#337 Debdas Sen: Why AI Without ROI Will Die (Again)

What does it actually take to prove that AI delivers real value in the industries that keep the world running? In this episode of Eye on AI, Craig Smith sits down with Debdas Sen, CEO of TCG Digital and Joint Managing Director of Lummus Digital, to explore what serious enterprise AI looks like when it is applied to some of the most complex, high-stakes problems on the planet. Problems like compressing years of catalyst research into weeks, predicting refinery failures before they happen, and accelerating drug development timelines that could determine how long a life-saving medicine takes to reach patients. Debdas has spent nearly 30 years in data and AI, living through every hype cycle from the data warehousing era of 1997 to today's agentic revolution. He makes a compelling case that the AI community has one defining job right now: prove the ROI, or risk another AI winter. We also get into what makes TCG Digital's platform mcube™ different. It is not a horizontal tool. It is a domain-first, agentic AI ecosystem built for the kinds of massive, multi-variable problems that horizontal platforms cannot touch. Debdas breaks down how mcube™ bridges legacy enterprise infrastructure with cutting-edge agentic systems, why hybrid modeling beats pure AI in energy and life sciences, and how the platform keeps private enterprise data protected while still drawing on the best of what public LLMs have to offer. Finally, Debdas shares where he sees the industry heading next, a future where agents from different providers can reason together in a neutral space, where inference and reasoning keep improving, and where the companies that go deepest into domain will pull furthest ahead. Subscribe for more conversations with the people building the future of AI and emerging technology. Stay Updated: Craig Smith on X: https://x.com/craigss Eye on AI on X: https://x.com/EyeOn_AI TCG Digital Website: https://www.tcgdigital.com/ TCG Digital on LinkedIn: https://www.linkedin.com/company/tcgdigital/ (00:00) Introduction and Meet Debdas Sen (01:30) 30 Years in Data and AI: From Data Warehousing to Agentic Systems (03:02) What TCG Digital Actually Does (04:32) Inside mcube™: How the Platform Works (10:06) Domain vs Horizontal: Why Specificity Wins in Enterprise AI (18:29) Catalyst R&D: Collapsing 12 Months of Research Into One (30:38) Predicting Plant Failures Before They Happen (36:51) Solving the Trust and Hallucination Problem in Enterprise AI (44:51) The Six-Layer Architecture of mcube™ (47:05) What Is Genuinely New About Agentic AI (49:22) What Young People Should Study to Work in Serious AI (53:14) Velocity to Value: Why ROI Must Be Tracked From Day One

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Craig S. Smith?

  So I'm gonna start by asking you to introduce yourself. I know you've been in the AI world for a long time, worked for several Fortune 500 companies, now CEO of TCG Digital, part of the ChatterG Group, which is a multi-billion dollar global conglomerate. So why don't you introduce yourself, give some of your backgrounds, a part that's relevant, and then we'll talk about what TCG is doing in the digital transformation world.

• 0:34Speaker 01

  Wonderful, wonderful to talk to you, Craig. In fact, so I I started my career in the area of what used to be called data warehousing in 97, a long time back. So it's coming up on 30 years and I've I've been through all the changes that this area of data and AI has gone through.

• 0:58Speaker 01

  When I was doing my first set of courses in data warehousing, artificial intelligence was in vogue. And and then somehow that that word got lost and we kept talking about analytics, with data, and now it's a it's a full circle. So I've spent a lot of time for all my career in this area of data and AI. I used to run the competency in data and analytics for one of the large consulting houses in Europe and Middle East before I joined the TCG Group about 11 years back. And ever since then, I've been evangelizing this area of now what we call AI

• 1:41Speaker 01

  and trying to help clients around the world. We have a AI platform that we use to to very complex real world AI. So yeah, absolutely wonderful to talk to you.

• 1:58Craig S. Smith?

  Yeah, and and TCG Digital is a digital transformation consulting company, right? I mean, you.

• 2:07Speaker 01

  Well, it's an AI, it's an AI platform company and which helps transform our clients, particularly in the area of energy, in the area of airlines, travel, in life sciences. We also do sports. And so so we do transform companies, but we do it with our AI platform, which we call NCube.

• 2:32Craig S. Smith?

  I see, I see in-house tech. So you're not like a systems integrator where you're you're pulling together tech from outside?

• 2:42Speaker 01

  No, we we well, see nowadays it's become an ecosystem play. So there's always an outside element to it. And and I wouldn't call it outside because now everybody needs to talk to everybody. So we would, so what we do is we have our our our platform, which is NCube. And then we talk to what's happening in Claude, Anthropic, what's happening in Gemini in Google, what's happening in OpenAI, what's happening in Nvidia? And and then we have the cloud providers, you know, AWS, Oracle, Azure, et cetera. So we work in that ecosystem, but the core value that we add is we add it through our platform.

• 3:24Craig S. Smith?

  Yeah, and the uh NCube, the platform, can you describe that for us and what it does?

• 3:35Speaker 01

  So we we can pull data, and this is where it started from. It's gone through a little bit of an evolution in the last few months, I would say. But let's let's trace back to, let's say, what it was two years back. So you would pull data from wherever data is, bring it to a central location, create something like a data lake or a lake house. And all of this is happening within the NCube platform. And then we would have reports, charts, graphs, which is business intelligence. Or we would have machine learning models, supervised, unsupervised. We also have a knowledge graph through which we are able to contextualize data. We in fact have a very large knowledge graph in the area of life sciences,

• 4:21Speaker 01

  where we store a lot of information and being able to.

• 4:27Speaker 01

  To create insights out of that. So and then there is a AI

• 4:34Speaker 01

  layer, orchestration layer with which users can can create models. There is model management and all that to be able to do that. So we work in, we work in large, complex problems, constructs. For example, let's say optimization of a large refinery. So that's a very complex problem. The net value added is usually very large and it takes several months to get that done. So we come in with our platform. We, we, we look at all the data that's relevant. And then we are able to use the machine learning models. In this particular case, we use a hybrid model.

• 5:17Speaker 01

  So we use kinetic models in conjunction with AI models to be able to, let's say, improve the yield or improve reliability or improve just the visibility of the refinery. So that's what it was, let's say, a year, two years back.

• 5:33Speaker 01

  Where we are now is it's become a lot more agentic. So instead of just being a monolithic platform and we used to say that, oh, give us the data and then we will be able to do magic with that data. We often reach out to where the data is and use the external information. For example, if, if Claude has some good information in the, in the particular context that we're working on or OpenAI or Gemini, we would work with that, but keep the private data private. So we work with large Fortune 100, Fortune 500 companies

• 6:15Speaker 01

  where they are very, they need to be very sure that their data does not go out, their model says does not go out. So we take the best that's there in the external world, bring that in. And then we corroborate with the internal data to create insights. So it's, it's a, it's, it's a completely an agentic platform today where we have our in-house agents which interact with the agents from the outside world. And, and these agents could sit in any of these language models or any other platform to be able to create insights. So. Today, we don't say we are a platform player anymore because I don't think there is anything called a platform player anymore.

• 6:59Speaker 01

  So we are a AI ecosystem player. We contribute to to the AI ecosystem so that we can interact in an interoperable manner to be able to create insights. And that's what we do. The core capabilities of the platform still remains. It's deep in machine learning. It's deep in business intelligence, very complex data management structures. But things like model tuning or even inference or reasoning capability, that we use some of the external models to do that.

• 7:38Speaker 01

  So AI is super invasive today. I mean, we can't say that we do it all. I don't think anybody can say that we do it all. So that's what we do right now and it's picking up a lot. Because what earlier enterprises used to do is they should see that here is our data, give us insights from it. Now the problem statement is, here is a business problem, doesn't matter where the data is, get us the best insight. So we try to make the best of what's there in the enterprise and what's there in the external world and to be able to generate value. Sorry, I interrupted.

• 8:19Craig S. Smith?

  No, I was gonna ask if we could sort of zoom out to forty thousand feet and talk about digital transformation, that whole market. Because I hear a lot of, I talk to a lot of companies, I hear a lot of companies that are saying similar things. And from an outsider's point of view, it's very hard to differentiate

• 8:49Craig S. Smith?

  what companies are doing in one domain as opposed to another domain or what one company is doing as opposed to another company who has similar marketing language to be frank.

• 9:07Speaker 01

  So when we do our AI, when we do our digital transformation, we are very domain led. And this is where the ChatGPT group comes in. We have a sister company, Lumm's Technology, which is the largest chemical engineering process licensor in the world. They have three thousand refineries and petrochemical plants and other things in energy where their licenses are run. So MQ works in that space, in that ecosystem, to be able to, you know, increase the yield of a refinery by two percent

• 9:51Speaker 01

  or improve reliability of a petrochemical plant by X percent. So this is real world, large problem sets that we focus on. Or we have another sister company, the name of the company is Labvantage, which works in the area, it's the largest laboratory information management system company in the world. And we work in the area of pharmaceuticals, in the area of life sciences, where we're accelerating the R and D of novel drugs. And that's again, a very large envelope problem statement. And we bring in complex domain encapsulated in the knowledge graph that we have,

• 10:37Speaker 01

  have the AI overlay with that with very, and then take advantage of what's there in the external world. Let's say what's perplexity has or what anthropic has and be able to see what a scientist needs to be able to accelerate the output that the scientist is looking for. So from a differentiation perspective, we are talking about very, very deep domain, whether it's kinetic models and chemistry or whether it's biologics from a life scientist's standpoint. I mean, we've got thousands of PhD scientists who whose day job is to just do that core chemistry or biology. And that's what we bring into play.

• 11:20Speaker 01

  And we do a hybrid solution where the core science is a part of what we provide to the client. So it's not pure AI. So that's our differentiation.

• 11:36Craig S. Smith?

  Yeah, and then when you look at what other companies that are

• 11:42Craig S. Smith?

  Either have platforms or are are building tech but providing consulting to to customers on on how to implement the the tech. Do you see it breaking down by domain expertise? Or is are these a more or are many of the companies more horizontal? I mean, there's this world of companies helping large enterprises transform digitally.

• 12:20Craig S. Smith?

  And that means today that, you know, building agentic systems, you know, in the business processes. And I mean, how how do you see that? Is there just so much business available in the market? Everybody needs to go through transformation that these more horizontal plays

• 12:45Craig S. Smith?

  will find plenty of takers? Or or do you think it's evolving into domain expertise, more vertical plays?

• 12:57Speaker 01

  When business value is added, it's almost always there is a domain specificity to it.

• 13:07Speaker 01

  And that is very critical to the solutions that we provide.

• 13:13Speaker 01

  We we typically work with core functions of large companies.

• 13:19Speaker 01

  If you're talking about enable function, enabling functions like finance, HR, supply chain, etc. That's where the horizontals can play very well. But what we are working with, you know, the type of work we do. So we work with accelerating R and D, or we work with making manufacturing more optimized. There's a huge domain play in in in such cases, and and that's our niche.

• 13:46Craig S. Smith?

  Well, you were talking earlier about insights about that that is the output of of your system. Is it insights or is it implementing agentic workflows?

• 14:02Craig S. Smith?

  Yeah, I mean, it's a business. Yeah.

• 14:06Speaker 01

  At the end of the day, it's it's impact and business value.

• 14:11Craig S. Smith?

  Mm hmm.

• 14:12Speaker 01

  The the means to an end could be could be one or two things, and I'll come back, come to that in a moment. But the end goal is to create business value for our clients. Because for us, return on investment for our clients is super important. If somebody is paying, you know, five million dollars, they should get back fifty million dollars. That's our rule of thumb.

• 14:37Speaker 01

  So the means to that end is a combination of what's there in the legacy infrastructure? And what can we bring in with the agentic infrastructure?

• 14:51Speaker 01

  Trying to say that for these large organizations who we work with, trying to say that, oh, we will completely ignore legacy and build an agentic system, that ain't gonna happen in six months. So the ability to connect to legacy, and there are tools available to do that right now, like the MCP protocol and all that, is is a part of the solution. And agentic and reasoning solutions are very much a part of the solution. But we cannot ignore core data management. Because at the end of the day, you still need to do a fair amount of plumbing because,

• 15:34Speaker 01

  you know, it's it's it's really surprising. I mean, a lot of companies still run mainframes and RDBMSs. And so we will still need to connect to those data systems and be able to integrate that data before actually creating the agents.

• 15:57Speaker 01

  And that's that's the enterprise data. I was talking about the fact that we we we play both with enterprise data and the external data. The external data that's sitting in the language models, that's all done by by the language model players, multimodal model or language model. So we take that information in and be able to corroborate, validate with the internal data. And that internal data, there's a lot of data management that needs to get done with that.

• 16:27Craig S. Smith?

  Yeah.

• 16:28Speaker 01

  So it it's it's not one or the other. It's the we we think value is to be able to make the new data interoperate with the old data to be able to create contextual insights for the enterprise.

• 16:46Craig S. Smith?

  Yeah, uh, can can you give us an example, say a large petrochemical plant uh, comes to you, uh.

• 16:59Craig S. Smith?

  What what's the what's what's the process? You sit down and do an audit? I mean, yeah.

• 17:05Speaker 01

  So so we we always start with the uh, the business problem. The front end of our our uh, services is very much consulting. We sit with the client and try to understand what is it that they are trying to solve that that piece is still traditional, old school consulting. Maybe now it's two weeks, earlier it used to take three months. But that two weeks is still there because I still believe that defining the problem well is actually half the solution.

• 17:45Speaker 01

  Once we do that, and I'll give you an example. So we are working with a a top five oil and gas major. And uh, they are trying to

• 17:58Speaker 01

  understand how catalyst can be developed faster. And catalyst is a is something that they use to use in defining and petrochemicals to be able to uh, catalyze the chemical reaction that happens.

• 18:12Craig S. Smith?

  Mm hmm.

• 18:13Speaker 01

  How catalyst can be developed faster? Now, this is R and D information.

• 18:20Speaker 01

  And this R and D information that this particular organization has, it's very, very uh, uh, private to them. They don't want to give this information to anybody else because this is this is core IP. However.

• 18:37Speaker 01

  The world of catalyst development has significant amount of external information. Earlier, to be able to formulate a new catalyst, and this this area is called formulation. To be able to formulate a new catalyst, they would go back to the experiments that they have done before, what materials was used and and try to come up with a new catalyst. Now, you go out to the external world and see what all information is available. Of all research that's there in the public domain, and there is a lot in the public domain. I mean, every time we go to a chat G P T or a Javna or a Plexid, we always get surprised just how much is there.

• 19:16Craig S. Smith?

  Yeah.

• 19:17Speaker 01

  Then we bring that in. And we contextualize it with the internal information. And effectively run millions of models. You know, in the virtual space, saying that millions of different formulations in the virtual space of with agents reasoning amongst each other and so on. And then come to five or ten or fifteen, saying that go and test this out in the lab. So that's a, this is a real world problem. Now, to be able to come to those five candidates to be tested in the lab earlier, it used to take twelve months.

• 19:57Craig S. Smith?

  Yeah.

• 19:57Speaker 01

  Today, it takes one month.

• 20:01Speaker 01

  So, R and D is completely collapsed. You know, giving giving another example, and this is a very large

• 20:12Speaker 01

  refinery in, this this particular one is in India. And the company is called Hindustan Petroleum. And they have come out with one of the most sophisticated refineries in the world. It's a it's a Lamas technology process called L C Maxx. It's a Chevron Lamas, there's a joint venture between Chevron and Lamas called Chevron Lamas Global. It's their process. And they wanted to make this A I ready on day one.

• 20:41Speaker 01

  Because they wanted to run this enabled by AI on day one from a reliability standpoint, from a optimization standpoint, from visibility standpoint. And the reason we were able to do that is because we were able to take all the information that's inside and be able to bring in what happens in these kinds of refineries from outside to make them run reliably and optimally and be able to do that. And, and, and these are, these are typically multi-billion dollar revenue ecosystems. So that's, that, that's what we are focused on, trying to take really large problems

• 21:23Speaker 01

  and try to see if we can optimize half a percent, one percent of it, which translates into several tens of millions of dollars of, of, of value.

• 21:32Craig S. Smith?

  Yeah. And, and just on the catalyst example, what would be, what would get you that one percent? I'm just curious, what is the process?

• 21:42Speaker 01

  We are trying to bring down the development time, cut down the development time to one third of what it usually takes. So innovation cycle is cut down to one third of it. That's value. I mean, that's, I mean, if it takes five years, they would get it in a year and a half, which means that for the rest three and a half years, they would have a far higher yield going forward.

• 22:10Craig S. Smith?

  Yeah. But, but specifically, I mean, obviously there, there would be different points along the way that you would optimize. Is it providing, you know, the, the chemical engineers with?

• 22:29Speaker 01

  So let's, let's take an example. So I'll, I'll, I'll probably simplify this. So they're trying to come out with a new material effectively.

• 22:39Craig S. Smith?

  Yeah.

• 22:40Speaker 01

  And that new material has ingredients and then do, they do something called a formulation. You effectively mix different ingredients in different ways to come out with a new material. That, that's a problem statement. And that material needs to help some reaction happen faster. That's the objective. Now, there are millions of materials and they can be combined in another million ways to come out with this new material that can be used to catalyze a reaction. So a million materials in a million ways is something that it's humanly impossible to figure out.

• 23:26Speaker 01

  But in the agentic virtual world, we can run as many experiments as we want.

• 23:33Craig S. Smith?

  Right.

• 23:34Speaker 01

  So that's what happens. So learning happens in the virtual world. Agents talk to each other, they reason amongst each other. And then they come out with saying that, ah, this is one combination that could work. So we never go around to one combination. We give them five combinations, ten combinations. Then they test it out in the lab in real scenario to see.

• 23:58Craig S. Smith?

  Right. But you're doing that work of coming up with the five candidates. You're not giving the engineers at the petrochemical plant the tools and the training to do that on their own. You're you're acting as.

• 24:19Speaker 01

  So we are we are giving this tool to the research scientists in the labs of that that particular petrochemicals plant, let's say. So it's it's not the engineers who are running the plant, but it's the engineers who are doing R and D for the plant.

• 24:36Craig S. Smith?

  Right. But you're giving them the tools that they can use. Or are you running, you know, finding the five candidates and then giving them the five candidates?

• 24:50Speaker 01

  We work with the engineers and we work in a collaborative manner to come up with the five candidates.

• 24:56Craig S. Smith?

  I see. And and is this on a on a problem like that? Is is and then you're done or or is there a continuing a relationship?

• 25:09Speaker 01

  There is a continuing relationship because they would still need the platform, number one. Number two is our involvement comes down over time.

• 25:19Speaker 01

  So the first catalyst, they might need, you know, maybe sixty percent of the stuff is ours, effort is ours. By the fifth catalyst, maybe five percent of the effort is ours. So, our objective is to enable them because this is a core function. As I said, we work with core functions of organizations. So we don't, we, we, we don't work on the enabling function. So they would never outsource something like this.

• 25:46Craig S. Smith?

  Yeah.

• 25:48Speaker 01

  So, so, so our objective is to enable them. We leave the platform in. I mean, that's how we make money.

• 25:54Craig S. Smith?

  Yeah. And, and the, uh,

• 25:58Craig S. Smith?

  the, you're combining insights from their data

• 26:05Craig S. Smith?

  with, with insights from your knowledge graph, from your own bank or knowledge source.

• 26:11Speaker 01

  Knowledge, we have, we have that, yes.

• 26:13Craig S. Smith?

  How important is the knowledge that you're bringing in?

• 26:17Speaker 01

  It is quite important because we have a lot of chemistry contextualized in our knowledge graph. So we, we do that. So we, we, we take an external data, let's say all experiments done for various things, whatever is available. And then we curate it into the knowledge graph.

• 26:42Speaker 01

  And, and, and we do the same thing on the biologic side. So it, it's three, three pieces of information. It's our knowledge graph. It's the internal data of the organization and the information that's sitting in the public L L Ms.

• 26:55Craig S. Smith?

  Yeah.

• 26:55Speaker 01

  All three come together.

• 26:58Craig S. Smith?

  Yeah. And, uh, and what differentiates you uh, maybe from another player in that domain is, is the curated knowledge graph, I would guess.

• 27:10Speaker 01

  If competition is another player in the domain, what differentiates us is the platform, the depth of A I in the platform. If the competition is a horizontal A I player, then what's differentiates us is the domain.

• 27:24Craig S. Smith?

  I see.

• 27:24Speaker 01

  The ability to play with domain, data and technology is what differentiates us.

• 27:33Craig S. Smith?

  Yeah. And on these problems, you know, taking the catalyst as as an example, not that I understand. I mean, I know what a catalyst is, but is the tech improving or or the depth of the knowledge in the domain accruing so that you're you're finding better and better solutions? Or are these

• 28:03Craig S. Smith?

  really, it can only, as far as you can go?

• 28:08Speaker 01

  It always learns. It always learns with every single project we learn. We learn from the information that we get in our projects. Of course, we leave the confidential information with the client. We don't touch that.

• 28:22Speaker 01

  But the metadata is is available. So

• 28:28Speaker 01

  and then we always learn from the external language models.

• 28:34Speaker 01

  So it is always learning and and every project that we do is a little faster than the previous one.

• 28:42Craig S. Smith?

  Yeah. Yeah. Where do you see this going? So you're you're optimizing industrial processes in these specific domains.

• 28:55Craig S. Smith?

  I I would imagine a couple of things are happening. One, companies that that undertake this process get a competitive edge over other companies that don't. But also, there's got to be an accumulated economic effect as this kind of

• 29:21Craig S. Smith?

  optimization works its way through an industry and through the economy. Do you see that? I mean, how do you see the the future, given this kind of thing?

• 29:34Speaker 01

  It is, it's actually so relevant with what's happening in the world right now. So if we can get a 2% improvement in the yield of, let's say, motor oil or diesel of refineries, which doesn't need to pass through that little strait of water, which everybody's fighting over. It's huge. And that's happening for us right now. And

• 30:01Speaker 01

  More than optimization, reliability is also very important. So all these plants are being pushed around the world to produce more and more and more, which means that they have a higher chance of failure.

• 30:15Craig S. Smith?

  Ah.

• 30:16Speaker 01

  So we, we predict failures.

• 30:22Speaker 01

  We predict failures and we are able to do this using our, in fact, it's just right behind me called the Agent to Command Center, where there may be a petrochemicals plan running in Nigeria or somewhere deep in Africa, where the subject matter expert may be sitting in Houston. And to be able to look at the reliability of that plan to see that, oh, it shouldn't break down and be able to give advice is critical.

• 30:52Speaker 01

  And it is, it has a far bigger business impact than, than the 1%, 2% improvement in yield. Because once these plants go down, they take like weeks to come back up. So in, in these trying times, loss of production is crazy. And that's what we're helping, helping clients with. So, so the economic value is incredible. I mean, it's, it is very focused on the economic value for us. And while the R and D economic value is not very easy to calculate, our manufacturing size of the use cases, it's quite easy to calculate that.

• 31:33Craig S. Smith?

  Yeah, and as I was saying, that, that accumulates that economic value. I mean, do you, you'll start seeing an impact in, in national GDPs as large.

• 31:47Speaker 01

  I was just talking about this one, one plant in, in India. It's the largest refinery in the country, multi billions of dollars. And so particularly for, I mean, percentage of GDP is a, is a calculation depending on where it is. But it's definitely relevant for large enterprises. It's big enough to be relevant at large enterprises level.

• 32:17Speaker 01

  Now, large enterprises' output adds up to a country's GDP. So I, I, I wouldn't go there, but, but, but certainly it is, it is large enough to be relevant to, uh, to large Fortune Hundred, Fortune Five Hundred companies.

• 32:34Craig S. Smith?

  Yeah. Yeah, I, I mean, it's, it's just, it's an exciting time because of all of these optimizations across industries, um, or, or, you know, it's a, a little bit like, uh, you know, the, the internet, when the internet was introduced and suddenly, uh, information could flow much more easily and, and there was a, a big productivity leap. Uh, we seem to be in, uh, that era again today.

• 33:11Speaker 01

  I think there is a ten percent productivity leap possible today with the Agentic systems. And we are working with clients across various roles to be able to, uh, create productivity gains. Whether it's in manufacturing, whether it's in R and D, or whether it's in supply chain. These are the three areas that we primarily focus on. We don't really look at finance, HR, et cetera. So

• 33:42Speaker 01

  I think the productivity gains are incredible. Some of these are, and, and, and I'm saying to the extent of ten percent. I mean, particularly, let's say in a supply chain. If we are able to optimize across the supply chain, let's say for pharma, right from where you are bringing in the, uh, the drug development process through the, uh, the, the phase one, phase the, the clinical trials process. And can we collapse that by six months?

• 34:17Speaker 01

  That means that it gives the, uh, the, the pharma company six extra months to be in the market. And for some of these drugs, it, it can be, it, it, the, the, the, the values are staggering. So reducing the innovation cycle and, uh, and then of course optimizing operations, I think the overall long-term impact of the P and L can be as high as ten percent.

• 34:47Craig S. Smith?

  Yeah, uh, you're, you're, uh, you know, there's been a lot written about the slow uptake of agents in industry because of trust and reliability uh, issues. How have you guys uh, managed that? Or have you encountered reliability issues?

• 35:11Speaker 01

  So, uh, so, so that's our differentiation because the agents that we use are within the enterprise boundary.

• 35:22Speaker 01

  Some of our clients, particularly in the Middle East, are, uh, this is, MQ is running in their data center.

• 35:31Speaker 01

  We might interoperate with agents that may be sitting with a language model or, or a cloud provider. But the final decision making happens validated with the internal enterprise information. So, so that's our answer to the very question that you asked.

• 35:52Speaker 01

  So by definition, the process that we follow does not have this trust issue. Because it's validated with internal information and it's tested enough with internal information. Uh, does language models hallucinate? Yes, it does. Uh, can it be brought to a place where the percentage error is acceptable? Yes, it can be.

• 36:18Speaker 01

  And, and that's exactly where we, where, where, where we focus. Validate the information that's coming from the external world. Make sure that hallucination is to a acceptable number. Because when it comes to management decision making, they don't care about the third place of decimal. That, that's actually, that's specifically our answer to the question that you asked. We are, we're actually doing very well with exactly this. Because the tech is ours. The, uh, the information, the data is validated with the client. It is validated with the domain. But we take advantage of what's there in the outside world.

• 37:01Speaker 01

  But we, we validate that with what's there in the enterprise space.

• 37:08Craig S. Smith?

  What what are the problems or what are the mistakes that large energy clients, for example, make in trying to implement AI? I mean, do you see, do people come to you after having tried to do it on their own or tried to do with a?

• 37:26Speaker 01

  All the time.

• 37:27Craig S. Smith?

  Yeah, and what are the mistakes you see?

• 37:30Speaker 01

  So in the energy space, what what we have seen is they either go first principles, which is, I was talking about optimization, so you could use something called a linear programming or a multivariate programming to optimize a process. Or they go all AI.

• 37:49Speaker 01

  You know, throw it to a neural network, whatever comes out, let's do that. I think the answer is somewhere in between. That's where we sit in the CUSP. And they so so that's from a risk perspective where they fail.

• 38:07Speaker 01

  But the advantage, the opportunity loss is not taking advantage of what's there in the public domain. Because that today is also quite beneficial. So so hybrid modeling, taking advantage of the public domain data is the answer that we have seen. And the mistakes that clients do is go one way or the other too fast, too far. Because marrying the two is not necessarily easy. Because in the energy space, there is something called mass balance, which is like you have a chemical equation, the two sides have to balance, going back to high school chemistry. And the the chemical engineers like swear by that.

• 38:49Speaker 01

  And that is important. That's what's sitting in our knowledge graphs. Making sure that the answer that we give is something that can get validated by the chemical engineer. So that hybrid modeling, hybrid between chemical kinetics and AI is what makes the difference.

• 39:11Craig S. Smith?

  How long has MQ been around or how long has the agentic layer of MQ been around?

• 39:22Speaker 01

  Well, we MQ has been around for about seven years. I mean, we really launched it about just around the pandemic. So maybe, maybe five, six years.

• 39:36Speaker 01

  The agentic, so we do a release every three months. And we did our first agentic release in October last year.

• 39:48Speaker 01

  No, sorry, in, in July last, in the middle of last year. So we have been, we've been around for about maybe six, nine months.

• 39:56Speaker 01

  So we, we would, I, I would never say would complete. But one version will, we will, we'll reach a milestone of everything that we want to do by our release in June. Because the area of inference and reasoning is evolving so much. That's the area that we are very heavily focused on right now. And the other thing is, there is no real industry market space for agents from different places to come and reason.

• 40:30Speaker 01

  They have to get to one of these platforms or have to stay within the enterprise and reason.

• 40:38Speaker 01

  I think that is a industry innovation that is going to come up. Somebody, somewhere is going to be able to create, I mean, there is the A two A protocol from Google. A lot of agents use that to be, you can do that. But, but that neutral area to reason is not yet there and I think that's where the industry is going. When I'm saying industry, I mean, the AI industry is going. And that's the evolution, that's the roadmap that we are all in. So we, the, the way, it's, it's such a fast moving thing. It's never done. So we've been there for about nine months now with the agentic solution and agentic layer. And it keeps, we, we keep learning from the industry.

• 41:23Speaker 01

  We keep bringing in the pro, new protocols that are coming in.

• 41:27Craig S. Smith?

  Yeah. You said there's a new release every three months?

• 41:32Speaker 01

  We actually do three releases a year.

• 41:35Craig S. Smith?

  Oh.

• 41:36Speaker 01

  But if you really compress it out, it's about three and a half months.

• 41:42Craig S. Smith?

  Yeah, so you're, I mean, that's another question I had. How do you keep up with state of the art? Because it's moving so quickly. So you have people constantly working.

• 41:57Speaker 01

  So we use a combination of traditional R and D and Skunk Works. And if I may say, the Skunk Work team, they don't follow releases. They just do what needs to get done to keep up. So while a release is there every three and a half months, there is always a patch going in every month.

• 42:17Craig S. Smith?

  Right.

• 42:18Speaker 01

  I mean, it's rare that a new protocol has come out and we have not been able to put it in the patch in a month.

• 42:26Craig S. Smith?

  Yeah. Can you talk about the architecture of M Q? I mean, how many different systems are working together?

• 42:38Speaker 01

  So we have a data integration layer.

• 42:42Speaker 01

  We have a data store layer which stores data in raw data and structured data and the knowledge graph. Then we have a traditional machine learning layer with an orchestration on top of it with model management and all that. Then we have a front end layer which used to be more B I and charts graphs. And now it's becoming more formless conversational.

• 43:14Speaker 01

  And then we have a generative layer which accesses all the language models which is outside. And then it also has the ability to do private language models, particularly using Gamma and Lama.

• 43:30Speaker 01

  And then we have the agent declare, which is A two A and ADA MCP and all that. So these are the six components and it's, I mean, MQ can still be installed on a data center, but the footprint is becoming larger and larger and probably going forward SAS is, we, we have a SAS solution, but SAS is probably going to be the only way MQ is going to be provisioned.

• 44:02Speaker 01

  But we do have clients who want it on data center, even private data center, we deployed on Nutanix and VMware and all that. But it's these six modules that gets deployed separately.

• 44:15Craig S. Smith?

  Yep. And then

• 44:22Craig S. Smith?

  from your vantage points, what is genuinely new about the current wave of agentic AI for enterprises versus the, the hype cycles we've seen over the last decade?

• 44:41Speaker 01

  I think it is around inference and reasoning capability is becoming better and better. That's, that's the real, I mean, delta in the last few months. I'm still waiting for that neutral playground for agents from everywhere to be able to negotiate. That's not really there.

• 45:06Speaker 01

  So, but, but I think we've, we've seen the release from Anthropic recently on, uh, to be able to, especially in life sciences space from a domain standpoint and, and of course, rarely people write their own code anymore. So even my R and D is completely transformed where they use significant amount of code generation capability from these language models.

• 45:35Craig S. Smith?

  Yeah.

• 45:36Speaker 01

  So that has significantly, so, so software development is completely changed right now. I mean, it is, uh, it's no longer about what we used to learn about requirements, design, not dev test, but so.

• 45:56Speaker 01

  But but I think inference and reasoning is the biggest thing. And that that's also the biggest area at which growth is happening.

• 46:04Craig S. Smith?

  Yeah, you know, I get asked a lot by young people. If a twenty two year old is listening and wants to work on serious high stakes AI and industry rather than another consumer app or optimizing advertising or some of the other things.

• 46:28Craig S. Smith?

  What very specific things would you advise them to study? Because for a while it was you had to learn to code. And now, I mean, I use Open Cloud.

• 46:43Craig S. Smith?

  Created all kinds of little interesting apps for myself. And I don't code at all. So what would you advise young people to be studying right now?

• 46:58Speaker 01

  I think AI used to be a department. Then AI has become pervasive.

• 47:09Speaker 01

  More of a horizontal. I think the next evolution of AI is going to be sectoral.

• 47:16Speaker 01

  Right from the chips to

• 47:21Speaker 01

  to the use cases. So I think understanding of how the hardware software user interface works. We we we used to use the word called full stack programmer. I'm extending the full stack programmer a little bit into the business side and a little bit into the hardware side. I think that's that's a that's an area that's really going to take off. And the second one is

• 47:55Speaker 01

  so much of business, so much of investments have gone in. I think some four hundred billion dollars went in twenty twenty five or something.

• 48:03Speaker 01

  People will ask, where is the ROI?

• 48:06Craig S. Smith?

  Yeah.

• 48:06Speaker 01

  And the ROI will come from business context. So because actor go deep into the sector, understand the business process that can get impacted. It's it's not AI for AI sake, because little P O Cs will continue to fail. But when you take up a large problem, it it has a higher chance of succeeding succeeding. And to be able to tackle a large problem, one will need to understand the various facets of that large problem.

• 48:38Craig S. Smith?

  Yeah, that's interesting. I'll I'll keep that in mind next time I'm I'm asked.

• 48:47Craig S. Smith?

  I I don't have a lot of other questions. Is there anything I haven't covered that you think listeners should know?

• 48:54Speaker 01

  No, it's absolutely wonderful talking to you, Craig. It's we're in a I keep saying we are in an era of of AI. And as a community, and I'm saying as an AI community, it is it is for us to make sure that

• 49:15Speaker 01

  ROI happens, return on investment happens. We have to take ownership of that that statement as a AI community. Otherwise, I heard of AI in nineteen ninety seven.

• 49:31Speaker 01

  And then it died for, I don't know, fifteen years. Now it's back. Otherwise, it's again going to die. We cannot let it die the second time.

• 49:42Craig S. Smith?

  Yeah, I I mean, that's a good point on ROI. And and that's really interesting, the catalyst example. And you you know, the the time savings and and what the enterprise it it has that much more time in the market or has that much more time uh to develop other things. Because there's there, you know, at the beginning, the chat G P T moment,

• 50:10Craig S. Smith?

  there was a lot of talk about productivity gains, but then.

• 50:16Craig S. Smith?

  Companies realize that, you know, an employee saving an hour of his time doesn't necessarily

• 50:26Craig S. Smith?

  accrue to anything across the enterprise. It just means they take a longer lunch break, maybe. And, and so, yeah, tying the, the, these solutions to real ROI is, uh, is important. And I would imagine that's the focus of the consulting phase that you go through. Is that right?

• 50:50Speaker 01

  So, the, I mean, that, that focus is all true. I mean, one of the, our tagline is velocity to value.

• 50:58Craig S. Smith?

  Right.

• 50:59Speaker 01

  So, value creation is right from the beginning at the consulting phase, and we track that all through the project. And we don't stop till that value is added. And what we commit to our clients is that you'll get to value faster if you work with us.