運用知見からエージェント型産業AIへ――IFS CTO ヴァイバヴ・クマー氏に聞く

Welcome, everyone, to the Emerge AI in Business podcast. Today's guest is Vaibhav Kumar, Global Chief Technology Officer at IFAs. Vaibhav joins Emerge's Aishan Yaziki to discuss how industrial teams can capture and access operational knowledge as experienced workers retire. He explains how conversational AI could make business information easier to use while keeping people in control of agent action. Today's episode is sponsored by IFAs. In this episode, we cover how leaders can prepare their people and identify the business data needed to put AI to practical use. To go deeper on this topic and learn how leading organizations approach AI investment more like a venture portfolio and why interdisciplinary collaboration is critical to defining the right data for AI success, Download our free PDF report, beginning with AI, at emerge.com slash AIK1.

That's E-M-E-R-J dot com slash A-I-K-1 to download your copies. Now the conversation with Babel. Let's start from what you're seeing on the ground in effort-intensive industries, so like manufacturing, energy utilities. You have a background in energy as well. So these businesses have complex equipment and infrastructure that needs to keep running. So most often they have aging assets, they have supply constraints, they have regulatory demands. So where are teams feeling the greatest strain today, and how is that affecting day-to-day operations? So there's three macro trends that I see within all of these industries which are impacting each of these industries quite significantly.

The first is the fact that there's an aging workforce in these industries, which is going out of the door. And a lot of institutional knowledge goes out of the door with that, which is creating a massive skills shortage in these industries. And it's quite evident across the world. The second one, I would say, is the fact that there are some unsolved critical business outcomes or business challenges that these industries face, which technology has the promise to unlock, but companies have just come to reconcile with the fact that their operations will run like that. But there's a massive opportunity for business efficiency, which they're struggling to unlock completely. Think of things like unplanned downtime, which is a massive thing. If you're in aviation, if you're in manufacturing, if you're in oil and gas, all of those, that costs real money and can get you fines and whatnot,

if you're critical manufacturing or aviation or whatever. And then the third one I would say is that given the geopolitical climate, it, there's a lot of disruption to supply chains, which means companies now have to start to become even more flexible and adaptable, right, as these things happen, which is again putting a strain on companies' ability to move fast, make quick decisions, change, and then, you know, decisions made with half the data, of course, you know, can be super risky, right, and cost you money. And tell me a little bit about the workforce from the conversations, you know, You mentioned before that work for some of this critical knowledge, you know, living within the workforce, which is now leaving at its peak. So tell me a little bit about the problems with that. The reality is, I mean, when you have people that have spent so much time in your company and they build institutional knowledge is latent. And the problem that you've had with software for the longest time, enterprise software is it's just super inconvenient to go in and get data into the system.

right? So it's a friction, right? It's a point of friction. Using that software is a point of friction for your most experienced worker who is well respected by everybody else. So they don't, right? Conversely, the people who are diligent and that do maneuver these screens and go into that data, that data is just sat there but not being used. So you have heaps of potential knowledge that you're just not using because, again, these screens are a friction for you to operate and be able to find that data and make sense of it. That's what AI has the ability to unlock. AI has the ability to build super consumer-grade experiences that allow you in a frictionless manner to... It's not that people don't want to share their knowledge. It's just people are frustrated. The frustration that you experience when you use bad software is super high. So AI has the ability to build consumer-grade experiences, which takes that friction away.

Equally, it has the ability to actually unlock all of that latent data and the knowledge that sits within there for your incoming workforce. And from your own experience, could you walk us through what happens when one of these challenges turn into an actual disruption, maybe, with an example? I mean like the example that Mark gave this morning is a nice one right Like Boeing is basically predicting that because of the skills shortage it going to have planes on the ground because there nobody there to maintain those planes So like you know I think there's probably enough of these kind of micro-level examples in different countries and different industries, right? And definitely these micro-examples in different specific companies itself, right? Where because of that skills shortage, your operations come to a halt. And so you need to make it

more convenient for them to enter that data into the system. And if you can do both these things, you create a flywheel. And that costs you millions, if not billions. Perfect. That's helpful. And maybe let's get into what AI is changing here. It was a big part of today's conversations as well, having insight into what's happening is one thing, but also understanding what to do and getting it done is a completely different beast. So how is AI helping industrial teams close that gap and which capabilities are maybe making the biggest practical difference for like difference on the front line, basically? See, at the end of the day, what the frontline worker requires is very purpose-built, persona-specific experiences that they can put to use in a frictionless manner, right? So they require software that is just easy to use, right?

You look at IFS intelligence, right? That is an example of effectively bringing consumer-grade AI experiences into an enterprise setting, right? It's the same experience that you're familiar with Claw or ChatGPG. Now you're layering that on top of an enterprise business context. That's the power, right? So all of the products that we're building for the front line are products that are very much grounded in a consumer-grade experience because our belief is you give them a consumer-grade experience and they will use it because they're very used to using those consumers, right? Don't make it complex like enterprise software is. And tell me a little bit about how do we keep human in the loop in that process? Because going back to the first question as well, you know, there are a lot of regulatory aspects of these conversations. So within these different processes, how do you decide what AI can handle on its own and where a person needs to basically stay involved?

for the fundamentals of when we build software we want to leave the human in control of the agent which means the human drives the agent so if we showed you in stage 2 years with intelligence it's the human that always drives the agent it's only by a human's choice do you decide well actually I want agents to just work in the background for me and when you do decide that you're then given the option and the ability to say, okay, when do you bring me in the loop, right? So you can decide those things. So by design, what we're doing is we're building the capability for humans to harness agents. And we're also giving the humans to go further and say, okay, I want to build background agents. So it's always about keeping the human in the loop and keeping the human in control. That's perfect. And let's make maybe this a little bit more practical for leaders listening today and this episode as well.

So for folks who are listening, for leaders who are listening and thinking, how do we get our organizations ready for this? Where would you start? So are these workflows become more intelligent and autonomous? What needs to be in place across their data, maybe their workforce and the way they oversee AI's decisions? you cannot underestimate the importance of foundations before you can get AI to work so like with what we did with intelligence intelligence only works because you have a foundation of IFS systems in your landscape so having the foundation is super duper important, now the data is important now aspects like data quality and data cleansing that actually get helped when you put agents over the top because at the end of the day, the power that you have within the LLM today can actually recognize duplicate data, can help you cleanse that data. So I don't think you have to be fearful of the fact that my data needs to be 100% quality

before I apply AI or put AI over the top of it. But you need to have systems with data. That's number one. Number two is you need to be choosy and very, very deliberate about which AI you use and which AI you don't. In the world of technology, there has been buzzword bingo for the longest time, right? So you see everybody package their integration APIs into MCPs and sell you MCPs, right? But you have to think about, you know, meaningfully, what are the products and capabilities that are able to actually give you the generalizable consumer-grade AI experience versus not, right? So it's both about the foundations of the data and then being very deliberate about which AI products you choose and what capabilities they give. Yeah. And that data readiness that we're talking about here, when is that data incomplete and how do you decide whether it reliable enough to act on with your customers or your clients I think incomplete or not putting AI over the top of it is useful Okay And the way I think about it is data has gravity right

So where is the gravitational pull? Which systems have the most meaningful data in your company, right? Go and think meaningfully about how you can put AI over the top of that, and you'll start to see good things happen. Interesting. And if the leaders could just take one practical step in in the next few months, what would you recommend as they are preparing their organizations for the shift that's happening right now? I think the most important is the literacy of people with the capability that AI has today. there is a lot of because of the public discourse around AI there is a lot of people that misinterpret that public discourse to say well AI is not capable that's not the point the point is what you're seeing is you're seeing a lot of innovation in AI that has happened in the lower layers of the stack and true value for a business is created when that AI starts to get into the application layer

and the enterprise software layer that's what companies like us are doing now But that's not been the case and isn't the case today across the board. So the more you see that happen, the more this narrative and the public discourse will change. But just because you don't see that happen today or not enough of it doesn't mean AI is not capable. So the best thing the leaders can do right now is to prepare their people to be capable and literate of understanding AI. and that doesn't just mean your engineers and your doers. That means everybody, top down. I think there's a big opportunity for leaders to set the right example. Just using the consumer-grade tools like Cloud and ChatGPT requires a level of knowledge. What does context mean? Why do I need to start new threads or a new chat, as some would call it? What is the skill? And how does a skill get used?

These are the difference between an output and an input token. These are the kind of basics, like super fundamentals, that everybody should be literate about. You're telling me basically getting, understanding what's happening is the core piece of how we need to get ready for the stuff that's happening in the industry. And that includes everybody from boardroom to frontline workers. It's the same thing that happened with cloud, right? Cloud was this new entity in 2010, right? and by the time you went into 2015, 2018, even the executives were able to articulately talk about cloud and hyperscalers and regional redundancy and all of these concepts, right? And I think that's the phase that we're in right now. We need to be super attentive, right? Building the awareness is the first step to understanding how you can apply this technology and where you would apply it. Is this one of the things that you would say if overlooked, it would be a hard thing to fix later,

or is there anything else that you... You will lose out. You will lose out. I mean, at the end of the day, technology, innovation, once it sticks, it becomes the next layer on top, right? That's what you're seeing. AI is the layer on top of, particularly in the enterprise context, AI is now the layer on top of your enterprise systems. If you don't get involved and don't deliberately look at how you can apply AI to your enterprise software, you're going to lose out because that is going to become the norm in terms of what your users expect. And that is going to become the norm in terms of the type of business productivity and business outcomes that you can achieve. That's going to become a point of differentiation and in the years to come, it'll just become a point of being in the competition. It'll become table stakes. And from all the different verticals that we were talking this morning, in these different keynotes, energy, utilities, manufacturing and whatnot. Which of these verticals do you see are making the headway when it comes to adaptation?

That's an interesting question. I think the reality is because this wave of AI is so prominent and so popular, I think that every board is telling every CEO to drive an AI strategy, right? So I'm seeing it from the big companies to the small, from the manufacturers to the utilities to the airlines. everybody is figuring out what their AI strategy is. So I think the propensity to change is there. There's actually a panel that we do tomorrow morning which you can come to. The purpose of it is literally to try and distinguish between what are breaks that people may be applying erroneously versus what are actual safeguards that you need to consider when you build AI for critical industries like we serve. And I think that's the reality of what's going on. right? There's leaders who want to figure out an AI strategy, but you have middle management who are not incentivized to actually be, you know, interested and involved around understanding the AI capability and in the

earnest applying it because perhaps they're scared for their future, right? Or other things, right? Maybe they're just, like, you also become a bit cynical about new technology as you grow older, right? So for all of those reasons, there is a massive dichotomy that's happening within the organization at this moment in time and there is a conviction that the enterprise world is going to find soon enough when enterprise software with native AI capability becomes a norm they just going to have to invest in this right And that's why I say that right now, the first movers who invest into this in the US, that will be a point of differentiation for businesses. And soon enough, it will just become table stakes. Yeah. And now, nowadays, everything, even a month, makes a big difference. And you mentioned a disconnection. Like we were talking about AI, you know, taking people's jobs. And now you're also telling me there is a disconnection between, you know, C-level management, metal management, and possibly the workers and frontline workers as well.

So how do you see some of your clients, customers overcoming this? And we talked about like getting ready for AI, but educating ourselves and whatnot. not how do we overcome this connection between the different levels of management and their incentives applying AI across their workflows and organizations? There's a lot to be done around human-centered AI, right? Like this narrative that took shape within the industry, and this is my personal opinion, of course, and this narrative that took shape within the industry that, you know, AI is going to take all your jobs, which is so hyped, right, for a whole year. and then you actually go into actually the amount of jobs are increasing rather than reducing. It has kind of poisoned, let's say, the opportunity with AI. And I think the experiences that you have in the consumer world or even like the productivity tools that you use in your business day to day, right? So you use ChatGPT or Claude. Like none of that actually claims to do your job. It makes you super powerful.

what took you maybe three hours to do, you're now doing in minutes, right? Like before, if you had a PowerPoint that you wanted to create, right? If you're an executive and you wanted to create a presentation, that's a three-day job. Build a storyboard, create the slides, you know, maybe form it out to someone else and then review meetings and blah, blah, blah, blah, blah. Now, you still need to do the work. You still need to be the creator of it, right? So you need to, and I do it when I want to create presentations, right? I'll write very, very long prompts, right? I describe exactly the detail and the intent of what I want. But then you feed it into Cloud, and out comes a presentation within 15 minutes, which then you can spend the next maybe half an hour, 40 minutes, you know, perfecting, and suddenly you have it, right? That's superpowers, right? So that's exactly what we are trying to do with IFS Intelligence on top of your enterprise data. You bring that consumer-grade capability into the enterprise context,

and you start to democratize the access that people have. Like today, if your CFO wants to ask a question, your CFO asks the question who never, ever goes into the system where the data lives. So he just asks the question, and then the VP has to go and answer that, and the VP never lives in that system. So the VP asks the manager. The manager probably doesn't know how to maneuver the screen, so the manager gives it to an analyst. and by the time that Chinese whisper reaches the analyst, the analyst probably doesn't understand exactly what the CFO was asking. Analyst pulls out something in Excel and then perhaps uses a plot to build a nice presentation that goes back up the chain. CFO goes, I don't know what you were listening to or not, but there's not what I wanted. Now change this, blah, blah, blah, blah, blah. Versus a CFO just going into intelligence and saying, I want to know this, and within a minute. So that democratizing of knowledge, of insight, through a consumer-grade AI layer that sits on top of your enterprise software, I think is a game-changing capability which will make businesses super productive.

And you just need to make sure all of your employees, all of your organizations, members understand that to fix that disconnect. And now everybody has an incentive to make it work, to implement it. Exactly. Perfect. Okay, great. So I'll thank you so much for your time, Wade. I think we are done. Perfect. This was very enjoyable. Just in time. Just in time, exactly. Wrapping up today's episode, let's look at the three key takeaways from our conversation with Vaba. First, examine both sides of the knowledge gap, how experienced workers record what they know, and how other employees find and use information already stored in business systems. Second, define which actions people direct and when agents working in the background must bring a person into the process. And finally, identify the systems containing your most meaningful business data

and built AI literacy across the organization, including executives and frontline employees. In this episode, we covered how leaders can prepare their people and identify the business data needed to put AI to practical use. To go deeper on this topic and learn how leading organizations approach AI investment more like a venture portfolio and why interdisciplinary collaboration is critical to defining the right data for AI success, download our pre-PDF report beginning with AI at emerge.com slash AIK1. That's emerj.com slash AIK1 to download your copy. For further executive level analysis and to join our network of leaders delivering workflow impact with AI, visit Emerge.com. On behalf of the team Emerge, we'll see you on the next episode.