Driving Manufacturing Efficiency With AI: Pirelli’s Daniele Petecchi

When you think about digital twins and generative AI, you probably don't think about the tires on your car. But on today's episode, learn how firms are using AI to develop this key component.

I'm Daniele Petecchi from Pirelli, and you are listening to me, myself and the AI.

Welcome to me, myself and AI, a podcast on artificial intelligence and business. Each episode, we introduce you to someone innovating with AI. I'm Sam Ransbotham, professor of analytics at Boston College. I'm also the AI and business strategy guest editor at MIT Sloan Management Review.

And I'm Shervin Khodabandeh, senior partner with BCG and one of the leaders of our AI business. Together, MITSMR and BCG have been researching and publishing on AI since 2017, interviewing hundreds of practitioners and surveying thousands of companies on what it takes to build and to deploy and scale AI capabilities and really transform the way organizations operate. Hi everyone. Today, Sam and I are very pleased to be talking with Daniele Petecchi, head of data management and data science at Pirelli.

Daniele, welcome to the show, and let's get started. So, for those of us who are not familiar with Pirelli, I happen to have four of your wonderful tires on on my car, actually, eight of them on my two cars. But for those of us who might not be familiar, can you describe the company and then your role at Pirelli? For sure. Pirelli is a manufacturer focused on the tire, as you said, and in particular tire for car, moto and bike.

Our target is to be on the cut edge of the R and D and technical point in the automotive industry. Pirelli is an Italian company, is an old company, is more than 150 years old. So during the time Pirelli changed, and now we are leader in the prestige and premium segment, the high value, but also we are known also for the unique supplier of the Formula One. If you see on Netflix, drive to survive, that is the series that is related Formula One.

So now I know what kind of car Shervin drives, he's he's he's the only one driving the luxury. He's a Ferrari Lamborghini.

you see, I'm a high-value customer. Yeah, Ah, you're a high you just buy it. value. Good, good, good. Maybe you tell us a bit about your role at Pirelli. Yes, I am responsible of what we call data management and the AI, but at the end, I deal with all is tree related to data. From business intelligence reporting, dashboarding to the data engineer, data science, AI, deep learning, general AI. We work with all the data. We work with data from our plant, we work with data of R and D,

and we work for business with the AI, also on commercial supply chain and so on.

Maybe give us some ideas of how data and AI is influencing all of the elements that you talked about, the whole value chain from maybe production and R and D to maybe supply chain. Can you give us some examples?

Oh, yes, yes, the business model is a simple business model, but is for sure powerful. Our target is working with our car maker. So this is our idea. We work with original equipment and then we can support our replacement market. Moreover, we do this with the benefit in terms of time frame of our business. Car maker collasa to develop the tire for a new car three years in advance of the launch of a new car model.

Then the new car model is launched. After four years, we have the first replacement wave. After four years again, we have the second wave of the replacement. At the end, this provide us a visibility, this time frame of eleven years. So when we start with the new business, we predict the next eleven years. Due to the fact that we launch the new car, the car maker launch a new car, and then we work on the replacement.

So this is the. This is powerful for our business model, but is also challenge for the use of data because you have to work over this eleven year to guarantee the supply of the product, the quality of the product. Consider that we are working on the prestige and premium market on the value, Ferrari, Lamborghini, BMW. I don't want to say all the brand, but are are the top brand. Based on this, let's go to consider that every year we produce more than sixty or seventy millions of tire.

We produce this tire over eighteen plants around the world. We provide our tire of almost sixty market around the world. And to manage this complexity, because each year we produce two, 3,000 of different SKU. The tire is pretty different inside for each model because there are the inner rim, there are the inches, there are the winter and summer tire and so on.

All the speed ratings and all the letters

Speed rating. and everything. Yep, You are expert, you are expert. So you have to manage this complexity and to guarantee the quality of our product for this brand that have, let's say, strong requirements in terms of technical requirements, in terms of performance requirements, and to do this, we need data. The data is the only way to manage the complexity. As I said, Pirelli, 150 years ago produced tire, but was a different scenario. The competition today is pretty different.

It's a global competition. The challenge to face with the market is for sure different, and the requirements, as I said, of our customer are really challenge. Strong. In terms of example, I go straight to the example where we apply the AI, and as I say, the data is the building block of our business body. If you example, we can start from R and D, for example, particularly in the past. When you develop a new tire, you start with the technical requirements, you design a new product, then you produce a prototype, test it, and then update the design of the product.

So it was a kind of development based on different loop due to the fact that you have to try different recipe of each tire. I can tell you that the tire is like a pie, okay? There are the ingredients, you prepare your pie, and then you have to cook it.

I've had those kinds of pies that taste like a tire.

Yeah, no way.

How long does that process take? That that iterative process you're describing, how long is that taking?

Uh, the starting from the body, thirty minutes to produce one tire.

How about the design process, though?

Design process require, in particular for the prototyping phase, require several months in the past. And for this, artificial intelligence is able to support R and D because you can virtualize your product development by A I.

A digital twin you're talking about.

Yes, digital twin. You have to consider digital twin for this kind of product is based on A I. When you want to predict, for example, the noise of a tire, there is no a simple equation that say, okay, guys, for this tire, this is the noise. No, no. And today, the noise is really important because, as you said, with electric car, there is no noise in the engine. The only noise of an electric car is related to the tire.

Right before, the engine would cover up those noises and you didn't have to worry about it. But now, with new models of cars, then you're a lot more pressure.

Yes, yes, there are more pressure and the tire produce noise. Moreover, there is the ESG stream, so it's important. The noise is also related to the ESG requirement. At the end, if you want to predict what will be the noise of a tire, I can tell you that it's hard with the theory. We trained our neural network and a neural network model, and we developed this kind of model that was able to predict the noise of the tire.

But we are lucky, we had the data. We had the data of the experiment of the prototype that are testing the past. So by this kind of data, we develop a huge amount of data. We develop this kind of model, deep learning model, that are able to predict the noise of a tire based on the design of the product and other technical parameters without the prototyping phase. So by this model, we are able to support R and D, reducing the time to market.

As I said, several months to develop a new product, we can reduce the time to market and then also the cost of the development of a new tire.

You have the benefit of data and the ability to experiment more than a brand new brand. Yep,

yes, and we based in our leverage on our historical data. We trained our model, and with this model, we are able to predict not only the noise with the AI, we suggest also the proper recipe of our tire for a new product according to the customer requirements.

so you dramatically reduce the prototyping and testing phase before you go to full scale production, and it's sort of like what we talked with Moderna, for example, and coming up with new vaccines, right? Because this whole process of trial and test and trying different variations or coming up with new formulas, this is quite quite fascinating. I can also imagine the manufacturing process once you have defined the recipe and the design that that manufacturing process of nearly 100 million tires a year with plant time and production line and all that must also be quite complex.

Are there examples of AI helping you run that more efficiently, more effectively?

Oh, yes, in the manufacturing, in the production process, the the target, as I said, R and D target is virtualization. In the production process is the efficiency. So we apply our model to provide more efficiency in our production plan at the same time to guarantee the quality. Because, as we said, a car customer are strong in terms of quality and technical requirements. And with the AI, you have to change the approach of your plans, starting from a reactive approach to a preventive approach.

And this is what we put in place with the predictive maintenance, predictive quality. You have to be able to predict the quality of your product before you complete the production. So it's important that if you want to provide the efficiency in your production process, they cannot wait the end of the process to say, ah, unfortunately, this product is not good waste. No. During your production process, you have to be able to, let's say, provide early warning of something that could be goes wrong, and then react before the something happened.

And this is based on artificial intelligence for us. So during the process, collecting the data with IoT data because we collect the data from our machine, our machine, we are able to detect if something is going wrong and could impact the end quality of our product, and then we have to do something during the process.

So, rather than stopping the process, you modify and adjust during the process.

During the process, the challenge is to do something during the process. If you stop the process, it's

Yeah, big deal.

a big deal, because it's for sure inefficient to stop the process and then start again. So, when I think

About, for example, the complexity of what you're doing. You made the analogy to a pie, and if I think about a pie, I might have some flour and some sugar and a filling, like a pecan or a pecan. If you're where I'm from, these fillings, you know, there's like half a dozen different ingredients you could combine. But my guess is the complexity for these people you're working with is much larger. What's the number of ingredients in the pie for a tire?

Yeah, more than one hundred ingredients, and if you consider the tire, you have twenty six parts of the tire that are built and then cooked. It's for sure. When I moved to Pirelli, because I had experience in other company, other industry, I was surprised about the the complexity of the production process. But I said, well, I thought, if you consider a tire, a tire is under a car, a car is more than one tons of weight.

If you consider electric is more than one, almost two tons of weight, and if you consider this, if you consider that a tire has to support a car, and I don't know a Ferrari for 300 kilometers per hour of speed and temperature, because the weather condition could change. So we can start in winter in Italy with a temperature that is, uh, zero degrees to forty or forty five degrees in summer.

That's like 130 fahrenheit for the American audience, right? It's pretty hot.

yes, yes. But if you consider all these conditions, you say wow, tire is not. It's not simple, the life of a tire

And so what that makes me think about is how much the potential for generative AI is. We think about generative AI, we think about, you know, language models that people are used to here. But you're not just predicting noise, you've got a whole bunch of data that will tell your workers what are potential combinations that might work well. That's a very different generative world than our generating sound or generating language. That generative design seems different.

generative AI is something that we apply. We apply a model of generative AI, in particular related to the language. But to be honest, I can say the AI in Pirelli split in two mainstreams. One is the stream that I described before, the stream of the operation, and this is something that, let's say, is based on deep learning, it's based also on generative AI, but generative AI not related to the language. With generative AI, we cover something that was a new opportunity in particular for the division, for the colleague that work with text.

But to be honest, what I described before generative AI is not only text related to the suggestion of a compounding of the ingredients of a new tire is generative AI. So is generative AI, as we might discuss about the idea that our model are able to propose something that in the past was proposed by only by human. So Gen AI, as I said today, by the hype of Chat GPT, is in particular applied for, as I said, the staff function.

But behind our predictive model, our model that virtualize the R and D, also behind this model, there is a generative model.

Right, that seems particularly interesting. You mentioned before you didn't realize this before you came to Pirelli. Well, tell us about your life before Pirelli. What did you do? What's your background? How did you get interested in Pirelli and interested in tires?

My background is I'm an engineer, I'm an engineer in digital signal processing, and after the university, I support my professor in Italy for a few researches related to the data in telecommunication industry. Then I started in a consulting company, then I moved to telecommunication industry. Then I had also the opportunity to see the utility and oil and gas industry and also retail. So I saw. I had the opportunity to see the customer base, so the consumer in telecommunication and in utility.

I had the opportunity to see the behavior in terms of data of the shop in the retail, and then I said, Oh, what is missing manufacturing? So I when had the opportunity, uh, with this challenge in Pirelli, I say yes. It's really a good opportunity because this data strategy is inside in our digital transformation in Pirelli. Here in Pirelli, we started in 2018, not only with the data strategy, data-driven approach, a different data-driven. We started with the new digital transformation to support the business model that I described before.

We started this kind of transformation that are based on data, and over data we have also transactional process based on CRM, based on product life cycle management tool, based on a new tool that support the production process, the IOT project. If you consider the production process, the point is that you can see, you can touch the benefit of what we do. If you, let's say, are able to provide a better quality in the product, you see at the end of the process, what are the benefit of your model of your work with data?

I like the way you tied all those together, too.

Yeah, it's a wonderful narrative. So we have a segment of our show, it's called five questions, where I'm gonna ask you five questions. You don't know what they are, hopefully you don't know what they are. Just tell us the first thing that comes to mind, just rapid answers, quick ten second answers is what we're looking for. So, ready?

oh, three, two, one. Oh, I'm a little bit stressed. Okay,

What do you see as the biggest opportunity with AI right now in the world?

the world, for sure. I think better support the human in day by day work. I think this will provide only benefit for people. They will leave boring activities to be focused on value added activities. This is the main benefit of AI.

What is the biggest misconception that people might have about AI?

That AI will substitute the human in their job.

What was the first career you wanted? What did you wanna do when you grow up?

Work with data in any case, with the new era of

As a as a five the next year old, you wanted

gen of AI.

to work with data?

In any case, the data with the next gen of uh data,

Is there such a thing as too much AI? When is there too much AI?

it's not too much AI, it's a bad marketing of AI. When you do something, there's, oh, this is AI. If you the sum of two numbers, sometimes is AI. So there is an overselling, the point is the overselling of the AI.

Good, good, good perspective. What is the one thing you wish AI could do that it cannot do right now?

I think that AI today work with the industries, with the process. I think that in the health and safety could be better, the support of the AI.

Very good.

We really appreciate you taking the time to talk with us. Something like tires is probably not something that we all think about a lot. They just work, and until we hear them or they start making noise, we don't worry about them. But what you've done is given us an insight into how complex that process is and how humans and the machines can work together to learn how to make better tires for us all. I think that's really interesting.

Thank you for taking the time to talk with us today. We appreciate it.

Yeah, thank you so much.

Thank you, thank you, thank you for this interesting discussion.

You also got us to think about pie.

Are you hungry now, Shervin?

I am very hungry.

For your cook the next I'm turning into pie, please.

pies. I want to drive fast and get pies.

Remember, yeah, yeah, remember the the the tire when you prepare the next pie.

Thanks for listening. On our next episode, we learn how AI and generative AI help people collaborate more effectively with Jackie Roca from Slack.