Introduction

Hi. I know it's working. Okay. Perfect.

I'm Luis Carlos Pietro. I'm the head of AI Center of Excellence at Ferrovial, a Spanish company.

Some people probably say from the Netherlands, it's not true. We are Spanish, right?

The Evolution of AI at Ferrovial

So we are working hard in order to create different, not only AI, but generate AI, different solutions and all over. the different evolution in terms of the technology involving AI to generative AI, right, is It's absolutely clear that the time we need to be absolutely rocket up, the number of users for the different products and solutions is completely changing. It's completely exponential, right?

So it's changing a lot. But we need to balance with this speed, with the technology and so on, how to impact with the different businesses. We consider that if we impact on technology and processes, we will create problems with our users because they need to understand what happens with the different process.

This happens in the company. If we work with only technology and people, we only create chaos. absolutely chaos inside the company. It's absolutely impossible to manage, right?

And obviously, if we impact people and processes, our colleagues, our employees, they will suffer frustration a lot with the process. We need to work the three dimensions in order to impact directly to the business process and so on.

Historical Changes in Technology and Work

Historically, the technology was changing from the very unprecedented behavior with our colleagues from the 50s, where we were working in a synchronized organization. We were working face to face in the same office and so on.

After that, the story changed with internet in the first years of this century. Just with internet, in order to share in a synchronous way, in order to share the different tasks with different projects, we don't need to be absolutely synchronized with others and we can work on different times or including different offices all around the world.

But pandemic changed the rules of the game, right? We can start working in our homes and the rest of our colleagues in the office or other countries or time.

The Advent of Generative AI

But what's the generative AI change really nowadays? the new dimension for the possibilities we can improve our results and be absolutely focused in the very important part of our business and left all the boring processes and so on to be absolutely automated. Genitive AI is not completely new, it's just only an evolution from the very first artificial intelligence models and so on. Remember, the first neural network is from the 50s, right?

It's not new. The only chance is that we can compute it in our watch or whatever other device we have in our, not only companies, but in our wearables or any other device we have. In terms of corporate, this is slightly so important. But the most important is the tiniest number.

0.1 and 0.6% of working hours will be recovered just only using AI. It's incredible amount of money, not hours, incredible amount of money with the different companies. If we can improve the results or increase the speed, we're using it to respond to several projects.

Understanding Generative AI

What's really generative AI is, remember, LLMs is not the knowledge platform. It's just a model that predicts perfectly the next word in a text. Just only that. Predict the preferred word after the first one, right?

Different companies are working just only in creating rough projects, so just only automating the different documents, information, to create a perfect response. But we consider it's not enough. We consider we need to change the rules of the market to create actions to other things that is not only using document information or to create a perfect response.

Why not to create a direct relationship with our customers or employees and what's for or include look for new other possibilities. With different flavors we can create different solutions. We consider not to create a front-end to use that one. We need that, okay, the people is not boomer than me, probably prefer that one.

Ferrovial's Jarvis Project

In order to create the Jarvis project, just only a single point in order to respond to any query you can launch to the system, no matter the information is, no matter the destination of this information will be launched or act, right? This is our internal project inside the company, Our platform is our generation 3.

We are working in generative AI to 20,000 employees since February last year. Right? So this is our generation 3. But our intention, obviously, several assistance that is securely protected. It's not shown on this slide, right? But we have...

the project in order to automate the different processes inside the company. Because our idea is to accelerate our employees. It's in order to give our employees superpowers in order to be absolutely more productive, in order to improve their results with less time to do it.

The Impact of Generative AI on Productivity

Obviously, generative AI can't do anything. Genitify can't finish a building. It's absolutely impossible.

The idea is not to fire people. The idea is if we have 15 engineers and we can respond to 20 projects with the same 15 engineers, we can respond to 40. And if we can't respond for it, probably we will hire a new engineer in order to respond 50. This is the idea.

To be absolutely focused, to be more productive, and be absolutely different in the market with the rest of the competitors in order to respond to other projects, including harder projects that we consider so difficult, but we can assume these challenges because we can accelerate our internal processes.

Ethical Considerations and Compliance

We need to create, obviously, the different limits. We are so, so compliant with AI Act, in fact, because we are creating the platform with complete transparency, explainability, and stability, and so on. The models is not possible to explain because the model learn by themselves, but we can explain how we are using our datasets and the different models. But the most important part of that is the bias.

Remember, all of us, we are in the middle of incredible moment nowadays when a technology is created and we are in the middle of that and we are creating the steps for the rest of the people because we are working on it, right? It's so, so exciting moment, right?

The Importance of Culture in AI Integration

We consider that in order to create this kind of platforms, we need to use five principles, pillars for that. But the most important for that is not all of them, but the culture.

The culture with AI is absolutely necessary to be implemented from the top to the bottom. I mean, from the CEO of the company to the rest of the employees. We are absolutely happy with our president inside Ferrovial because he always say AI in all the different process in the company and is absolutely sent to the rest of the divisions and employees.

Because if you work from the bottom to the top, you will find barriers in the process. And finally, the AI will not be possible to create for the company.

Remember, the tiniest numbers. Why? Because it's money. And this is a complete example of that money.

It's a Microsoft study from seven different months How many working hours will be recovered on the different industry sectors and so on?

It's incredible. 100 minutes in just only one week. Per week. Minutes per week.

It's incredible. Try to think the number of hours with all your employees. Right?

Innovative AI Applications at Ferrovial

How we are creating that?

We are trying to think not only on the top of the iceberg, but the bottom. Not only document analysis and so on, but create way not contact center. Not only for transcript on real time, but transcript on real time, creating cross-selling and upselling pop-ups to the different operators in order to improve the results with our customers.

Another example, virtual assistance for drivers, we have highways and so on.

Health and safety, we have currently on production a project where we can upload a video with health and safety incident on the platform, read the video, capture with LLM for all, in fact, the video motion port animators and create from scratch a complete reporting with the different steps to solve using the procedure database. This accelerate incredibly the number of hours for our health and safety colleagues to create the proper reporting for their general managers.

It's incredible. And we solve a problem because they was absolutely overloaded with the number of incidents, right?

Integration with Robotics and Other Projects

Other projects we have is in order to be integrated with robotics using our conversational platform in order to dialogue with different robotics. We are investing in robotics and so on.

From Generation 2 to Generation 3 AI Platforms

A lot of people probably think, obviously, generative is so new, you are creating, as you can, the different parts of the business. But it's not real. This is our generation two structure that we launched last year, right? We were absolutely focused on governance for that project, but this is not our current platform.

After that, we changed to a new philosophy. We are currently working generation three.

Agents is your best friend in the future, right? We are creating a platform based on absolutely automated agents that dialogue between them in order to create the proper response.

is incredible because we can including create an agent connected with our internal ERP or including human resources application and including an employee can launch a prompt to the system how many holidays I have this year and the platform response you have 25, my holidays, but you have only five this year because the agent is connected with World Day.

And using the platform in the future, we are working on it, the platform, we be absolutely prepared in order to receive a new prompt saying, okay, but I want to from August to 15th. And the platform responds, OK, I reserve for you the days. You have just only three days, because 15 is free.

Try to think how many iterations you will create using agents. That dialogue within them is incredible. This is, in fact, how it works.

If we create an agent that is an expert to extract information from a database, and we have other agent that is perfect in order to identify which chart is proper in order to represent information from the data, and we have other agent that can execute a Python code, you can prompt, please draw a complete image with the data from that database. And the system says, OK, I have three different agents. I will split the different tasks for them and create a dialogue between them to create the proper chart for the user. It's working currently in our company, right?

Collaborations and Platform Development

The reason we are accelerating the process is that we signed an agreement with DXC in order to create that platform, the most advanced genetic platform, being absolutely agnostic with the technology behind. We are using Azure, AWS, and including Google in order to provide the preferred response for each piece of the platform.

The most important for that is a global cost control because you need to pay to the different models or not, I will explain what, including the security and so on. We can use the different part of that impacting to the platform.

This is part of the complete pillar where we are working on it. This is our generation three. This is real. This is the platform we are creating currently.

This is completely updated, right? We are including Jim and I, and we are working on it.

We are working on it not only in Teams, but including WhatsApp to interact with the platform, obviously with completely security parts of that project and so on, right? And yes, we are using this platform as integrated with Teams. It's inside Teams. And we are creating our virtual colleague through Teams, a colleague that I can call in an internal call using Teams to create a dialogue with the platform.

Project is incredible for that because we are considered to integrate anything, not only our internal developments, but third party, needs player, or any other application with API. We can create an agent to interact with this platform and integrate all of that into the same, just only with a conversation interaction.

The Future of AI in Business

Oh, excuse me.

But the most important part of that, you need to be your own hero. You need to create the different business cases for the company and just only start to think how to solve. You need to take the different pieces to create the proper project.

And yes, the technology is evolving in an absolutely insane speed. It's absolutely true. The technology is evolving in a speed we can afford.

We are suffering all the days with the sensation that we have not enough time. Yes, but create the different business cases and just working on that. Because if a commodity appear after months you are working on it, you are learning with the process.

You can change the tool or the path of the project with other new commodity that solve part of the project you are working on it. But in the middle of the process, you are learning how to solve, how to use that technology. And just when this commodity appear, you will be thinking the next step, not the previous, right?

Closing Remarks

Be your own hero.

Thank you. Any questions?