ML Journal

NTT DATA’s Prasoon Saxena sees agentic AI as the next leap in manufacturing—necessary to accelerate cycles, fill labor gaps, increase global competitiveness, and reshape how humans and AI work together 

 

Agentic AI is shifting the conversation in manufacturing from simply generating insights to executing real-world tasks. Prasoon Saxena, global president and co-lead—Products Industries at NTT DATA, Inc., explains how this new class of AI is already helping companies slash engineering cycle times, support an overstretched workforce, and unlock new levels of productivity. In this special Future of Manufacturing Project-focused Executive Dialogue, he shares why trust, governance and training are critical as manufacturers prepare for a future where humans and AI agents operate side by side.

STEVE MODKOWITZ: It seems like talk of agentic AI has overtaken conversations about AI and GenAI. Can you tell us a bit about your perspective on the market right now?

PRASOON SAXENA: Is agentic AI overtaking the conversation around the GenAI and other AI topics? The answer is absolutely, yes.

What we have noticed in the market is that over the last four or five years, the agentic AI has been in the conversation. It started with GenAI, of course, and with the introduction of ChatGPT and with Copilot, etc., it was definitely helping us with the overall insights.

But what is missing is the execution with the with the GenAI solutions, and that’s where I think the agentic AI is going to take over. Where the clients are looking for not just getting the insights, but the execution of it. So insights-to-execution is where it is heading. We are seeing several use cases that are coming out in the market where agentic AI is being used.

SM: Can you describe a smart AI platform?

PS: Specifically NTT DATA, Inc., we have developed a smart AI platform that is basically helping our clients to have an agent sitting with the humans and work jointly on the plants and on the shop floor and the supply chain and whatnot.

We’re really taking that seriously and there’s a lot of interest in the market.

SM: Can you share an agentic AI use case?

PS: We see with a company called Continental—and this is a public information that’s in the press—it is a $38 billion automotive supplier, and, if you see the engineering in the automotive, the engineering cycle is about six to eight years before you can see the products coming.

And, the Chinese have been able to release the products in two years. Why? Because they have adopted agent AI solutions. So we worked with Continental, looking at the 27 different processes in engineering, and we picked about six processes. So we looked at six processes and we deployed the agentic AI solution. And we are seeing a productivity of almost reduction from two years to six months.

So we are able to reduce the cycle time to six months, which is going to really help Continental with the overall benefit.

So yes, agentic AI is definitely taking over in the market as we see, and more and more you’ll see a lot more manufacturing companies will deploy it in their plants.

Agentic AI solutions are fairly new in the market, and we need a governance mindset while deploying the agentic AI with the humans.

 

 

SM: Will agentic AI play a role in closing workforce gaps?

PS: I do want to highlight that in the North American market—especially with the tariffs situation and whatnot, and with manufacturing coming back—there are almost half a million jobs open right now.

And we expect that to go up to almost two million plus by 2030. And, there are not even enough humans to really do the work.

So it really makes sense for the companies to deploy agentic AI along with humans to develop the products in manufacturing.

I was with a CEO of a forklift company in Tokyo last week, and he believes that 70% of the work that they do in the company is repetitive. So he is looking at deploying agentic AI solutions so that he can invest the human capital in more innovation and driving transformation.

SM: What kind of cautions would you place on agentic AI adoption?

PS: Agentic AI solutions are fairly new in the market, and we need a governance mindset while deploying the agentic AI with the humans.

Do you trust the actions the agent takes? We don’t know that yet. Right?

So there’s always “good-to-haves” and checks and balances as you deploy the agentic AIs to make sure that actions that are taken by agentic AIs are in line with what is expected of them, and gradually increase their responsibilities.

Also, our human workforce in manufacturing have to go to a change management and training themselves—how to coexist with the agentic AIs.

So it’s a journey that’s going to take some time. And it’s a trust that one has to build together, with the humans and Ais coexisting together to really delivering the services for the plants.

It’s more, right? It’s not just looking at the data, but also trusting the actions of the agentic AIs because, the way they interpret the data and the way they’re going to execute it, may not be exactly what you want them to do. And there are some mistakes that have happened in the plants that I’ve heard.

So we have to just watch that carefully, and once we have built the trust with them and then start giving them more responsibilities and start expanding into other areas as well.

Yes, there is a technology that’s available, but it’s up to us how we adopt it, how we embrace it, and how we deploy it.

 

 

SM: What is the future of agentic AI?

The answer to that is the multi-agent environment. You will see that there will be multi-agent environment that is going to get deployed in the plants now, or just in the manufacturing overall in different processes. Whether it is inventory balancing, production planning, you can have two agents talking to each other and getting the work done.

SM: Can you give an example of a multi-agent use case?

PS: Just as a case in point, we are working with a large automotive OEM.

A little bit of a background to that is that there’s an individual. His name is John, right? His real name is not John. I’m just making up the name.

We name this project as a digital John. And the real John is going to retire in two years, and he has immense knowledge about a plant in Kentucky, but nothing is documented well.

So we’re using different techniques, AI and whatnot, to capture the knowledge and document it.

And then, we took one of the lines in the production plant and deployed agentic AI.

So, one, we captured the knowledge from John as he is retiring. And, two, is that we use the knowledge to automate one of the assembly lines.

And we are seeing over $2 million in savings by deploying agentic AI. So that’s really driving the productivity with this particular OEM.

There is a lot of interesting stuff that’s happening in the agentic AI space. The future is really bright as I can see with this technology.

A lot of manufacturers are still lacking the basics: having the right data governance structure, capturing the right information and whatnot.

 

 

SM: Can you talk about smart AI solutions for manufacturing?

PS: For smart AI, we have partnered with OpenAI to build solutions. So now what we are doing is we are basically taking it to each of the industries. Like in manufacturing, we are taking the smart AI solutions and looking at each of the business processes, and we are deploying the AI agents in first the horizontal, like the procurement services, the indirect procurement, the supply chain, etc., now and then taking it to the assembly line and the production plants.

We’re making the agentic AIs really learn from being deployed there and then driving the automation.

SM: How will humans and agentic AI coexist?

PS: Within North America, we will continue to have challenges with the trained workforce.

We have two issues. One, we don’t have enough labor to work in the plants. And the second is the training of the workforce itself with the technology. We have to train our workforce to be ready to work with the technologies like agentic AI.

So we have some work to do as manufacturers to bring our workforce up to be ready for the future.

And then you train them. You go through a change management process, so that they are comfortable working with the agentic AIs.

So, yes, there is a technology that’s available, but it’s up to us how we adopt it, how we embrace it, and how we deploy it.

The important message from my perspective is that AI is pervasive, and it’s here to stay. Embrace it.

 

 

SM: What role does infrastructure and data play in this?

PS: The second part of that is around the infrastructure and data.

Within just north American market, what I have seen is that a lot of manufacturers are still lacking the basics: having the right data governance structure, capturing the right information and whatnot. Connectivity is an issue, for example. So, there is an issue around the foundational block. And, I think, that’s a real opportunity for the manufacturers to build a solid infrastructure as they deploy the agentic AI and what not.

So one is the training and the change management with the humans. The second is the foundational block, which is making sure that you have enough of a strong foundation to really deploy these type of solutions. And at NTT DATA we have all the capabilities to help you—not just the foundational blocks but also rolling out the agentic AI or any AI solutions in the future.

SM: If you could leave the audience with one final important takeaway about AI, what would it be?

PS: The important message from my perspective is that AI is pervasive, and it’s here to stay. Embrace it. Otherwise we will be all behind. And as we have seen, especially in the automotive sector, the Chinese have really taken the lead. So I don’t think we have enough time now. And we have to accelerate the adoption of AI solutions.  M

Portions of this interview have been edited for clarity and length.

About the author:

 

Steven Moskowitz, Ph.D., is senior director at the Manufacturing Leadership Council.