Introduction: A Real-World AI Project Request

So, I'm here to present one of our projects we are trying to implement. We are meant to present some real project, but let's say POC at least, we wanted to present

some real use case and what you see here is basically the picture of a shop which is located It's located in Germany, it's called Rang. It's a shop with, of course, sell and install parquet during all of Germany.

And some time ago, we were giving a talk, similar to this one, into an association of craftsmen, like Comfortigianato here in Italy. And we were presenting a POC based on AI, basically, like today. We were speaking about agentic AI presenting, if I'm not wrong, something related to semantic kernel at that time.

And what happens after the talk, it's like maybe we can't sound really familiar today. because after the talk let's see the CEO of rank this company came to us and said we want to use AI in our company to improve the efficiency so he didn't provide any the scrap description of the process of for example the tools already in place the data something like that he just found a solution and the solution was basically to use the AI.

So this is the start of our story, because then we, of course, started to think about it.

The Core Message: AI Doesn’t Fix Unclear Problems

And just to not to be so rude today into this conference, I think it's a perfectly reasonable reaction for every one of us, because we are basically

bombarded by news about the AI. AI can do this, AI can do that. It feels like magic today.

But the reality, at least the reality that we face every day, tell us that there's also a risk. And the risk is this.

1If you don't clearly understand your problems, AI won't fix them. It will simply make them worse.

Starting Point: Understand How Work Happens Day to Day

So instead of starting with tools, we decided to start with implementing posing some simple question how does does the work really happen day by day in your company so we have to start before to think about the process how the work is done this is a real company

so they have workers that go around Germany you can see later and so how How do they work? What do you expect from the AI?

So instead of thinking about tools, we start observing them and how the work was actually done.

And surprise, this is what we found.

What We Found: A Paper-Based Workflow Drives the Business

Rang has many employees, which are basically workers which go directly into the construction site, like, for example, a building, an house, or something like that to do their job install a floor and every day they have

after the work has done has been done fill in a paper document like this one you see this is not optional this is mandatory for every employee and I already told to you it is a German company so they are supposed to be as as accurate as possible. They like to be accurate.

So you could imagine that, for example, this tiny piece of paper could also become even larger in some cases.

This document will contain something like hours worked, arrivals on site, breaks, notes about what happened during the work time, the note about the material used, used, and everything happened on the working place. So everything starts from this document. The process starts from this document in this company.

And only after this paper is filled in and put on the table of one lucky person, which has the duty to fill in and move into a digital form, then they are able to finally create an invoice. voice.

So I don't know if it sounds familiar for some of our business today here in Italy, but I think it's not so uncommon. This is a real company.

And at this point, we didn't see an AI problem. We see really a process problem.

Why Adding AI on Top of Paper Would Make Things Worse

Now, guess what would happen if we just added AI on top of this paper -based process. For example, yes, if we try to create some sort of, I don't know, scan for this paper document, try to access this information just to be able to show our, let's say, capability in using these AI models.

Data was unstructured, information was incomplete, and most of of it existed only on paper. Processes were written only on people's heads. And error -prone operations, because somebody has to basically translate what is written on the paper into a digital system.

We have different people, different habits, different interpretation, different writings. So I don't know your point, but it's simply like chaos for us.

and it's not really a good idea mixing ai with cows i think the image show speak for itself so here i can start with the demo then i have to switch will be not simple but before to start

The Practical Approach: Replace Paper with a Simple Digital System

let me say that what we have basically i can also describe what we basically did it's really intentionally simple because I don't know how many of you have tried to face the paper once in their life I did several time and the paper is one of the medium that it's difficult to end so when once a company work with paper and

it's fine basically because this is a company which is selling multiple floor or so it's a, let's say, well -established business. So you have to go there and propose to change. And maybe the CEO wants to change because he wants the AI.

But what about the workers? What they think about it? They want the AI, too, or not? And the problem is that if they don't accept the AI

or a software system there, it's a problem for the company and a problem also for us. So we had to design an application that is really as simple as possible to help them to adopt quickly.

We designed two applications. One application is just a simple web application that they can use from their mobile phone. And the other one is of course designed for being used by the workers that go around.

and on the other hand there is another application which is the under administration app which is designed to be used for the unlucky person before to just be then it's a look into a system directly which how is the timing for their for their workers so just it I will show you so totally forget I have

i have to stop the share i have to do this i have to move this one here then i have to switch here and then i have to share the screen and then i am fine so another screen clean it and then i can basically show the application really i don't i i don't know then i would like also to know

from you if you like the application or not because I find it of course simple but maybe not fancy just it this is the application for the workers and I'm really sorry because the application is in German only but it's also something

that I would like also to remark because this is not really a start -up we are a well -established consulting company we work for many customers and much of them them, don't allow us to show their project in public. Run, it's one of them, and allowed us.

So it's a trade -off. It's in German, but we are allowed to show. So I'm happy that it's possible for us. So I will be the translator for you.

App #1 (Workers): Time Tracking, Project Logs, and Notes

It's just some few German words. These are just a small application the worker should basically use from their phone.

If you noted, when I opened it, basically there are some time counter, which were already running, because I started to work 14 hours ago. And this is the label, a bite size,

which means in German, working time, basically. It is meant in their system that will be the time where where a construction worker will simply jump off his home and put his seat in his car and try to reach the first

construction site. So just it. It start to work. And then there is the project time.

I said before that it's German. So they need to split by time with project time. Project time is meant to be set in place when they arrive on the construction site.

So they have to start the time there. And on the other hand, they have also a pause time. I think they will be really happy for this one, because we count even also the seconds,

if they take the coffee, for example. I can give a brief look. look. So basically, now I am starting to work on a project.

This application, I don't know if it's this version because it's a demo, but the version they are using, it's also GPS located. So we will see later that into the administration app, you have the construction

sites and you can basically, the application match the position I have with the construction construction site that I'm probably near to, so that the construction worker does not have to basically specify where it is, where he is. And then, as I said, we have a fantastic button for saying, hmm, I am in pause now, a big

coffee. As I said, I would be not probably happy for it, because I tend to stay in pause too much time. Then I answer it.

and then if I end with the project time I can basically close the project and here there's the possibility to just let's say log the time or basically just let's say take a notes here I can just write with my hands. The point is

that there you can also for example click this button here and just speak I don't know that now how it's my English today but it sounds like not so bad so the system is able to understand what I am saying and it's able to write into this text of course given given the problem okay I can always for example

decide to basically let's say do on my hands and refine the message or I am saying like that because we have to come let's say try to persuade or make the the construction worker to use these features no so every feature is like an obstacle for them so we have to convince that it's better for them so it's not

like I put a button that will press. It should work, they should be confident of it, but it's a bit difficult. Then I press the

button and then you see that the project time is just stopped and basically my log time is taken and it's given back into the administration app.

App #2 (Administration): Review Worklogs and Consolidate Data

I can now show briefly the administration app here also it's not really a fancy application it's like a simple web application we can just let's say look into it we have

the list of the workers we have the construction zones and we have also the working times the working times I can select a worker in this case it's myself and this is let's say the kind of log i created and if you see here there is the this

tiny one put here which is my last note which will contain basically the log time together with the message i basically set in so this is our implementation and now we have to switch back back to the slide here, OK?

Where AI Fits (For Now): Small, Optional Features That Help Adoption

So just to not lose myself, most of you have seen here. I don't know what you think about is not AI at all. This is simply a digital system composed by two web apps, web pages, form, buttons, data, services.

There is, anyway, a bit of AI. AI, and it's the small speech to text feature we saw before, which is made optional here.

So why did we implement it like that? It was on purpose, as I said, because at this stage, our goal was to replace the paper with a simple, fast, reliable system.

And only once this system starts to be used every day by everyone, adding AI could make sense.

I can continue because this is one important point you could guess I don't know this is due to particular customer because of its paper -based process you could guess that this data does not it's not the case when you for example work

with our customer which other is has already placed a digital process maybe well -established with I don't know a well -defined strategy for example so they're using the AI will be simpler and then you have to follow me into the

Scaling the Lesson: AI in Complex Manufacturing Environments

white room now this is a picture this is a white room to give you an idea this is what you could find in a real production plant a place full of machines which cooperates with between them and this environment is representative so of course this is just let's say still an ai image of course but it's representative of one of our biggest partner

which sadly I cannot even name in public because it's a big big company and what you see for example it's also not allowed to take pictures so I can also take in court so it's really clear this is not real it's fantasy from the AI but it's representative so I can say that I see something I saw something similar

some weeks ago if you imagine a situation like this one where machine are let's say tightly together they work for for getting fine a final result many of this machine for not to say everything of this machine are created created uniquely from this customer.

So there are no other machines like these ones around. So what would happen also if we put AI here?

Why “Jumping to a Solution” Is Risky When Mistakes Cost Thousands

So 1at Atartiz, we mainly work as integration and consulting partner for large enterprise and manufacturing companies.

And in environments such as this, this. We face complexity, really. And we pay every mistake with thousands of euros.

Because if one of these machines, for some reason, stops because, for example, you messed with one of these thousand lines of configuration, because these machines not only are designed one time,

but they take in a configuration which can be also changed based on the process they have to do. so if one line is fine not fine the process the process could stop the

company lose money it can happen also during the night time so you can maybe also take taken out from your bed put on the work solve the problem for me and it's even also dangerous for the people that are there so when you work in a such of this place

jumping directly to a solution is risky even if it's AI

even if it's AI it's capable of doing basically everything and that's why we follow the same strategy as we

saw with the rank customer we observe and we iterate on the project

on the problem we observe the problem and we do the software that is needed I already said in this one so and we have really to work that way and by doing this I don't

know how many for example month we will have or years because they call the the hype will continue to to treat us that we will not be developers forever okay OK, but as for now and for decades,

we did software which is tailored against this machine. So it's designed to serve these machines. And we are the contrary of what you can call a department

store, which it's fast, cheap, and for many situations is more than enough. But in this kind of situation, reality is complex.

and always you cannot stay with the standard solution you have to implement something different which is unique for their environment so sorry sorry because we we are into the white room right now so I will like to close with this idea idea.

Conclusion: Be Pragmatic—Use AI Only When It Truly Makes Sense

It's totally serendipity. Some days ago, I was taking off the dust from my bookshelf.

A Reminder from The Pragmatic Programmer

You know, every one of us has a bookshelf full of books. And as a developer, I have some books. And one of these books is called The Pragmatic Programmer, which gave me, it's like a Bible for every programmer.

I don't know if you know it. and some of the sentence I have not the time to note that them down all together because multiple sentence they are screaming out loud with how the industry is leading right now but I would like to remember you one of the

sentence that it's really funny because it's really appropriate to describe our teaser today if you only have one number every promise look like a new so

AI Is a Powerful Tool—But Not Every Problem Is a Nail

So today, AI looks like a powerful learner, isn't it? Are you agree? So I expect it's used to something that is a nail, while probably it's not.

So we need to be pragmatic, at least in our environment. And being pragmatic today means understanding that this is just another tool, and we have to put it in software only when it really makes sense thank you thank you antonio