Legacy Rules, New Intelligence

Your core systems might be old, but the logic inside them still runs your business. We break down how to modernize those foundations without losing the rules that protect revenue, compliance, and reliability. With guest Ankit Shah of Netweb Software, we unpack what AI truly does well—rapid code analysis, business rule extraction, dependency mapping—and where human judgment remains non‑negotiable: understanding intent, setting policy, and approving change.

We start by reframing modernization as risk reduction, not just code migration. Instead of “lift and shift,” we advocate a risk‑first approach that identifies operational bottlenecks, compliance gaps, and talent risks before choosing the right pilot. From there, AI becomes the accelerator, compressing months of discovery into days and making it feasible to build a living catalog of business rules that everyone can read—engineering, audit, product, and operations alike.

Then we get specific. In manufacturing, inventory management, and production planning are prime candidates for business rules rejuvenation, especially where legacy assumptions about lead times and batch sizes no longer hold. In healthcare, patient‑centric data flows and medication logic demand rigorous compliance; AI can surface undocumented exceptions while clinicians and governance teams decide what to change. We also explore continuous improvement: AI as a control tower that monitors rule drift, runs what‑if scenarios, and maintains documentation, while humans approve deployments and uphold accountability.

Whether you lead a global enterprise or an SME, the path is the same: start with risk, pick a focused pilot, let AI handle the heavy lifting, and reserve human expertise for intent and compliance. If you’re ready to turn opaque legacy logic into clear, governed, and adaptable systems, this conversation gives you the playbook. Enjoyed the episode? Follow the show, share it with a colleague who’s wrestling with legacy tech, and leave us a review with your top modernization challenge.

Transcript

Alex Romanovich 0:27

Hi, this is Alex Romanovich, and welcome to Global Edge Talk. Today we're joined again by Ankit Shah of Netweb Software. Hello, Ankit.

Ankit Shah 0:35

Hi, Alex.

Alex Romanovich 0:37

So we're continuing on this series that we call the AI, you know, in legacy transformation. And today's specific topic under that umbrella is going to be on business rules rejuvenation, as I call it. There's a lot of talk today about legacy systems, legacy applications, and how AI is going to revolutionize and the entire environment and is going to capitalize on this amazing technologies and models and so forth and so on. So let's dig into this topic a little bit because I'm sure a lot of the listeners that are going to be joining us will kind of be interested, you know, what is this legacy transformation NAR is all about? A lot of the critical systems today, as we know, are still running on legacy systems. And, you know, as you and I know, we uncover, as we talk to different customers worldwide, we uncover some really interesting old artifacts like AS400. I mean, I couldn't believe recently I was talking to somebody and they're still running on the AS400. Some people don't even know what that is, right? It's a very old, it's like you know, running on the original Mac Apple computer. And then, you know, a lot of them are still using COBOL, they're using Power Builder, they're using, you know, uh Microsoft Visual Basic applications, and of course, a variety of mainframes, right? So, yeah, that's a lot of iron, and that's, you know, this thing's look bulky and so forth, but maybe a modern uh iPhone uh 17 Pro Max could probably do the uh computationally do the uh the same amount of power. But how do we reconcile AI and this new tech with the legacy systems? Is there actually hope that we can extract something useful and valuable from those systems and then you know eventually retire them?

Ankit Shah 2:21

Alex, yeah, absolutely. I think uh uh you know knowing that they are still AS400 or COBOL systems is is it's amazing in 2026. But uh, you know, it's it's important to know that these applications have been working for years, I think 20 years, 25, 30 years, and they're still working for a reason because they're highly reliant. The transactional integrity is is extremely good, right? Uh so most of the enterprises who still use, uh, especially the AS400 and COBOL, uh, you know, they are going by the principle that let's not try to fix a problem that isn't there, right? Let's not try to solve some a problem which is not there. But with AI, I see AI, at least at the moment, as an accelerator to modernize and to transform these legacy assets, right? Whether it's AS400 or uh you know an application that was written just about 10 years ago. It's it's classic.NET or Visual Basic, like you said, right? So AI certainly is a great tool, it's an accelerator, but it's not something at the moment where you can just feed something in uh like an AS400 application and out comes like a perfectly written uh microservices-based, cloud-based uh AI native application. I don't think we are there yet, but it's a great accelerator to kind of uh modernize some of these old applications.

Alex Romanovich 3:50

So what you're saying is that uh if the enterprise decides to go into AI, and by enterprise I mean, you know, a larger organization, a larger entity, if you will. And not necessarily in terms of billions of dollars, but also in terms of complexity, it could be a hospital system of five hospitals, for example, right? It could be a um a uh large manufacturing or medium-sized manufacturing company, it could be a logistics company, and so forth, right? So by enterprise, we mean multiple locations, multiple business units, multiple systems, obviously, right? It could be a you know, standalone system writing for accounting and billing, could be another standalone system uh for inventory management, you know, and so forth and so on. And of course, as you know, and as we know, a lot of these systems were built in the past as kind of this um almost semi-isolated monolithic systems. And then there was always an issue, okay, how do we make them communicate, right? How do we extract data? You know, we were designing uh centralized data warehousing with data markets. Remember that word data mart? Some people don't even know what that means, right? But um it's all coming back though. But now we have some really cool new tech, we have some really great new technologies. And the question is, is it realistic to take a semi-itonomous system, old system, and extract the business rules, extract the data, make it uh, you know, kind of virtualize it completely, right? And on top of it, on top of it, have AI edit, right? Go, you know, have AI go at it. And um, you know, is that the right approach? Or should the approach be, hey, let's isolate, you know, let's go into a pilot stage first. Let's prove to the enterprise that it can be done. Let's uh maybe even train the model, maybe even create your own model, which is very possible. And then we're gonna deploy this across the board. What is the right approach?

Ankit Shah 5:54

Based on the type of application and and what AI can do at the moment, it's it's very important to understand what AI can effectively do for enabling modernization or accelerating it. It can very easily extract business rules, right? It can it can uh allow you to understand, analyze the legacy code, which earlier would have taken months, now it can be done in in days. But what it cannot really do at the moment very effectively is understanding the business intent, right? So for example, you are having a, and I'll make it very simple, you have a financial application, and there's a there's a rule that uh do not process uh any transaction over 10 million after 4 PM, right? Now uh AI can tell you that there's a rule like that, but it cannot tell you why that rule is there, right? Is it because the legacy system had limitations that it could not process 10 million as an amount? Or is it like uh a legal requirement for regulatory reporting, right? So so you can map the rules, you can extract the rules, but uh you absolutely need to know the business intent, which again uh AI cannot really tell. So, you know, use the word reject rejuvenation, right? Business rules rejuvenation. So that, in my opinion, is still highly human-centric, right? Uh what you call the human in the loop, right? You have the rules extracted by AI, but how to change the rule, when to change the rule, why and why not, is something that has to be governed by humans at the moment.

Alex Romanovich 7:31

You know, you it's funny you mentioned this world word uh rejuvenation. I use this word actually as almost an analogy to the wellness and well-being, if you will, of a human body, right? I was involved with uh some of the wellness and the well-being applications in the past. We actually built the system, and I was involved with some really cool uh wellness projects in the past as well. The word rejuvenation doesn't just imply, hey, let's convert some legacy code into new code or Python or something that AI would understand. I think what I meant by rejuvenation is that as we're converting, right, as we're doing the conversion, we're also, and maybe even extracting business rules. You know, these business rules of a semi-autonomous legacy systems or system needs to be enhanced, needs to be sort of uh given some light or given some energy and needs to be integrated. Maybe there's something that was going on in the past 10 years, right, within the company. And this one thing, one system that's kind of sitting to the side, and we can't really get rid of it because there's very limited documentation. Or there's some uh, you know, for some reason, you know, a lot of things are tied to it, maybe some transactions, maybe there's some dependencies on outside partners, whatever the case may be, right? So there's a certain level of hesitation. Plus, it's a it's an issue of resource, right? It's an issue as, hey, where do we focus within the enterprise? Do we focus on this new thing, or do we focus on maintaining the or or you know, converting the uh the old thing, if you will? And I didn't mean to simplify this, but the word rejuvenation to me means not only convert, but also um add, also enhance, integrate, bring in, adapt, and and so forth and so on. So what do you think about this? Is is you know converting code to new code, is that just enough? Or uh the does the enterprise have to really plan this thing out in a lot more careful way?

Ankit Shah 9:34

In my opinion, it is not enough, right? I mean, modernization is simply not just taking old code and and converting it into new code or a new architecture. I think modernization is essentially a controlled way of eliminating or reducing your operational or business risk. So unless the enterprise thinks how and what operational risks are there, right? What are the risks that they want to really reduce or eliminate? And what is the strategy? Just changing or migrating the code would be part of the strategy, a small part of the strategy, and which is where the largest gain of AI is, right? Uh, but but you know, unless there's an absolute competitive need that I have a product, I have a CRM product, and you know, when I go out and sell it, it's not being sold because it's in Power Builder. Now, in that case, you simply need everything to be translated or migrated to a newer technology with all the errors or all the processes as is, right? So if if there if you have a competitive pressure like that, yes, a simple code migration is is warranted. But otherwise, I think it's not just a code migration, but but more of a strategy to kind of minimize or eliminate the operational risk that the legacy application is is uh you know.

Alex Romanovich 10:54

So what you're saying is that it's not just, you know, if you have a conversion or modernization project, you you you don't start with the project itself. You start with the risk assessment first. Absolutely. Absolutely. That is very important. And then the other piece that's important that you and I, you know, are uh you know coming across a lot of is um you know compliance, right? For compliance heavy, regulatory heavy environments, healthcare, financial services, to a certain extent, energy, you know, um some of the other ones. We encounter a lot of these uh systems that need to be, and you know, of course, compliance changes, you know, it has changed in the past 20 to 30 years, right? So that legacy system may not even be compliant anymore, correct? Uh to some of the new standards, whether they're OSHA standards or IEEE standards or ISO standards, whatever the or you know, uh anything that we can think of, right? So part of rejuvenation, if you will, or part of conversion, we have to worry about compliance, right? And then there's another risk involved. Um I would like to I would love for you to give me a couple of examples and our audience in maybe some of the industry segments, right? And what type of legacy systems or legacy logic that might be critical, that might be more critical, let's say in the manufacturing environment, right? Manufacturing supply chain environments and so forth. I know you have a lot of experience with manufacturing systems, so having worked with them for the past 15, 20 years. So now the world of AI is here. What would be some of the examples of the legacy systems in the manufacturing environments that would be really good candidates for AI transformation?

Ankit Shah 12:36

In manufacturing, I would say uh the way inventories were managed in the past, supply chain, especially, production planning, some of the business rules, some of the way uh a lot of these legacy applications manage the inventory or do production planning were based on the old ways of doing it. With the newer technologies, newer ways of doing things, uh, a lot of these processes have changed, which is the main drawback of a legacy system that it cannot change as fast as some of these newer processes or techniques that come up uh in your operations. So that's uh obviously an area where uh in a manufacturing supply chain where uh modernization or regeneration would help, right? Because that is still for any manufacturing, your inventory management and production planning remains like the most important, the heart of manufacturing. Uh as far as healthcare is concerned, I think uh patient-centric data. We all know how how important that is. The way drugs are handled, right? Uh, there may be rules in legacy systems that kind of uh you know interact or allow or enable or uh not enable how uh certain drugs are dispersed or prescribed or diagnosed. So those are obviously some of the things that probably need to be transformed, right? Now, again, some of these rules, uh, and that's where what we talked earlier, Alex, as to why the business intent or understanding the rules are important, because the legacy systems are full of uh uh logic or or business logic which have been added by an enterprise as a result of that experience. So some of the exceptions, some of the rules, maybe to remain compliant, maybe to remain uh you know uh legally or or to to uh you know uh uh be statutory compliant. Just by because you are rejuvenating, you have to be very careful that you're not kind of uh you know messing around with those rules that have been undocumented but but are there for a purpose.

Alex Romanovich 14:50

Are we moving towards an environment where not only we're transforming some of the legacy systems and um you know extracting the business rules and and uh you know converting the code or whatever, but we also have a great opportunity now to set up a system so that AI, potentially, can continuously update and refactor these systems, right? Or at least the uh the instance of the of the uh of the enterprise architecture, right? Have AI now document it and track it for any type of changes. Or maybe track it against the ever-changing regulatory environment, or track it against new systems being brought in, or so forth. Do you think that AI has a role like that now?

Ankit Shah 15:36

Absolutely. Absolutely. That's that's the uh that's the best use of AI in in modern systems. Again, uh what we think of when we think AI is that AI is there, it it does everything and it puts into production, so your production systems are being continuously upgraded and and you know enhanced. That is the scenario which I believe is we're not ready for that yet, right? I mean, agents can do it, but uh for all the reasons that we've uh that we discussed earlier, it's not practical for AI to make independent decisions, you know, especially in regulated environment. You you cannot have AI or the agents make decisions for you and put it in production. So monitoring, analyzing, running what if scenarios, uh those are those are the things that can be done in real time in a in a way predicting certain things, right? So so you you have AI to help you do all of this, but the actual implementation is still and should and the governance should still be in the hands of humans. I don't think we are in a state where we can just hand it everything to an autonomous agent. But everything before that, absolutely, AI, AI is is enough that can certainly help in that.

Alex Romanovich 16:50

I see, I see. Well, I I think I agree about the human part. Speaking of humans, let me ask you last questions for this podcast as we continue in our series and tracking updates, tracking trends uh and developments within the enterprise. For the CIOs and CTOs listening to this uh broadcast, what do you think should be the first practical step towards legacy environment transformation, modernization, and as we call it, rejuvenation?

Ankit Shah 17:21

I I sincerely believe that a risk assessment, uh, you know, uh starting from top up rather than because AI typically works from bottom up, right? It it looks at the code, it analyzes the code, extracts the rule, and and presents something. But uh rejuvenation or transformation is successful and will be successful if the intent, if the risk is correctly identified, and and what the strategy is correctly defined. I think uh that would be the first thing to do. Uh, you know, we have those 30, 60, 90 rules for CIOs, so all those strategies should be done and and then let AI do the heavy lifting and so on. Uh that's what my advice would be.

Alex Romanovich 18:08

And I also think that this is something that's reachable for small and medium-sized business as well. I mean, we're used to uh big players like Accenture and McKinsey and BCG and IBM Global Services, you know, spend millions and millions of dollars on strategies and risk assessments and so forth with large organizations. And they're, of course, getting the benefit of it. They're getting the benefit of the latest technologies, they're getting the benefit of the latest trends in transformation and you know, moving forward. But I think small and medium-sized business, with the advances in AI, has the same opportunity now. You know, working with smaller companies like NetWeb Software and Jem and some of the other ones, our partners, I think we, you know, the medium-sized business has a great opportunity to be in step with the uh, you know, with the big guys and um, you know, take advantage of the similar technology.

Ankit Shah 19:00

Absolutely, absolutely, because some of the things that we talked uh that AI can do very fast, which used to take months, now it can be done in days, uh, you know, uh, is where SMEs were constrained earlier with resources. They couldn't really afford or have the wherewithal to put a team of resources to kind of analyze, reverse engineer. But now, with the help of AI, those things can be done as fast and as efficiently as possible with a smaller team with minimal investment. So AI certainly has a great help and to SMEs. And we're seeing that. And we are seeing that a lot of SMEs coming up to us, regardless of the domains, asking us how can AI help us? What is the low-hanging fruit? What can we do here? So absolutely, SMEs are the ones that are most excited about how AI can be leveraged for what they're doing.

Alex Romanovich 19:52

Absolutely. Ankid, thank you so much. As we move forward with our series of podcasts on AI transformation within the legacy and small, medium-sized environment. I want to uh thank you. I we appreciate your insights and uh be well. And I know you're traveling a lot. You're you're a global trotter, uh globe trotter, so to speak. Uh and uh we hope to see you very soon.

Ankit Shah 20:16

Thank you so much for Alex uh for for having me here. And I I know you travel as well. You just back uh from Europe, so safe travels to you as well, and see you soon.