Ben Ko on Why AI Adoption Is a Trust Problem

Jeff Kavanaugh:

I’m Jeff Kavanaugh here at the Infosys Connect Conference in Los Angeles. Here with Ben Ko, head of Kaleidoscope, part of Infosys. Good to see you.

Ben Ko:

Thanks Jeff.

Jeff Kavanaugh:

Looking forward to talking with you. It's always exciting. You are leading a lot of the product innovation, product design for the company, and we're talking here about AI value, how to unlock it. What are you seeing with your work and your clients in this front end of the business with AI?

Ben Ko:

So you're right, Kaleidoscope Innovation is the product design and engineering arm of Infosys and for three, four decades now we've been inventing the future products that each of us use, either as consumers, video game players.

Jeff Kavanaugh:

Surgeons.

Ben Ko:

Drivers, surgeons, nurses. We touch a lot of products that we use every day in life. These are physical products and digital products and then the whole supply chain to make that a reality after you discover what problem needs to be solved. So AI is manifesting a whole lot of ways, as you can imagine. Very different ways across these...

Jeff Kavanaugh:

Because on the one hand you might think, a physical product, maybe AI wouldn't play as big a role, but you're saying it does.

Ben Ko:

It absolutely does. Think about discovery of what needs to be solved. Use the operating room for an example. There's so much data in an operating room, but much of it is inaccessible. It's qualitative data. It’s in someone's head. It's unstructured, it's in someone's head. A nurse knows this, a technician knows that, the hospital administrator knows something else. AI is allowing us to bring that together and say, there are physical and digital product solutions to these problems, but we have to know what the problem is first. So how are we using AI with leading hospital systems on the provider side? How are we using it with the pharmaceutical companies in drug discovery and accelerating clinical trials? How about surgical robotics, where yeah, it's a robotic physical system, but there's a lot of artificial intelligence, whether we're talking about data AI and truly software AI, or physical AI, which is a massive part of Kaleidoscope’s business. We talk about robotics and automation and how that looks in an operating room is completely different than how it looks in a manufacturing environment. But the fundamentals of how you discover a problem and then reduce that to a potential solution, it's very, very similar. And that's what we call product development. And so our teams are using AI across all of those systems, sometimes to help product development be more effective, to be faster, to be higher quality, but sometimes it's even discovering what are the next things that we need to be working on and focus our attention on those big problem statements.

Jeff Kavanaugh:

You can buy something, you can implement it, but how do you get people to adopt this?

Ben Ko:

AI is very much a change management problem as much as a technology problem. And when we talk about rolling out a new piece of tech in a product, the old way of talking about it was, there's a user error, they didn't understand how it functions, and usability gets in the way. That's what we call human factors engineering, human factors design. When you talk about implementing these kinds of technologies within a company, not only do you have the user interaction and the understanding of how the tech works, but you've also got this element of trust or “that's not how we do it.”

Jeff Kavanaugh:

That was where I was headed with the trust part.

Ben Ko:

Absolutely, and so by no means does this technology supplant the need for good change management within a company. And so part of our team's work… we have a whole design research team where they immerse within one of our clients' businesses. And it could be in a warehouse environment where physical AI, robotics are coming in and working side by side with humans, or it could be in more of a corporate design office where AI is being used to accelerate graphic design, user experience design. And our design research team are embedding in these spaces with the designers, the researchers, the associates, the manufacturing operators understand their opinions, their feelings on these autonomous systems, understand what keeps them up at night or what do they wish was easier about their job, showing how this technology can actually solve their real problems, not replacing them. And then you start to turn these potential resistance into advocates. But you have to do the work of being there with the humans who are doing this and bring their voice to the forefront and then incorporate that into your design process.

Jeff Kavanaugh:

Getting specific, what have you seen in a company or a warehouse or a surgical suite where someone has created this value, how they’ve seen it, how are you measuring it?

Ben Ko:

It's as varied as the environment in which it's being deployed in. Sometimes it's throughput, it's yield, it's failure states, how many course corrections can you do?

Jeff Kavanaugh:

But, you’re seeing a lot of it already?

Ben Ko:

Absolutely. We’re seeing very real measurable outcomes where it’s this metric increased by 28% because of automation, or the decision making was able to be accelerated by 15% because you have better data insights earlier, faster. But it's not just more data, it's interpretable data. So it's contextualized, not just saying here's a long list of metrics.

Jeff Kavanaugh:

Insightful data.

Ben Ko:

And you can now take multimodal streams. So what I'm getting from my corporate dashboard is aligning to what I'm seeing in the manufacturing environment dashboard. And here's my inventory management system.

Jeff Kavanaugh:

You're bringing together these disparate systems.

Ben Ko:

And that's really the nature of physical AI. We talk about it a lot as if it's all robots and humanoids and things. But if the robotic operating system isn't connecting to your warehouse management system, to your manufacturing kit execution system, to your ERP, to all the other three-letter acronyms that run these businesses, then you're leaving a lot of opportunity on the table for actual efficiency. And each one of those tools and the environments in which they're used will have their own metrics of success. And so again, it comes back to your research, design, engineering teams need to understand what makes those leaders successful in each of those environments and those tools, and customize the solution.

One size does not fit all. The great news about AI is it's very adaptable and flexible, but you still need to have the expert context to ensure that when it is rolled out, the change management’s done well. And that you can prove the value because you’re spending a lot of money on this and corporate leaders want to know what's the value that I'm creating.

Jeff Kavanaugh:

I see two big buckets. One is you're working with a product development organization for their products to drive their innovation, or it could be for their process improvement teams, like in the warehouse. What's been the biggest resistance and what do you have to overcome for people to agree with either your numbers or the value proposition and actually adopt it?

Ben Ko:

If I use the example of the teams who are using AI to… I'll use design and engineering, the R&D teams. There is still a lot of concern about intellectual property, privacy, data security. If I'm using AI to help me invent a new piece of technology…

Jeff Kavanaugh:

Will it go up in someone else's cloud and my competitor gets it.

Ben Ko:

We don't want any accidental breaches or disclosures of intellectual property prematurely because then you can risk your patent ability. You've got concerns about what if I'm uploading patient data. All these concerns we've been talking about for half a decade now, or really decades, but with AI in particular, it jumped to the forefront five, six years ago. Those are still there. Now the good news is there are contextualized tools like Topaz Fabric that can live within our clients' data systems where their data does not have to leave their environment, does not have to leave their sandbox. So you've got a much more mature set of tools led by the large LLMs and the large hyperscalers, but it’s where we're seeing Infosys, Kaleidoscope Innovation, all of our companies have built very, very mature models now and environments where we can reduce the concern because they're made specifically to address those challenges. No one is trying to sweep the privacy concerns or the data breach concerns under the rug. We're facing it as an industry out in the open and developing tools specifically to address that. And I think that's where we're seeing adoption accelerate because the legal teams, the compliance teams, the regulatory teams are all saying, finally, I'm being heard. I see tools that meet my needs, not just the cool stuff that the engineers and designers want to see.

Jeff Kavanaugh:

What's the one thing that you're most excited about in the next six to twelve months as it relates to AI in your field of work?

Ben Ko:

I'm really excited to see how it is making the product development process more accessible. By that I mean we have designers and engineers who spend their careers learning how to illustrate beautifully, learning how to work with computer-aided design programs, CAD programs, to come up with gear systems or come up with really complicated electromechanical assemblies, of course, highly complex code. But you've got business leaders across all of our clients, even our own companies, who might be non-technical experts in their field, or they’re experts in a non-technical field I should say, but need to be involved in the product development process. These might be marketers, they might be finance folks, they might be sales folks who are working with the consumers, the customers, the users of the technology that these companies are trying to invent. We can now involve those leaders in the product development process and their ideas can be brought to life.

Jeff Kavanaugh:

So you’re democratizing the product development process.

Ben Ko:

I’m very excited about this because we've seen it with GenAI image generation, but that's just like the baby step. What happens when you've got natural language processing for an engineering CAD system? The difference between the image, where a “oh that's cute”, there are dimensions to this, there are tolerances, there are specs. Is it manufacturable? How much is this thing gonna cost?

Jeff Kavanaugh:

And will that be able to be passed it down to the CNC machine and actually make it.

Ben Ko:

Exactly. And that's what we're doing this within Kaleidoscope’s facilities where we've got a machine shop.

Jeff Kavanaugh:

So, you're living it. You're living it before you...

Ben Ko:

Our designers and engineers are, you know, they're working in their CAD systems, they're getting 3D prints made, they're going to the machine shop, they're cutting steel, they're walking it into our operating room, and then that cycle repeats in a couple of hours, not weeks, or months.

Jeff Kavanaugh:

That has to help you as you’re doing it with your own clients.

Ben Ko:

And it’s incredibly fast for a set of professional teams like we've got, who this is their world. But now we can start to bring in folks who aren't part of this world into our process. And that's really where co-design… the phrase I used earlier is design with, not design for. We're designing with these folks. And we're involving nurses in our process. We're involving manufacturing associates in our process who have never worked on an engineering problem before. They’ve got great ideas.

Jeff Kavanaugh:

What about these agents? How do the agents play into this?

Ben Ko:

Agents, I think, become an augmenting tool. So certainly there's opportunities for efficiency. Processes can go faster, but it also allows us as a company, or any of us as business leaders, to redesign our own processes. So I think if you just take agents out of the box, they're custom agents, but you just take them out of the box and you apply them to your existing systems, they'll go faster. They'll allow your human team to focus on the problems that they really need to focus on. But if you really take the time to redesign the processes for the agents working with humans and you're transforming yourself into an AI-native… this is again the hard change management problem.

Jeff Kavanaugh:

Oh, you're talking to the non-AI-native area.

Ben Ko:

Now you’ve got the exponential opportunity where it's not just, okay, we went from here to faster. Now look what we can achieve. I think that's the unlocked potential of agents, but it's gotta be through really good, detailed service implementation.

Jeff Kavanaugh:

In product design you have regulatory bodies, both within a country and across countries? What's the role they need to play, whether it's for patents, whether it's for safety, as AI makes all these things faster and more possible?

Ben Ko:

We as humanity need to get our heads wrapped around open systems. Even though we're talking about artificial intelligence and I'll use medical diagnostics as an example. Open models, AI models that are continuously learning from new data, are highly restricted or not permitted by FDA because there's a question of as that model keeps evolving, it's now different than what is approved.

Jeff Kavanaugh:

And how do you revalidate it? How do you prove that it's just as safe as what the data showed in the clinical trial system or the medical device approval process?

Ben Ko:

So the regulating bodies are getting their heads wrapped around how AI is being used to accelerate some of these steps, but not necessarily as the product itself in a fully adaptive, evolving software model. I think that's a huge next step that we're gonna have to confront.

Jeff Kavanaugh:

What if medicine can all of a sudden have a cure for cancer or have that wonderful outcome but it's not approved?

Ben Ko:

I mean, the reverse is true as well. Something bad can happen, you prevent.

Jeff Kavanaugh:

But it's a balance, isn't it?

Ben Ko:

It absolutely is. I had the pleasure of listening to Dr. Francis Collins, who was the director of the NIH, talk about a patient-specific treatment. This is unthinkable. I mean, it was science fiction. It’s that nirvana for life sciences now. And now they can develop a gene therapy for a specific baby, for her specific disease, based on where her progression is. It still requires the discovery, the clinical trial, the validation, the safety. But because of technology, you can accelerate so much of this to make it possible for one person. Now, that's one example that happened just in recent year. But how do you scale it now?

Jeff Kavanaugh:

And I think that's where we're heading. And make it affordable.

Ben Ko:

Make it affordable, absolutely accessible. And it has to be safe all along the way. But this is where I think the science fiction now just becomes the science future of how do we get there? It's through hard work and dedication of scientists, engineers, developers, regulating bodies, but it's happening and that's incredibly exciting to me.

Jeff Kavanaugh:

Ben, it's always a pleasure talking with you because you're doing such interesting work. It's both important technology-wise, business-wise, but you're also solving problems for real people.

Ben Ko:

It keeps it fun.

Jeff Kavanaugh:

Good seeing you, Ben.

Ben Ko:

Thanks so much.

Jeff Kavanaugh:

Take care.

Ben Ko:

Thank you.

Jeff Kavanaugh:

I'm Jeff Kavanaugh. Until next time, keep learning and keep sharing.