# AI-Driven Data Democratization for Product Teams

**Podcast:** Stories Connecting Dots with Markus Andrezak
**Published:** 2026-04-01

## Transcript

Stories Connecting Dots.
A business podcast to discover a world of possibilities through stories told.
Ja, hallo, schönen guten Tag.
Heute mein Gast Nico Noll.
Nico Noll ist seit Jahren schon ein Produktspezialist, hat viele Leute in Produkt trainiert und arbeitet aber aktuell an einem eigenen Produkt.
Der Traum von uns allen, das eigene Produkt.
Nico is living the life, with all the ups and downs, which are always important to me.
You're always looking for Product Analyst AI.
There are already two things.
It's a Product Analyst, it's a thing that's with AI to do.
What would you first interest, Nico, is what is...
The current product, what is the current product, what is the line that you have seen with the product?
Hi Marcus and everyone, it's nice to be here.
I work at Product Analyst.di because we have identified that many of the Numerous data and the product data, which we use to make decisions, use to get to find, but mostly getracked and run.
And that AI can help us to increase the interpretation and the use of data to increase.
That means just simple things to answer like what do users do in the average, after they are in the average, after they are in the average, what is our rate of the average in the Abos, depending on the segment of our users.
That are so questions that we would like to ask.
We can't answer it.
We can't answer it.
We can't answer it.
We can't answer it.
We can't answer it.
We can't answer it.
We can't answer it.
We can't answer it.
where you put something on the side, that's also worth it.
We try to make a flexibility of definitions and statistical functions, so that the AI Guardrails has good answers based on the data that already has.
That's the idea of the product.
And that's what we're currently using with the pilot customers.
And where there is also a lot to build and find out what works well and how the UX works.
And we can talk about all of that.
But first of all, the need is clear.
That's pretty good.
And now is practically the phase of the process.
How can you get that for the people right?
Can you maybe a question of a question that I find especially...
good answered, also to better understand the difference between what tools could actually be able to answer or could be answered and where you then the Lücke further schließen.
And I think a part of it is Skills, a part of it is also Menschlichkeit.
Just like that, if I can't bore and do and do, could that be there, but it doesn't spring to me in the eye.
That you maybe this Spannungsfeld a little deeper explain.
Absolutely, very much.
Also, just the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the Nutzungs...
the And what we help is with the access to and the analysis of it.
That's when you ask a question, if you ask a question, which is what's going on, what's going on, if a customer stays or goes?
That's a question, which has every Abo-Modell.
And we have worked with the Kunden, we have worked on our pivot and worked on what else we have worked, what has been brought in these conversations.
And we were so, Ja, aber warum habt ihr denn die Antwort darauf nicht?
Und dann ist die Antwort gewesen, naja, wir haben keinen dedizierten Datenanalyst.
Oder der oder die ist immer überarbeitet.
Das waren die zwei Hauptantworten.
Und die dritte Antwort war, und selber kann ich es nicht.
Das heißt, ich weiß, ich habe theoretisch die Daten, weil wir tracken ja, wann die Leute sich einloggen.
Wir tracken, wenn sie das Feature nutzen.
Wir tracken auch, wenn sie abbrechen.
But I have not the skills, to do that.
I am a Gründer, I am a CPO, I am a...
And I myself have either the time or the skills, not that to do.
That is, that is very concrete the Lücke.
There are companies, which have them for themselves filled.
That is mostly a relatively expensive occasion, because they must have people for themselves to answer, which have every single question.
And as I worked as Product Manager at Xing, I had the...
That's a wonderful situation, that I had both a Frontend as well as a BI-Analysten available.
That's not possible in every company.
And we believe that AI can fill this Lack of the time.
Not for the big strategic Wahnsinns-Analyses, but for these questions like, which is what's going on?
That's the one thing that it doesn't have data, but the most...
They are more in a flood of data and events, which are tracked.
We know what?
So, before 10 years was it was different.
But now track all the people data.
That's not the bottleneck, but now, from the data, information, information, information, decisions to make, is now the bottleneck.
That's also an experience, which I share.
Also, my best friends in the firm are often data engineers, data scientists, with whom I did exploratory sessions for data.
so eine Augenöffnung Entdeckung gemacht habe, nachdem wir nach 50 Winkeln gesucht haben.
Jesus Christ, das war eigentlich immer klar, aber jetzt wissen wir es, das und die, weil und so weiter und so fort.
Das heißt, da ist ja auch dieses explorative Element drin.
Erzähl doch mal, wie die Technik das Problem löst, was ich jetzt, keine Ahnung, in mehreren zwei Stunden lang Sessions mit...
with a data scientist, how it comes to the technique now relatively quickly and on the point the pattern shows.
That's often things like, what we often say, after the first three sessions, six sessions, how it goes after 30 days, 60 days and so on, so je nach Model that you have to sell.
How does it work that your technology, the pattern so quickly or even even really knows?
My mental model is that we are now with AI in the ability to work from above to above.
From Junior to Senior, from a good defined task to a better defined task.
So is that when coding happens and so will it also happen to the data analysis.
That means, to say, what can we not do, what we do not do is one shot.
One shot is the wrong word, because it speaks with machine.
Also, was ich machen konnte, ich konnte einem super seniorigen Datenanalyst ein relativ hingerotztes Ticket schreiben, wo ich sagen konnte, ja, ich muss mal das besser verstehen, ich habe kein Verständnis vom Datenmodell, aber kannst du mir mal helfen?
Und dann haben die sich das angeschaut, haben mir Gegenfragen gestellt, genauso ein bisschen, wie du es beschreibst und haben mir wahnsinnig fundierte Antworten zurückgegeben mit einem wahnsinnig tiefen Verständnis für unser Business, unser Datenmodell und so weiter und so weiter.
Ich sehe das nicht als disruptier...
innerhalb der nächsten Monate im Sinne von so, diese Menschen werden nicht mehr gebraucht.
Ganz im Gegenteil, ich glaube, der Wert dieser Menschen steigt, weil wir eben in Daten ertrinken.
Was wir gerade bauen, ist ein System, was die erste Schicht, was die meisten Anfragen sind, beantworten kann.
Also, was aktuell banal ist, ist, wie viele Nutzer zeigen dieses Verhalten?
Relativ gesehen.
Okay, ziemlich banale Frage.
Die Frage kannst du stellen.
Dann kannst du die nächste Frage stellen.
Was ist unsere Churn Rate?
Okay, auch relativ banal.
Lässt sich leicht beantworten.
Dann wird es ein bisschen interessant.
Dann bauen wir Kontext rein von deinem Unternehmen, sowohl als geben dem Agent die ersten Funktionen, könnte man es nennen, oder Tools.
Also die ersten deterministischen Funktionen, die er aufrufen kann, um etwas zu berechnen, was sich immer wieder wiederholt.
Da wäre zum Beispiel eine Beispielsfrage, wie unterscheiden sich denn diese Churn-Raten in den verschiedenen wichtigen Segmenten, die wir haben?
Dann zieht sich der Agent, also die Segmente, die wir definiert haben für dein Unternehmen, die sich auch nicht jede Woche ändern, die sind an Verhaltensmustern zum Beispiel identifizierbar, zieht sich die Segmente und nutzt die Churn-Kalkulier-Funktion für diese jeweiligen Kohorten und gibt dir dann also die Antwort darauf.
So, then it's a bit more complex.
And so I work, so I see it at all, we work practically in the Nudzlichkeit and in the complexity of the questions.
And that's the way our system, because that was the original question, the answer can be found.
more context, what is your model, what for segments you have, what for business models you have and so on.
So that's, that's, that's, we have a contextual layer built.
So that's, it's hard to say, a context layer, a context sheet, that we built.
And, and on the other side, we have specific functions, which are in the data analysis, which are always repeated, formalized.
And this can the agent now combine.
Also diesen Kontext und die Funktionen kann er kombinieren.
Und so kann er Fragen beantworten.
Und was wir in unseren E-Vals und so weiter halt betrachten, ist, wie komplex können diese Fragen werden?
Und um jetzt sozusagen final deine Frage zu beantworten, das, was du in mehreren zwei Stunden Sessions mit deinem Datenanalyst gemacht hast, würde ich ganz ehrlich sagen, können wir nicht ersetzen.
Und was wir aber machen können, ist, wir können einzelne Fragen beantworten, die the data analyst probably currently belated, because he or he always the same questions be answered and the other ones are not the exploratory and complex.
And we work in the complexity of the challenges.
And we see that these questions, which we can answer, are still very often the questions that teams ask.
So what kind of behavior is that?
That Pareto-Princip?
It helps a lot.
80% just to get it done, is cool.
Even if the other 20% are exponentially more expensive.
How much of the, what you just described, is then automatically automatically?
Or how much of the context of the way you speak, is that more individual brought in pro Kunde?
Where comes the question?
Not skalierbare Systeme.
Und stell dir mal fest, für meine eigene Automatisierung bringt es unheimlich viel, mir Gedanken darüber zu machen, wo ich was parametrisieren kann.
Also wo ist Kontext in die Variable reinschicken kann, damit dasselbe relativ komplexe Verfahren dann auch in anderen Anwendungsfällen verwendbar ist.
Und dieses Parametrisierbarmachen sorgt dann auch dafür, dass es halt allgemeingültiger ist.
Also wie weit konntet ihr das denn schon treiben mit der Parametrisierbarkeit, Allgemeingültigkeit und so.
Das ist ja eigentlich hyperkomplex.
Genau, das ist Nerdtalk, aber das dürfen wir hier.
Das dürfen wir, ja gut.
Ja, nee, sehr gerne.
Da habe ich mir natürlich viel Gedanken drüber gemacht.
Also wir fangen mit der Herausforderung an, dass es generalisierbar sein muss auf der statistischen, funktionalen, wie analysiere ich Datenseite.
Wir bauen keinen Custom, Code, um die Daten von Personen von Firma XY zu analysieren.
Okay.
Das heißt, da muss man mental schon mal so halt an ein Abstraktionslevel hochgehen und sagen, gut, aber ist Segmentierung von Daten nicht eigentlich immer, ich nehme die Gesamtpopulation und ich schaue mir das an und dann bleiben halt diese 200.000 Leute übrig.
Das haben wir gebaut und auf der anderen Seite, hochindividualisiert, haben wir das Wissen about the system, the context of the company and about the system.
That's what we do in the first four weeks, when we start a pilot, when we start a customer, we start a customer, we start a customer.
And that's often actually worked out in dialogue.
So it's a bit of a banal thing.
So what's like, the agent must know, which tracking events there are?
And there can also be payment provider events like Stripe Data or so.
Also it must not just Mixpanel or Google Analytics be.
But often have the doof names, which can't be in context.
And it can also be an agent.
So that's it, we have to add context to context, that says, hey, this event is important, because that is actually the conversion event.
That's not conversion and event, but that's actually a important event.
And this interpretation, which would be a different data analysis, we would have to give them.
That's what we're going to do.
Then we're going to have some details like, you have already existing Arten, like you for example, the Abbruchrate or the...
and calculate your different metrics.
If you have that, we can give that the agent, so that he doesn't make it new.
If you ask an AI, that probably all know, and ask, so, rechne me my Abbruchrate, then probably a answer comes.
It's just the question, if it's the answer, which is the answer, the one will.
We give the opportunity to give the context, the already there were.
For example, we calculate the churn rate.
And as last is there are also segments, that means that we have seen, that is just a repeatable thing.
There are some kind of segments, which you see as a customer, as a part of your own model.
That are not events, not any different behavior, but that is just a...
So ein Primitive, das ist sowas, das tritt einfach immer wieder auf.
Das heißt, das sind die Dinge, die wir alle in der personalisierten Ecke haben, die wir einpflegen können für den Kunden.
Man sieht das trotzdem in der UX.
Also das heißt, die Idee ist, dass das Self-Service werden kann.
Leute können selber diesen Kontext kreieren.
Ist aber auch was, was natürlich kein Kunde machen will, wenn er mal anfängt.
Der will natürlich, dass das einfach funktioniert.
Das heißt, aktuell ist das, was wir praktisch manuell für die übernehmen.
Das heißt, ihr trennt sozusagen, Verhalten oder Fachwissen eben über Reporting und Auswertung, das ist das, was generalisiert ist, aber firmenspezifischen Kontext und die Semantik dazu muss man reinpushen, was ja auch Sinn macht, weil wir wissen es, ohne den guten Kontext kann die AI im Hintergrund nicht gute Ergebnisse liefern.
Das heißt, da sorgt er dafür, dass das eine allgemeingültig ist und das andere so spezifisch ist, wie es eben gebraucht wird.
Richtig, ähnlich so wie man das mit einem externen Analyst machen würde.
Man würde sagen, man erwartet, dass der Analyst schon weiß, was Statistik, Signifikanz und Bonferroni-Korrektur ist.
Aber wir verstehen natürlich, dass der Analyst noch nicht weiß, was unser Segment ist.
Das ist so schön, dass du das sagst.
Die schlimmsten Horror-Erlebnisse in meinem Produktleben hatte ich, wenn Externe kommen und alles wissen auf beiden Seiten.
Also nicht nur Verfahren kennen, sondern sagen, übrigens E-Commerce geht so.
Oh Jesus, you can't do that.
You can't do that in the first presentation and say, E-Commerce is going so, if you don't even know our own business, maybe we can do that.
But you can't say, so it's not.
And all the people who work in so a company, they know, oh no, another one thinks, he's in the first hour, he's in the first hour, he's in the first hour and so on.
So it's actually nice to you, that that doesn't happen.
That you don't have to roll everything with the Jungs, we tell you how it goes, but you really have to listen to the firmness of the way and make a part of the system.
That's actually funny how analog that is, also to the human experience.
We see right now that people, who our system now don't necessarily use, but then we ask why.
Because it's okay, we want to learn.
And then it's like, I'm doing this right now in Cloud Code.
And I'm the biggest fan of it.
I'm using it every day.
But...
That means that there is no context management available.
This business-over-making of this is what it means to us.
We have our calculations for this.
Please use it always.
This engineering, this Übertragung, which you should have to give to extern, the most not done or just like that.
And then you wonder, Because of my personality is an AI, and the most LLMs are more like the Alptraum, which you described.
They make then just the Annahmen.
They are totally self-conscious.
They say, I have to calculate the Abbruchquote, it's not a matter of mind, if you have said it, how I should do it and that I have no plan for your industry and your business.
But I can give you super Empfehlungen.
That means, from the Art, from the Persönlichkeit, are the LLMs, the we are now having, not sonderly careful in their...
Fähigkeit, Lücken zu füllen und einfach mal mit Selbstbewusstsein reinzugehen und zu sagen, ich weiß es besser als du.
Und das versuchen wir zu vermeiden.
Genau, vielleicht an der Stelle nochmal so ein kurzer unerlaubter Education-Anteil hier drin.
Die Welt ist ja gerade unheimlich polarisiert zwischen AI ist das Tollste und bleibt mir weg mit dem Teufelzeug und das ist eh doof und Statistik und Worte vorhersagen und mehr ist es nicht und so.
Und alles hat ja was Richtiges.
Aber sozusagen an den Kommentaren, die man so gerade auf unserem Business Instagram LinkedIn liest, kann ich immer erkennen, was haben die Leute für eine Erfahrung gemacht.
Also sind die vorwiegend mit 20 Prompts in ChatGPT unterwegs pro Tag oder bauen die schon in Cloud oder bauen sie in Cloud Code und sorgen sich darum, dass der Kontext reingezogen wird, wie sie ihn brauchen.
Und die Ergebnisse, die man kriegt, sind halt, sie könnten nicht wilder unterschiedlich sein.
Das eine ist sozusagen wirklich so als Privatmann ChatGPT, oh Mann ey.
I'm going to leave you from the place and these Ratschläge brauche ich alle nicht.
And then when you start with Claude Code to build, then people are unleashed and very enthusiast.
And they build a company.
And then when Context Engineering comes, then it's a bit, then it's a bit of a critique from the results are bad and random.
Because the Verrückt is just...
Oder was ich dazu erklären wollen würde, ist, dass LLMs halt erstmal gar nicht so schlau sind.
Und wie du sagst, leere Lücken halt mit Annahmen oder Halluzinationen füllen und so weiter und so fort.
Deswegen ist das lange beschreiben, was man will, unendlich lange beschreiben, was man will, so ein gutes Mittel.
Und ich kann ein praktisches Beispiel geben.
Ich habe letzte Woche musste ich so ein Bootcamp-Programm reviewen.
Das erste, was ich mache, ist, ich schmeiße das Ding in die Maschine ohne viel anweisen und sage, Sag mal, was daran gut und was schlecht ist.
Da kommt ein hyper, hyper banales generisches Zeug raus.
Großer Gott.
Das hätte ich wirklich auch noch in fünf Minuten rausgefunden beim Überfliegen und so.
Und dann fiel mir auf, dass drei, vier tiefe fachliche Fehler drin sind.
Da habe ich gesagt, guck doch mal in diese fachlichen Dinge rein.
Was sagst du?
Und dann sagt dasselbe Ding, das ich vorher reingeschrieben habe, sagt plötzlich, ja, da sind ja diese folgenden Horrorfehler drin und so.
Und dann sage ich, plapperst du mir jetzt nur nach?
The LLM can answer, no, no, I have now learned in my model, that I have already taken care of my model, now I can't even look at it.
Now, if I look at it, I say, that here is fundamentally wrong in architecture and so.
And that, I think, is something that we can do the people's way of doing and say, kümmert euch drum.
Deswegen heißt es Context Engineering bei Entwicklern.
Kümmert euch drum, dass ihr...
If you have a good result, you will get a good result.
You will wonder how good the result is, if the context is good.
And that is the Sprung we learned in the last 1,5, 2 years.
That is so important.
And at the end it is what enables us to build the product.
Yes, and it is also also wild.
I think that is the people who already have this value.
And the people who say, that was a little generical or that I can't really start with.
And so dynamical the world is, I see our job in this, this at least at some point, to do.
Many, not all, will actually the perfect Context Engineering Skillset build or will always live in.
And the idea would be that...
What we are doing is saying, we can practically this context, the relevant context, the relevant agent with the relevant data, for you to mix, so that you can continue to ask your questions.
You must always say, I want to say, I want to answer these questions from these reasons and I want to see these interpretation or these decisions.
But that's practically...
The actual system is of course also very rudimentary.
We give practically every person access to...
It's also as if you, I don't know, HTTP-Pakete yourself would send or so.
Also in the sense of the browser has very much been abstrahered for you as a user.
And the question is, are tools in the future also a part of this?
So that the Durchschnitts-HR-Mitarbeiter, what else, not context engineering so much learn?
And there is, I believe, the Spannung there.
The, who can already do this, and the, who can already do it, don't do it.
And they can either learn it or tools can hopefully help, that they don't really need it.
Yeah, it's a bit like the Internet.
You said that already with the browser.
The browser came, HTML was stateless.
Did you really get to E-Commerce?
No, but some people saw that could go.
Then came the dynamic language language, which was the browser, the integration, and then it was gone.
The point is it was gone.
The dot com bubble, the dot com bubble, the dot com bubble.
where not everything works.
And I think this is a way to do this.
We are going to see how these systems are going to fit.
What chat GPT exposed was just that model that with all his own and own weaknesses is, then a few product decisions.
It always answers.
It's a bit so, so it's a bit, so it's a bit of the next question.
It's always a trigger to the next question, so it's used to be used and viral.
That's the first discovery when Cloud Code works, ...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
You search a special area, you can also find a more engering harness, which you can make more special information, so you can give more special information out of the box, but they can always bring more context, if I understand that.
Absolutely.
So, you can ask yourself a different question.
So, we want to use our data, which we all collect, we want to use.
We want business decisions, we want product decisions, we want to know what we need to know.
We want to know what people should be calling.
It's important in different departments.
In marketing is it also, theoretically.
So, and what we see is that, like, only product has access to Mixpedal.
That's why Key Account Management knows not, which is not what we need.
I hear every day in Customer Success.
We have no idea, which features the user...
That's a lot of granular for us.
But of course, if you ask the people, it's relevant for me to know, do they do the key features?
Have they their teammates invited?
We have made them in the last quarter.
Is that helpful for them?
Nutzen they that actually?
We can see them in a call.
We could see them, if they do it.
Honestly, it's already too far, so that it's marketing, sales, customer or whoever.
Often it's the product not really.
That is actually the real point.
You have totally right.
The real point is not just product.
And Marketing Sales, here are some people who have no access to it.
They can't know it.
And this is of course a totally overused word, but it is not wrong.
To democratize it is not so easy as, we give you all an account.
Because, oh now, it's just like this to be.
has not everyone the experience, a very complex tool like Mixpanel, Postalk, Amplitude or Google Analytics so to use, to get a better answer.
And so there are people who specialize in there.
The point is just that it doesn't have enough in the company that everyone just answers.
And what we now build a context and a harness, as you said, what the AI kind of thing is.
but create a flexibility for answers, so that you can actually start with something and maybe even even have to go on.
In the sense of, if I ask the chat GPT-Frage here is and people are doing this right, and they are in the position of the product teams, the whole product team, they take their CSV into the chat GPT and say then, what do I do next?
The question is probably a little bit differently.
But in principle, what should I do as a second?
And it's not fair, AI to use it, to just use it, to just make an exploratory analysis.
But of course, without any context engineering, it's a very wild thing.
And the lack of the answer is not, I would say, I would say, I would say, there should be a business pivot.
That is a bit of the challenge, this balance to find between the Superpower, which we give the LLMs, the freedom, just to ask each question, just to throw away every problem.
And then, because it's a business and a product and a product and a decision, a flexibility and a reproducibility to give.
Simple question from, if the board has the same question in the system as the product manager, then the same answer is coming.
That's not the case, if you actually your data in JGPT.
There comes another answer to it.
Because many of them are on the way and many of them are not there.
Maybe that's for those who have not seen it.
The Prompts make not a big difference.
It's more like how I'm going to set the agent, how the tasks in small pieces break, which specific context I give each other agent and how they orchestrate.
And the better I break and orchestrate, the more and more reproducible will be the results.
And that makes this difference between, well, so cool is it not, and a surprise how good it is.
And what I do now is, if I want to do something, I set myself to a decision and show him the difference in the way.
I have a bunch of prompts in chat GPT or anywhere I set and then a structure work, a cloud code or whatever I always show.
And the effect is almost immense, what people would like to do with it and that they would like to want.
And now you have Pandora's box then always opened.
And from me just the effect, I will not have it.
But the effect will I have, the you here just show.
And that is cool, I believe.
And that is so a chance, the in our time liegt.
What I would like to know about two things.
You can search the following.
I remember, if you remember the other one.
The one is, you do this with this, you want to do your own product perspective for two years.
There were quite many changes in the watershed moment in AI.
So came in.
So the question is, what What's your AI now as a company, what did you do before?
And what did you inspire?
And that's the other thing, because you didn't wake up and you said, I build an Analytics Tool as AI-Harness.
That's not the case.
It's nice, but it's not the journey.
You're looking for it.
You're looking for it.
You're looking for it.
You're looking for it.
I'll start with the story.
The story is obviously one of my own pain.
So often.
I don't know.
Sometimes I get tips.
You should actually build a company for frustrated Disabilities.
And then I'm like, I don't know.
And I don't know what they have for problems.
That's it.
Maybe it's theoretically a super business.
But I don't know.
I don't know.
What I know is.
I know exactly what it would be, to look at me before the data from what, before I have to go into the product manager, as product team.
That's been a long time.
And I have been able to answer that.
It cost me 10 hours, which I didn't have, because I'm in 18 meetings.
Or it's not...
It's not right.
that the most time data doesn't play a role in the decision.
Data from six months, where someone makes a report, now as a decision-based basis.
They have changed.
And granularly questions are extremely expensive, because I have a new ticket to someone with a new email to send someone and then I get it three weeks later.
But that team will move fast and break things.
And so I'm going to be so busy.
So that's my Fähigkeit, data, which we already have.
with the time to take into the day-to-day process, is the frustration that I have experienced as Product Manager.
And I know that I know.
And I think I just saw it now, but the one is now, if you have a really super Setup and everything you make and so on, can you do it now?
Das heißt, du kannst dir diese Fragen jetzt schon beantworten, du kannst dieses Problem jetzt schon lösen, diese Frustration abbauen, die ich gespürt habe.
Das heißt, das ist die Entstehungsgeschichte zu sagen, hey, das muss Standard sein für jedes Produktteam in der Zukunft und vielleicht auch für jedes Marketing- und Customer-Success-Team.
Ich kann sagen, hey, ich manage diese 150 Accounts, diese 150 Kunden.
Ich möchte jetzt mal wissen, We have this new, new thing released.
Who of them uses this?
That will I know, because then I can go with them in the next conversation and say, cool, you use it actually or you use it not.
That must still be available.
And for I don't ask the product, could you maybe a Amplitude dashboard build, what you're going to send me, then I don't have to go with them.
Or then I see there...
Also, that is what I feel like, what the pain is and that is the reason why we're there, where we're at.
The second question was, how I use it as a business as a business.
And what is possible, what was perhaps not possible was?
And you use it.
There must be a reason for that.
Also I tell you totally...
Man should the second chance give.
If you tried to make a trip to six months, then you should definitely...
So was wie ein strukturiertes Cloud-Code-Setup, wie du beschreibst, unbedingt mal versuchen oder sich zeigen lassen, um den Unterschied zu verstehen.
Für mich war das auch so.
Also ich musste der Sache mehrere Chancen geben, um zu dem Punkt zu kommen, wo ich jetzt bin.
Wo ich jetzt bin ist, das hat sprichwörtlich jeden Teil meiner Arbeit revolutioniert, was nicht heißt, dass ich keine Arbeit mehr habe, dass ich alles schon gelöst habe, dass ich nicht mehr denken muss.
But what it means is that we are two Gründers, that we are to zweit in many things the bottleneck of our decision, our strategy and not the use of manual tasks.
If I now need a list of 5.000 companies that...
really many data have, but maybe not enough data have.
And I will see, in this company I know someone?
As a company, how would I do that now?
Do I have to do that?
Do I have to do that manually, because I myself have the time for that?
And so on.
So, for example, there is a Python script that says, what I write with AI and not a thing.
So, where I can say, okay, here is my whole...
My whole LinkedIn Connections, my x-tausend people.
And then we can always ask with LM, at which company you work?
Then you can see, what is that?
What is that?
What is that?
That means that systematical thinking allows us to delegate, what is before at the time of the day to delegate, but in reality, and I believe I don't think that it's Jobs back to take away, is it of course that we can't delegate before, because we don't have it for.
That is, at the moment we can do work, die wir davor nicht tun konnten, weil es einfach wahnsinnig ineffizient gewesen wäre, manuell durch alle Leute durchzugehen.
Und das Gleiche ist bei fast jeder Tätigkeit, die unternehmerisch notwendig ist.
Also wir haben zum Beispiel...
If someone is for us for the product, then we have the same question.
Who are they?
What kind of problems can we do?
We don't want to have a problem.
We don't want to have a customer, who are not just a customer.
We don't want to have a customer.
We can't make a customer.
We can make a customer, which is just a cloud code skill.
So, hey, check mal, who has been new?
Okay, from the people who have been new, check mal, what for companies they work.
That's a typical question.
And that can we just automatically automatize.
Very cool.
And on the technical level, had it an influence on your product?
Yeah, also on the technical level, we have both the development process as well as the design process as well as the product itself.
Also, we had before a product built, which was basically software.
which we want to give the 2.0, 3.0, what else do we want, so not AI native.
And there it's obviously only in the development and design process, in the sense of the ability, faster banale Dinge zu machen und sich auf die strategischen und architektonischen Sachen zu konzentrieren.
Ich habe Informatik studiert, ich habe als Software Engineer gearbeitet.
Bin ich der schnellste Entwickler aller Zeiten?
Hell no.
Bin ich der korrekteste, detailorientierteste Entwickler aller Zeiten?
Absolutely not.
Das heißt, was es mir ermöglicht, ist zu sagen, okay, I make the architecture, I analyze the trade-offs, I make the security audit.
Those are the important things that I would make as a senior and a junior would give.
Hey, make a front-end for this input form.
That is already solved.
We don't need to find out how a drop-down menu is.
That is the thing that we can delegate in the development.
In design is the same.
We don't need to We can differentiate ourselves in the stage where we are now as a company, through innovative UX design.
What we want is that problem first to solve.
That's the best in B2B.
You can solve a problem and you can solve it.
You have to be paid.
You have to be a person.
Das heißt, das sind alles Sachen, die wir abgeben können zu einem großen, großen Teil und dann also eher die Manager, die Orchestrator, die werden von diesen verschiedenen Agents, die das oder AIs oder was auch immer, die uns da halt helfen.
Und dann im Produkt selber natürlich die Intelligenz ins Produkt nativ mit einzubauen.
Also statt zu sagen, hey, wir machen jetzt ein Python Notebook, was eine Datenanalyse machen kann.
was Leute sich runterladen können, können wir sagen, naja, wir können ja ein AI entscheiden lassen, welche Datenanalyse passieren soll, sodass das Interface Richtung Endnutzerin auf einmal so einfach wird wie ein Chat.
Wir haben eine Slack-Integration und dann haben wir halt jetzt, keine Ahnung, ich sehe das ab und zu, wenn das reinkommt, haben wir letzte Woche halt eine Kundin, die hat halt gefragt, hey, wir haben im Onboarding was verändert, wie sehen denn die Zahlen aus?
Erstmal relativ primitive Frage rein von der Satzstruktur her.
But there must a lot of happen.
And what now with AI is that we're going to build a product, which says, okay, you want Slack to be, you want our Web-Interface to use, and you want to not even a data toggle X-Achse, Y-Achse to make, but you want to just know, hey, has something in onboarding something that has changed?
And with the intelligence, the we have, which, as you said, is not perfect, we can just try to say, Naja, also welche Tracking Events hängen denn mit Onboarding zusammen?
Wie sahen denn die letzte Woche aus?
Wie sehen die diese Woche aus?
Und der Agent nimmt dann ein paar Denkschritte und antwortet dann, hier sind die Daten von letzter Woche, hier sind die Daten von dieser Woche.
Ich sehe hier einen Unterschied von 4,9 Prozent in dem.
Und dann kannst du als Nutzerin ja immer noch entscheiden, Nachfragen zu stellen, die Rohdaten anzuschauen oder zu sagen, hey, das ist Käse, ich wollte es mir anders anschauen.
But the AI allows you to answer the first time, the user question to answer.
Instead of saying, here is a interface, where you can answer it, if you want to answer it.
And that's of course for product-people, I think, also a dream.
If the product surface is, or not, it's not very important, but if the future is, I can ask a question, the question, the question I'm actually doing in my work, and I get a question on it, that's of course, how I want to do it.
I don't want to click on Bar Chart versus Stacked Line Chart.
That's not going to do it.
I know, you must go on.
A young, a young, a young entrepreneur.
The time is running away.
But first of all, I thank you very much for being here, that you took the time.
And between the lines, I think, came very much out of it, how you make it out of it.
And what that human image is behind it, I think that's really nice.
Just now, I see you very much out of it.
Thank you for that, because that's also the contrast to technology and full of the pieces and so.
There is a idea of the future behind you, which you search for.
And that's very cool.
Um it at the end to the point, you don't come out without telling me what you do, if you don't sit down the computer and don't drive your business, so that people know who Niko actually is.
Yeah, of course.
Very gerne.
What do I do?
I'm relatively in the mountains.
I think that this Gleichgewicht between the very statics and the other ones I'm just noticing, I'm just noticing.
I'm just noticing that I'm just feeling like this.
It's just like a bit more wild, but it's more like this.
And I'm noticing that because I don't have so much manual work, also not so many repetitive tasks, merke ich, dass die Klarheit meiner Gedanken die fundamentale Bottlenecks sind zwischen dem, was ich leisten kann in meiner Arbeit und nicht.
Davor konnte ich sowas sagen wie, ich habe jetzt nicht so die Geduld, das jetzt so durchzuziehen und man muss den Kopf einfach auf mal abschalten und einfach mal machen.
Und dieses Kopf mal abschalten und einfach mal machen ist fast kein Anteil mehr meines Arbeitsalltags.
Und das ist cool erstmal, aber das ist wahnsinnig anstrengend und es erfordert eine wahnsinnig hohe Konzentration.
Das heißt, mein My time in the work has been moved to me as a way to move on, that I make long walks, ideally in the mountains, when it's going, also Trailrunning, and I need to use new impulse.
Also, Reisen is something that I notice, which I can see, which brings me also out of bubbles.
Also Bubbles sind nicht gut oder schlecht.
Ich bin in der AI-Bubble drin.
In welcher?
Dann bin ich vielleicht eher so in der Programmierer vielleicht noch so, wie Leute Business-Owner-mäßig Cloud-Code nutzen, Bubble drin.
Es gibt ja auch noch ganz andere, die ich zum Teil kenne und zum Teil nicht.
Und dann gibt es aber auch zum Beispiel eine Area von Fashion-Fotografen, die sich gerade darauf konzentrieren, wie man AI-Fashion-Fotografie macht.
And what they have for a conflict between, I have 20 years long Fashion Models photographed and now I can do it myself.
And if I'm honest, I can even make the Narbe here.
And so that they really look like and really raw.
That's actually the creative...
Also that's a different bubble.
And if you're traveling or you're with other people, then you find people.
So one I've seen in Bali.
Nick, who works with me.
Nick and Nico.
Yeah, exactly.
And so, I think, that's the two things, I would say, that I'm just going to do my life outside of the work.
This is the different...
And of course, the normal people to see and to meet.
Also, if you're somewhere else and you see people...
I'm just in Montenegro for example.
And then it's so...
That's just cash society.
That's just bar.
And I have no problem.
I'm just so, wait a minute, if AI agents can then pay for it, then we can do that.
And then it's just helpful to know, oh, I have the 1,20 for an espresso, not?
Bar?
So.
And also, um auch mal so ...
The Erdung is so cool.
The contrast, which it's trotzdem.
This conforme innerhalb of the Bible, that's ...
total auf eine echte Welt trifft, wo man rumläuft, Cash braucht, Leute ganz andere Probleme haben, ganz andere Löhne kriegen, ganz andere Sorgen haben.
Das ist so, das passiert ja schon auf den langen Läufen, dass du sozusagen so, keine Ahnung, wenn es halt bergauf schwer ist nach einer Stunde und man doch so klein ist der Natur und dem Berg gegenüber, dass man sagt, mein Problem gestern im Schreibtisch war vielleicht doch nicht das größte.
That's also a huge thing.
This is such a way.
You have always the success.
That is almost like Social Media.
You have always the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the dopamine, the And that's also brought to the Boris Cherny's in so many 10 terminals, and they say ADHS-Centwicklung in principle, and from one Ping, where the terminal says, I'm ready, I'm going to ping, context fütter, spring and so.
Everything cool, and I will that not to take anyone.
But what I feel like is, I must be more than I do, I have to make more of this work, I have to make more rhythm, to make more of this work, to make more of this work.
to come and up to go to the morning, the same time to go to the morning, the same time to go to the bed, my sport to make, my pauses to order and so on, because it would be for me not healthy, always in this thing to hang, just because it feels so cool.
I could switch to my T-Max Setup, where I am at three parallel Workstreams at the end.
And that's right, right, to take the time for other things and to know that it's just...
And I don't know, for me is it especially important, too, for things that are important to take, not only in private, but also in private.
Also, no idea, for example, to take such a way to do this, is also a way to do it, because I don't have three things multitask at the same time.
And so, Naval Ravikant hat so diesen Tweet gemacht irgendwie so, der ist so, play long-term games with long-term people.
Und das ist halt so genau das Gegenteil von dem, was du jetzt gerade hauptsächlich siehst.
Also die, ich glaube, das irgendwie zusammenzubringen.
Das Hassling sehr im Vordergrund.
Ja, genau, wo du irgendwie drei Terminals gleichzeitig machst, aber gleichzeitig auch weißt, dass du in zehn Jahren auch noch arbeitest und auch noch Beziehungen und auch noch.
and also solve problems, which maybe not only for six months relevant are.
That is perhaps also important.
Many of them make their money, just to solve things, which are two-month problems.
Also, business-to-day-fliegen, that is also okay, but that is not what I do.
I think at the moment, you come from a certain ambiguity not out.
I think at the moment, it's really important to live and everything to experience.
And yes, everyone can have a long-term idea, but the long-term idea, also the long-term idea, the long-term idea is very hard to say, more difficult than before.
And at the moment, one of the strategical things that you can do is very tactical, namely...
to know how this work works and to learn what you can do.
And that's so weird, that so a tactical and strategic is now actually a strategical thing.
It comes to the Internet, boom.
You have to deal with it.
Now I don't know what's going to do.
Mobile-Kram came, boom.
Engineering.
what we have spoken about, CICD comes, you have to deal with it, you know what it will be, maybe it's a good idea in two years and so, we know in the moment not, but boom, now you have to do it and so.
And that's the moment where you get everything very hard to sort it.
What is now so nitty gritty in the detail and what is that long term value, super overjoyed, here comes this whole FOMO and I'll pass what, the biggest of us, the Carpartys, Chinese, all have FOMO right and think they'll pass everything and so.
and I don't want to go with it.
So it should be better.
Yeah, and the only thing that I have to do is really, just if you have not so much time, I spend 50 hours in the week in the midst of it, if I have no calls.
And that's of course a huge privilege, because I can't take a job, where I don't have a job, where I can't wait to get in the Jira instead of getting in.
What I could just recommend, and it's hard, but it's just signal versus noise, to say, three, four people find, which don't tell me the opinion, but say, so I'll do it.
And to follow them, because that's the shortcuts, to have the feeling, that you're not there, but not to be able to do this.
That's for me also the challenge, that to find.
I make sometimes YouTube videos and so on, where I take what I do, because I think, that's I had to have.
That's the question, if that's for others can be solved, because I'm not consistent enough.
But that would be the challenge, which we all have, this signal versus noise to find.
Because that's why one is just finished.
And then five minutes on the leery Wand to starry and to filter.
Cool Nico, vielen Dank.
Ich lasse dich weiterarbeiten.
Dankeschön.
Ich bin mir sicher, wir sehen uns wieder, bestimmt auch hier und so.
Sehr gerne.
Wir haben bestimmt noch viele Themen zu bringen.
Vielen Dank für deine Zeit und für die Menschlichkeit, Wärme, alles was da durchscheint.
Ich weiß sehr zu schätzen.
Dankeschön.
Bis bald.
Ja.
Discover World Follow Possibilities Through Stories Told.
Presented to you, Markus Andretzak.