parts. But for me it's a combination of several
31:57
countless sleepless nights and pressure to keep customers happy.
32:03
Yeah, but I would say whenever
32:07
the team is willing to go the extra mile, is willing to
32:11
go beyond, to basically be able to deliver
32:15
the results that we are promising with the agents, that's the
32:19
moments where let's say, well, that's our
32:23
source of energy.
32:27
Your source of energy. Okay, I get the
32:31
feeling that Asian driven operations are
32:34
here compressing time, which is just talked about quarterly to
32:38
weekly in the interview. Which
32:44
legacy workflow becomes obsolete here?
32:49
First.
32:53
White collar
32:56
tasks that require desk research.
33:00
All the research work will be automated in the
33:04
future. And you can see this already in consulting,
33:07
which they have huge pressures. It's the same for
33:11
lawyers, it's the same for any kind of research assistants. Because
33:16
I don't know if you have used deep research from
33:20
Gemini and ChatGPT, but it's incredible how
33:23
much depth and how much data can be processed with these
33:27
tools. And they can do it significantly better than humans can.
33:31
So everything related to research and analysis will be something
33:35
that will be obsolete in the future. And humans will
33:38
switch from doing the analysis to making the decisions. Basically exactly what we
33:42
discussed before with the leaders of margin drivers
33:47
and compressing months of analysis
33:52
into days, if not sometimes even hours. Making
33:55
fast decisions, making counterintuitive decisions
34:00
makes explainability become something strategic.
34:04
How do you maintain trust when high impact
34:07
recommendation contradicts their intuition?
34:13
I like that question, especially in
34:17
an In a machine learning driven world that we had
34:20
for the past, let's say between 2016
34:24
to 2022, explainability was
34:28
really hard because you had a result.
34:32
Your machine learning model told you this should be the new price,
34:37
but there was very little explainability behind it.
34:41
What is now interesting with generative AI is that the
34:44
models have reasoning tokens and they can
34:48
output and they can explain why they came to a
34:51
decision. And this reasoning is getting closer and
34:55
closer to the reality of why the model decided to do what it
34:59
decided to do. And this is actually one of the things that we have realized
35:02
with customers is they love to see the thinking
35:06
of the model. So not only the final output of why you did it, but
35:10
even the steps, the thinking steps that happened, the tool calls that
35:13
happened, the web search, which websites did it browse, all of
35:17
that to make it very very very transparent.
35:21
So I think in the last 10 years there was a bit
35:25
of the dogma to
35:29
hide, a bit of the complexity behind it, to not confuse
35:33
especially the managers, but also the category
35:36
managers and also the executive teams. But now with LLMs
35:40
you can be very very very transparent about why the
35:43
model did what it did. And very often it's you can find
35:47
immediately in the prompt why it did the
35:51
decisions the way it did and then you can together with the customers
35:54
define the prompt to make it basically just be more
35:58
effective for their categories and their company.
36:02
At the time of recording and admittedly we were recording this in early
36:05
December, Europe has a 180 day window before global
36:10
platforms widened the gap. What is the one move
36:14
retailers must make now?
36:19
I might be a bit biased but
36:24
it's the super obvious one. Skip
36:28
all this digitalization
36:32
work. Just everything related to pure IT
36:35
transformation needs to be done in
36:39
parallel to all the agentic more business related
36:43
work that needs to happen immediately. No retailer has
36:46
the time now to wait for months and months and months if not
36:50
years to do IT transformation SAP for
36:53
Hana Transformations. But at the same time they should be already
36:57
working in the business. Not specifically with it,
37:01
but the businesses need to start working on agent based systems and see
37:05
what's already possible with them. Play around with it, find
37:08
very very highly valuable use cases that
37:12
will make an impact on the margin within four to eight weeks.
37:17
And I would say if retailers, especially in Europe and
37:21
the US just feel more confident to do these things, they will
37:24
see how advanced retailers are. And that's one of the things that we hear
37:28
very often. Typically the response is after
37:32
a month. They say, I would have expected AI to be three
37:36
to four years away from what you showed us. And I think this
37:40
is where you need to get the retailers to want to work more with
37:44
it. I've heard over and over that focus
37:48
determines survivability, especially for you guys. What,
37:52
which idea did you kill that unlocked your current
37:55
trajectory? In the beginning of Zinlin, we were focusing
37:59
on the. On a very similar problem, on the assortment problem,
38:03
but for manufacturers which have very
38:06
complex portfolios, sometimes hundreds of thousands
38:10
of SKUs. And as soon as
38:14
we started talking to retailers who are the buyers of those, let's say, complex
38:18
assortment, complex portfolios, we realized that the interest and the
38:21
traction that we received was. It was instantly
38:25
obvious. And we basically pivoted our entire company within three
38:29
hours to this new idea because of the traction
38:33
that we received there. And I think
38:36
being able to take very, very strong decisions and
38:41
let's say, reshifting the organization's focus, not only individual
38:44
focus, but the focus of the entire organization on a new topic, new type of
38:48
customers, and then very narrow scope of,
38:51
let's say, features. This is what makes the startup incredibly
38:55
powerful. And to be very honest, it's not about killing
38:59
ideas very often, it's just about delaying them. You know that customers
39:02
ask for something, you know that customers are willing to pay
39:06
for something, but you just don't have the capacity to serve them
39:10
in the quality that you want to show to
39:14
them. So basically you just, you have a roadmap and
39:17
you put it off until the features that you're building
39:21
are brilliant and you can scale them first before adding the new
39:24
ones. I just want to ask you this
39:28
again because I'm not sure I heard it correctly. You
39:31
pivoted once with a decision within three hours.
39:35
Yes. I think that's the future of business.
39:41
That actually ties in into my last official question. Your product
39:45
velocity is unusually high. What principle
39:49
anchors precision without slowing execution?
39:53
The two things that we do brilliantly, and it's also related
39:57
to the one before that is very,
40:00
very, very high focus on the most important features,
40:04
only the ones that are really meaningful to the customer. And then
40:09
we don't obsess with
40:12
technical details. We don't obsess with the beauty
40:16
of the code. We obsess with the customer value
40:19
and we Every single day we question ourselves, is this
40:23
the most impactful thing we can do that really saves
40:27
margin for the user? And we have it on our boards,
40:30
we have it in our notions. The one key metric that we track
40:34
is did we improve the margin of our customers?
40:38
And in the end, if we don't do that,
40:42
we're building a bad product. So this is what has helped us
40:46
to keep the product velocity very, very high.
40:51
I see. That was the official
40:55
questions and we usually end with
40:58
our two standard questions. Are you open to talk to new investors?
41:03
We are currently well funded, so we are not talking to investors,
41:06
but we are aiming to
41:10
speak to investors again beginning summer 2026.
41:14
Are you currently looking for talented employees?
41:19
Yes, we are. We are looking for the best and most
41:23
ambitious and boldest engineers. We have two,
41:27
we're thinking in two teams. At Zenline, we have product
41:30
engineers which are close to the customer, but in the end
41:34
they're responsible for building features and building the product. And then we have
41:37
forward deployed engineers, which is a term coined by Palantir, which
41:41
is engineers who are managing the
41:45
customer relationship. So in the end,
41:49
Zenline is built purely by engineers. However,
41:53
we then distinguish between the ones that own the product and the ones that own
41:56
the customer relationship.
42:00
Albert, you've shown that the future of retail belongs to
42:03
operators who move faster, see deeper
42:07
and trust systems built for action rather than
42:10
observation. You clarified where margin
42:14
heights, where complexity wastes energy, and
42:18
why agentic decision redefines competitiveness.
42:22
For those who want to follow your work or understand
42:25
Zenline's impact more directly, Visit try Zenline
42:28
AI or connect with you on LinkedIn. Thank you for bringing
42:32
structural clarity to sector that really needs it.
42:37
Thank you so much.
42:43
That's all for Find more news, streams,
42:47
events and
42:48
interviews@www.startuprad.IO.
42:53
remember, sharing is car.