AI for the Glass Industry

This podcast is presented by NXAI, your partner for time series foundation models and physical AI. Hello, everybody and welcome to a new episode of our industrial AI podcast. My name is Rohit Wieber And actually agentics application LLMs or OCR topics are usually Peter's domain but today we are broadening the topic in talking about very specific industry Very specific domain The glass industry. my guest is Georg Katzlinger Zöllradel.

That's typical Austrian name difficult to pronounce. Welcome Georg! Welcome George. Thanks for the invitation, Robert.

I also want to offer it too. you let's stick with George right away? Yeah exactly so. everything else is too much of a tongue.

twit Exactly So once again cut sling as well rather for our English listeners very interesting Austrian name. but we will stay with George. Josh before We start talking about your approach please introduce yourself briefly To The Listeners. All Right Hey My Name Is George and i work on an AI Solution For The Glass Industry.

we've been doing that now for the past, well slightly over a year. Now I'm having a lot of fun in the topic as it's moving crazy fast right now. so what is your background? Well if i go back to university times and actually more into business and mechanical engineering like I did industrial engineering masters at TU Wien then joined company in international sales, they were producing everything if iified the needs.

then I moved into interlogistics doing aftersales there and now. The glass industry it's a very special industry. you're very specific industry. can you please explain a bit to our listeners?

Absolutely. Actually, I even have to narrow it more down Robert. Okay? It's not even the glass industry is the flat glass.

What does that? what is that? mean i guess Glasses known too most of the listeners to this podcast Is what you look through when You Look Out Of The Window. Pretty Much The Flagglass Industries Focused On Things Around Either The Windows Or Glass Doors or I don't know shower cabins like separate those glass walls that you have in your shower.

Pretty much, Those elements where glass is more or less flat? I have to say more or Less because In a sense also the car industry could be counted into the flat glass industry even though glass there's not perfectly flat anymore. Okay So why did You choose The Flat Glass Industry as A Domain for Your Approach? i Have To Admit That Was Not Your Idea.

No, actually I joined in a bit later. The idea where it originally originated was from Christian, Christian Kippuswanger. He was the founder and he was working with company for say twenty seven years something like that, so he was able to gather a lot of experience when it came to flat glass. When I came into the industry on how customers in this industry work...

I joined a tiny bit later and actually before joining Lumeso. well i knew Glass as a consumer but I didn't have the chance to do a deep dive on all the intricacies and specialties about Glass. So That's what started jumping into that. Can you explain a little bit the structure?

Are we talking about big corporates or is it more mid-sized companies, what are your typical customers. It depends on the region. I have to admit like here in Central Europe, the Dach region there's quite a lot of companies that we can count in small and medium enterprise area so companies with somewhere between fifty two hundred employees that do glass processing. There is also a few companies, or quite a lot of companies below that to be honest but they are rather into taking this process glass and mounting it at the customers.

so the german term for it would be glasereide. English equivalent is glaciers. we focus on these middle parts. We focus with our solution in the company's.

take huge glass as it comes from the glassworks and would take this huge glass, cut it down into smaller pieces or laminate together to make special glass out of it. Or build like a glass that goes in windows insulated glass units. This is what we focus on the middle step. Okay so let's talk about your solution because its all about inquiries offers from customers through ERP system right?

Yes. Can you explain? It doesn't sound like rocket science, it's OCR technical drawings. where are the pitfalls?

and we're the obstacles. maybe if he don't leave us Robert If they don't who was so? its rocket science? I wouldn't go that far but say it is not an easy task.

okay Explainers one with target. with it You know, whenever you start the startup. Whenever your focus on a topic...you look at it through the pink glasses and see that problem?

You have solution in mind to say okay this is what we are going for. let's do it! It can't be that difficult. And then step by step you realize the complexity.

What Is That We Are Doing? What Do We Focus On? The process from most of our customers right now..is they get a lot of orders via email.

This Email Can Come In all kinds of formats, with all kind of attachments. With All Kind Of Context That Is In The Email Itself. So When You Open One Of Those Emails And You Scroll Down You'll See I Don't Know. Two Months Ago They Started Talking There Was Quotations Sent Back and Forth.

Then Finally The Customer Sends In The Thing And Says This is What i Want To Order! And it's Actually Not That Obvious What they Now Want to Order. that's Why for a Lot Of Our Customers this Is Still A Really Really Manual Process. so there people, quite a lot of people sitting at our customers' companies that would take those orders and come in via email.

They will either print them like put the paper right next to their laptop then go one by line item by line atom to put into the ERP system. because this process taking custom order putting it on the ERPs they need to do quite alot data enrichment. Because typically the customer who orders a glass like imagine that you would order one of those glasses You were to write towards did, you know? To describe.

The product that you want and need. but it doesn't mean That this wording that you use is the wording. The producer actually needs to make sure that his system, his ERP system production and all of these subsequent steps work with it Is so specific what they are ordering? So its not a catalog.

They're going through a catalogue saying I want order number. we need one two three four five. Are there really specific products? Yes It's a make-to-order industry.

Okay So they get orders And customers typically configure products as they need. They say, this time I need a glass that insulates better and keeps more solar energy out. so it has different G value or light transmission values all those things. So its product can be really configured for the specifications of user.

Then receiving party our customers still needs to add information on how we want process it. You know you have to assign internal material like these processing parameters All those things, they need to apply historical knowledge. They do know especially the experienced people that work at our customers who enter orders have all this historical context-knowledge knowing if these customer uses this and that word what does it actually mean is? Okay so there also doing a production planning stuff.

right Production planning, yes in a sense but this would already be the second step. The first step is basically to translate the customer language into what it means their ERP system In the sense of identifying the correct IDs like putting the build up right and then next step after that they will have to look at when can we actually deliver it? Is this material on stock Can do now or need buy from another supplier because we don't have facilities for product ourselves? A lot complexity bundled through the operator that puts order into a system.

And this is where we come in, and support those clerks or people entering orders because you can cover one major step for you which is taking input and preparing it as good as possible so your work becomes easier. How do you do that? Well... We kicked off like many would.

We just said, I cannot be that difficult. Let's write the smart prompt and let's get the input converted to the output in a structured way And all good. actually this is also what many of our customers tell us and not be The difficult right? i mean i take one Of the pdfs.

did i get as an order lcr? yeah well Not even ocr. they say. you know What.

i Just go To claude. i got to chat chip it here i go to wherever And It gives me impressive results. We started in a similar way and then soon we realized okay if you want to do it on an industrial scale, You need to do differently. Because actually what customers expect is deterministic outcomes right?

So they sealed-in the order And they wanted have same result. always again. This where I believe that we can contribute or make difference. because If you do this now Or any listener could ask a question to JETGPT, Gemini whatever Claude Open three screens in parallel and see what the answer is.

And they'll say that, The Answer Is Different Three Times In A Row. So What We Need To Do And What we need to put it into is actually get this deterministic outcome. How do you do It? The Difference Actually Is The Learning Part Like how are able Add On To The Base Intelligence Of LLMs How Are able Create The Workflow That Enables The User to reduce workload.

But how do you do that? What is technical in behind? I mean, I cannot give you the blueprint. Yeah sure but maybe if us some hints yeah i can get your honor.

on a conceptual level it isn't multi agentic step That we apply in parts its conventional code because whatever works really good in conventional cold We try to do in conventional called. why not? when he then comes to the point where we see like easily do that in code anymore, it's actually smarter to use LLMs with calls. We would do it in calls.

what we soon realized is our prompts and how we talked to LLMS became bigger and bigger and the bigger the prompt got like The worse the results became. so we then started splitting It. we started splitting into multiple steps And now also technology evolved over the past especially months. That now become even a bit different in the sense that we start applying a genetic systems to it where We use skills.

In that sense You actually give a bit more freedom at the right point To the system so you can pull the tools and the functionality That it needs to get to a better result, that's like one path The forward path. equally important from our point of view And this is now what we can make. the difference Is deriving learnings? what we see that is correct from a user point of view.

I believe this is an advantage. actually in our setup, the users are very eager and they're very reliable in validating that the output is correct because actually when we present to the user they edit it then send over into the ERP system. so all big advantages get really good and accurately validated result. So there's still human in loop right?

Yes, yes. To be honest right now in our setup I don't see how that would work without the human. i'm actually convinced That with most of the AI use cases The AI Is to amplify the human and empower the human to create more productivity within the same time? Don't think that our solution necessarily Upforces human beings out-of-work.

it's rather amplifying them and enabling them to work faster. I just wanted to jump back to what i said before. So having this validated results that gives us a really, really good chance To derive learnings from that. Learning in...

I have to stay at the high level here. Learning is different dimensions so it's not one learning that will apply for all and every customer anyway but its the differentiation between customers regions where we can derive learnings that we can then in the next step for the next round of order, apply this knowledge. That right now sits with the operator. were they say hey I know when a customer writes this They always mean that product.

This is to learn it. so anyway We're kind building a dictionary. or you are building an inquiry model for the flat glass industry. Yeah You know, the thing is you set the flagless industry now and we get that question also from customers.

Hey with what I do here am i actually training? The rest of the industry as well. So am my training the competitor What we see? That's not the case because every company speaks their own language right.

Every company has their own Behavioral patterns uses their own words uses their on information to actually get a result. so what We do hear Is we basically our solution gives the framework to enable users To put this knowledge into data and to use that data. To enable users, to enable operators to be more productive. so actually the knowledge we create for a certain customer of ours is not transferable to another customer because it's dictionary which works in their customers but not anyone else.

How accurate your solution? Can you share some KPIs when it comes to your custom? Accurate are results. First of all I have add though depends And it depends on the quality.

Yeah, I mean you sound a bit like CEO to me. Like a customer CEO? We also want to always have precise KPIs and precise dates in everything. that's measurable...

Maybe maybe a flat glass industry CEO is listening now to this episode and he wants to get in contact with you George! He wants know some KPIs. Let me give you one KPI. Ah, let Me Give You two KPIs and it's I just take One exemplary Field That We are Extracting?

Lets Take the width you want to go for a guess first. or Shall i tell you right away what do you think is The Level of Accuracy For Width When we extract with the width Of an element like A glass can be as good in height? how Many Times Do you Think We get With Correct? Okay, it's actually nineteen nine ninety-nine.

Yes and this already also includes when its handwritten input. so But here is some context And this is why I want to say really depends if the input is machine written. So if it comes in a form that can be easily extracted there Is no OCR needed or not too many conversion processes? The input file is not too large.

its quality is really good, like really good. What decreases this probability? It's actually when there are a few factors at play that reduce the quality of input. say The document is scanned and maybe the documents tilted.

This where now models start to struggle a bit. When There Is Handwritten Annotations When There is Ambiguity in the Input. So Say That Party Ordering an order Or Ordering A Glass has a PDF attachment to it and then says in the email, but wait element number two you make with of one thousand five hundred. Because then that system needs to figure out.

hey what is the truth. now how do I rank relevance? And this is where we will go down from there hundred percent. okay What's the second KPI?

The Second KPI which one can i give you? let me think oh You have even more so We Can Go Further. Yeah No no no no need To Be Careful Here. Well, I tell you now a listeners.

We also did the same analysis once for only handwritten documents and we saw that actually were getting closer to what you suggested before. so back then when we did the analysis an only looked at Handwritten with. it was somewhere in area of seventy five percent seventy seven percent. And here this is where depends.

It's really about the quality of input, because as we all know handwritten is not always handwritten. You might have a prettier handwriting than I do so maybe this system will struggle less With your handwriting than it would with mine. Yeah, okay? Can you explain us a little bit the implementation process at the customer side?

where they did things get stuck At The beginning. what is the process looks like? can You share a bit? how difficult or easy is that to implement the whole stuff?

Absolutely and I guess this work gets really interesting Because we, I mean... We hear from Silicon Valley and all the AI hotspots in the world. All the time how cool AI is! And how much benefit it gives end-to-end.

The interesting part what we are seeing Is the reality of AI In a manufacturing setup. So our customers are typically as i said before Somewhere in an area of one hundred employees. Typically, in this size there is maybe one person responsible for IT or may be it's just half a full time. Maybe the CEO?

Yes! In-person yeah. Responsible for IT. so what we see like huge interest.

We get a lot of demand and they are...we see an absolute willingness to move ahead. What sometimes bit of challenge is internal availability of resources could be because it's such a niche, the ability to establish this end-to-end process. So to start in the email program of the customer connect to our solution and then connect back into the ERP solutions.

What is the standard APIs? or am I wrong? Not always APIs. In some cases we also have to go through conventional EDI import connections Because there is nothing more up-to-date present yet.

And this is, you know in a sense it's also limitation and the chance. at the same time I believe what i'm seeing that definitely helps. if the earpiece supplier already has some kind of API implemented we can easily connect to. That accelerates process a lot.

It just not always like that. Yet We see that the industry beheaded in that way. And on a higher level, I actually think it's absolutely necessary to don't see how an ERP solution without API or like a proper API strategy can be successful. and I don't know three to five years from now.

look at the CP what they announced just recently with sales for sales first what they're doing. Yeah. Or Ceylon is stuff yeah? Yep.

But isn't then running into cloud your environment Can you share a little bit? Yeah our solution is running in the Cloud we have Right now for Europe. We provide data locally actually on a server in Germany cloud server and Germany. Is it the topic for your industry?

sovereignty Data sovereignty. Customers from Europe definitely like to have their data in Europe. what we didn't experience yet is that customers insist On having it on prem mm-hmm, okay Actually I but isn't possible to run it on Prem. Everything is possible At the end of today, I guess it's a matter-of-price Especially because we connect to powerful models in the background and i'm not sure if any Of The Customers has the GPU power on prem To run all of that.

what kind of model can you name? The models actually. I do Not want That's okay. What I Can Tell You Is It's A mix.

it's A Mix of Different Models of Models with less parameters and models with larger parameters, because what we also see is that costs they should not be underestimated. You know it's just one call for a business case. It actually I think were doing somewhere in between twenty maybe to twenty five calls right now during One Business Case. And if you will go through the most performant models then it doesn't really become a business cases anymore.

And your customers paying for the tokens? or what is the business model? Now we translated into a language that actually speaks to our customers. In the beginning, We thought about the case should we just charge customers by tokens?

because it would be like back-to-back contract and pay is kind of what we charge out customer with change debt. We are charging customers based on how many line items run through our system cause. in a sense This is a degree for them to measure how much more efficient they can be. So basically, the pay-for reduction of workload in that sense and it's really well received.

I have to admit The industry is rather familiar with conventional software pricing. i don't know your seat or you pay once And then you pay yearly maintenance fee. We didn't receive any pushback. actually customers kind of like because They are going to pay if their using It They don't pay.

So it kind of fits their business model, okay? That's good. You mentioned at the beginning that you support your customer At The first step Extracting the information from the email From the PDF from the Word document for a picture form a drawing. Is there is second-step plan to help the customer too process the data now to the ERP system or do planning?

Or is there anything? you have a secret source, are you having an idea what's coming next. Yes Okay We actually great plans. okay share some ideas.

What I'd like to talk about staying within the industry and focusing on what we're doing right Now right and running really, really smoothly is this first end-to-end process of orders getting orders from point A to Point B. From the email into the ERP system. if This can be done If we can enable our users To become Ten times more efficient with that than they are Right now? This Is already a huge step.

what We added just recently's? also? we enabled them to do quotations in The same way because we got the feedback from our customers saying, cool that works for orders. Why does it not work for quotes?

We said yeah I guess we can do that. so we added quotes and a lot of our customers are now running quotes through Lumizo as well. So in that sense It's becoming quite powerful sales support. And this is where we see ourselves especially with this dull work of taking a lot data, preparing it in certain formats that can go into an ERP system.

We see ourselves as an add-on and tool to interact with the existing ERP systems. This is where I believe we contribute a lot. Some customers ask hey So you're kind of an ER P system And i always have to clearly say no We are not an earpiece system. I don't want to be at the AP system right here, we really want to connect your existing earpiece systems because we know that you know.

even thinking about becoming a new European system and maybe being part of implementing in New York pieces them. i did That In one Of my previous jobs? We got SAP. You're A big fan of SAP.

i can hear that. yeah now i see The tool is immensely Powerful. It's a lot of work. You know, like when you try to do heart surgery on the person while this person is running a marathon.

that what it kind felt. and I prefer if we become this add-on into an existing ERP system make sure We can support with really good data so for The customer process just becomes much more smoothly than so. much easier. are they other domains comparable to the glass industry, where you say oh this could be an interesting domain.

maybe same approach. Same problems? Same obstacles? do You have plans to scale your solutions?

two other domains? we actually get a lot of interest from connected domains. in that sense like quite A few window manufacturers already approached us for me it's important to get one thing Right before I jump into the next adventure. I think conceptually our tool can definitely do it.

But you know step one then step two, okay? What is on your agenda in the coming weeks George? In the coming week's customers Keep asking and more of them keep asking. so are kind of need to be able.

You can send them the podcast now. Yeah Yes maybe this instead of sales meetings that we're having Next weeks for me are really busy. Actually, I'm going to the UK next week To visit a few customers. i'm running in parallel A few demos with customers from pretty much all over Europe as well As The US because we get a lot of interest From there right now as Well.

And also I didn't tell you that We have a customer In Australia already. So could be them Spending their whole world quickly and You know making sure That then actually Also able to onboard those customers While at the same time keep adding functionality. What we see, and this is our big benefit I would say as a startup... We don't have too much legacy in our backpack.

so we can really quickly add new functionality based on the feedback that we get from customers. And this what will keep doing for next weeks or months maybe even years. George i keep my fingers crossed for you and your team. all the best interesting approach.

Thanks a lot, excuse me for playing the CEO asking for KPIs but I think your customer will like the questions. all the best and greetings to Austria! Thank you very much Robert, talk to.