Chainet LogoChainet
Back to Home
The Dirty Secret of Industrial AI Nobody Wants to Admit

The Dirty Secret of Industrial AI Nobody Wants to Admit

An honest conversation with Archit Naraniwal, co-founder of Faclon Labs, about the unglamorous reality of bringing AI to factories with 30-year-old equipment, skeptical workers, and the massive gap between industry hype and ground truth

November 4, 2025
14 min read
By Rachit Magon

Listen on

Everyone talks about AI revolutionizing manufacturing. What they don't tell you is that the hardest part isn't the AI at all. It's everything else.

The problem? Factories run on equipment older than the internet, operated by people who have done things the same way for 30 years, generating billions of data points daily from systems that were declared end of life a decade ago.

Today's guest has spent nine years in these trenches. Meet Archit Naraniwal, an IIT Bombay civil engineer who co-founded Faclon Labs with his college roommates Rishi and Utkash. They're building an industrial IoT and AI platform helping steel plants, cement factories, power plants, and FMCG facilities optimize operations and hit sustainability goals.

This isn't Silicon Valley AI. This is the real world, where you integrate with legacy systems nobody has documentation for, where COVID suddenly made digital transformation urgent, and where your biggest challenge isn't picking between GPT-4 and Claude but convincing a plant operator that a computer won't steal his job.

Key Takeaways: Industrial AI Reality Check

The Real Challenge:

Industry 4.0 Reality:

The Adoption Problem:

Q: Can you give us a background about yourself and introduce Faclon Labs?

Archit: I'm Archit, and I started Faclon with two very good friends, Rishi and Utkash. We all met at IIT Bombay, lived together for all four years in the same hostel, same floor.

When we started Faclon, we were a different version. We wanted to optimize India's water distribution. That was the mission. But the current version of Faclon is a data and AI platform company providing digital transformation for large physical infrastructure.

We primarily focus on manufacturing facilities. How can we help them increase productivity? How can we make their plants more reliable? How can we enable them to achieve their overall sustainability goals?

Anything under productivity, reliability, and sustainability is where Faclon works.

🔥 ChaiNet's Hot Take: Most AI startups pivot once. Faclon went from water distribution to industrial AI. The lesson? Sometimes the technology you build finds its real home in unexpected places.

Q: What's the most difficult part in introducing AI to these conventional, decade old systems?

Archit: First of all, there are two different things. We're not automating the factories. Factories are already automated to a certain level. That's how plants are running.

What we're trying to build is an intelligent layer on top of that. That's where the AI comes in. How can we enable these factories so they can move forward from automation to autonomous?

The automation is already there. How can we make this factory smart? That's what Industry 4.0 means. How can we enable these factories with insights and data driven insights so the people running the show can take faster and better decisions to optimize their processes?

🔥 ChaiNet's Hot Take: The sexiest part of industrial AI is the unsexy part. Everyone obsesses over models. Nobody talks about connecting 15 different legacy systems that were never designed to talk to each other.

Q: You mentioned Industry 4.0. What does that actually mean?

Archit: Industry 4.0 is a terminology and jargon being used, to be really honest. But ultimately what it means is: how can we connect the machines, the people, and the processes?

That's what we do. How can we connect all these three things that run the plant? There are machines, there are people, there are processes. Closely integrate all these three things.

Then build smart, intelligent, data driven decisions and applications and use cases on top of it. That's what 4.0 is. Going forward more than just automation, it's an intelligent layer on the data that's already there, lying there for years, with no intelligence being created on top of that data.

At the end, a lot of AI comes from data. You have to have the right data.

🔥 ChaiNet's Hot Take: Industry 4.0 sounds fancy. In practice, it means making systems built in different decades by different vendors actually talk to each other. Less buzzword, more translation layer.

Q: Walk me through what an AI implementation looks like when there's a 30 year old boiler talking to a 15 year old control system. Where do you come in?

Archit: Great observation, Rachit. We're talking about factories established decades ago. The equipment out there are legacy systems. When they were deployed, no data, AI, or digital transformation was in the picture at all.

It's a difficult and different place altogether where you have legacy systems, and there are different kinds of systems. It's very difficult to bring the data to a single platform.

The enterprises are talking about AI, which model to use and everything, which is for a different type of use case and space. Where we're operating, having data in a single place is the most difficult task.

I'll tell you, at some places we're even integrating with legacy systems which have been declared end of life by the OEM 10 years ago.

Rachit: Wow.

Archit: But you cannot change that because then you need a plant shutdown. For factories, production is the most important thing. As long as it's working, they will not change it. Obviously someday we have to change it, but as it is running, they don't need to change it.

That's a very difficult task. You have to integrate with legacy systems which nobody has any clue about.

We're also trying to work with people who have worked in a certain manner for decades. Factory operators, plant operators are working on the same machine, same plant for 20 years, 30 years. Now somebody is coming and telling them that AI is coming into the picture and will solve your problems.

Anyway, there is fear in people that AI is going to replace people.

By the way, just for our viewers, we are not here to replace any people. We are just there to empower those people so they can make their life easier. For those people to believe that, it takes time. You need to show some results. You need to take smaller steps so they start believing and then start trusting the AI results coming.

Once it's there, then obviously they will do wonders.

🔥 ChaiNet's Hot Take: You can build the smartest AI model in the world. If the plant operator who has run that boiler for 25 years doesn't trust it, your deployment fails. Technology is easy. People are hard.

Q: Have you faced skepticism where people think AI is mostly just marketing with no real value?

Archit: 100%. 100%.

The problem is that the business users who are the actual users for the solution, that is one of the biggest problems. In fact, we also face this. The biggest problem is not how we'll take the data or which model to use. I believe and strongly believe that the biggest problem is: how do you onboard those business users on whatever technology and solution you have deployed?

Ultimately, they have to use it day to day. Until unless they use it, no matter what you put there, it's not going to be valuable for any enterprise.

But you can't blame them. I mean, they've worked in a certain manner for years. And the environment is very different there. If you go there, nobody is free. Factory people are always on their toes because something or other is going wrong in the plant.

It's difficult, but now with all this, the way we're moving forward, everything, obviously the AI buzz, ChatGPT or whatever you say, other things coming in the picture, people have started realizing that AI helps in a certain manner.

So there is some initial level of acceptance. Obviously the skeptical mind is there, whether it's going to replace or what. But once you build that relation with them, once you help them understand how it's going to help them, and you actually solve a couple of problems for them, they start believing your system, they start believing in you, and then start adopting it.

🔥 ChaiNet's Hot Take: Industrial AI adoption is a trust building exercise disguised as a technology deployment. You're not selling software. You're selling belief to people who have every reason to doubt you.

Q: You work with steel, cement, FMCG, power plants. Do you have to rebuild data infrastructure for each customer or is there a pattern?

Archit: Great question. You rightly said that every factory is different, and that is the biggest problem we're trying to solve.

All the factories out there work in a different manner. They have different assets. They have different processes. And if nothing of all of this, then they have different people. People work in a very different manner. Human beings are going to work in a very different manner.

In fact, we've seen for the same customer, different plants, they work in a very different fashion altogether.

That's the exact factory heterogeneity problem we're trying to solve with the platform we've created. It's scalable, modular, and designed in a manner that any need and requirement of a factory can be solved through some drag and drop kind of stuff. You don't need to rebuild the entire platform or entire thing for every customer.

Obviously, there are certain patterns out there. A certain pattern in terms of how the data will be accessed, how the data will be stored. The core capabilities of the platforms are being used. But then obviously you have to give some customization angle or you have to give that context angle to a particular industry, to a particular plant. In fact, to a particular user who is going to use it.

So you have to curate those things for a certain set of users and then a certain set of plants and a certain set of industries. That kind of work we always have to do. And that is something which then becomes your differentiator. That also kind of becomes your moat.

Because if you've worked with so many people before, you know how these things work. And why a new startup cannot actually come today and start building what you've built, because it's not just software. It's people, it's devices, it's a lot more things than just code.

It's not just a piece of software. There are a lot of functional understandings and goals on which the deliverables need to be curated. Unlike a standard solution which goes and can be used by everyone. And that's why this particular sector is not very sexy for a lot of people.

Rachit: Absolutely. But I think it's the boring sectors where you can make maximum impact, right?

🔥 ChaiNet's Hot Take: The moat in industrial AI isn't your algorithm. It's the 50 factories you've integrated with, each teaching you something new about the chaos of real world manufacturing that no textbook covers.

Q: You mentioned that pre-COVID no one was really interested, but COVID happened and you suddenly saw a surge. What changed?

Archit: Especially in our case, pre-COVID we were working in a very different space altogether. Our primary customers were commercial buildings. The industry is something we just recently started, maybe 6 months prior to COVID. We started thinking we should work in the industry. Prior to that, we consciously took the decision that we will not work in an industrial environment.

But after COVID, you see, our business got shut down in a couple of days because no building is working for the next one year or so. We got a lot of chance to think: okay, something which we have built as a product, as a technology, as a capability, where can it be used?

During that time, if you see, the industries were the only thing which was working, which was operating. Those people were also facing a problem because they were not having any access on what is going on in the plant.

The plant which is running on a very bare minimum of human resources. It used to work with a thousand people working there. Now there are only hardly 50 people working in the plant, or say 100 people. How would you manage that?

So they wanted to have all their data on a single tip, which was not there earlier. When we used to go to the plant, they were like, okay, I can always get the data. I'll just call this person and he'll get me the data.

COVID helped those people understand that they have to prepare for something which was never seen. From that point, I believe, especially in the Indian context, the inclination towards having these digital solutions and technology is something which has started.

If you'll see, a lot of enterprises after COVID only started having a full fledged digital team, having a CDO as one of the CXOs. Earlier there was only CIO or IT head, but the digital officer, you will see nowadays a lot of enterprises have started having a separate team altogether who is going to lead this kind of initiative.

COVID changed the narrative. The acceptance got changed in 12 months time.

Rachit: Wow.

Archit: So people started understanding the importance of data and how data can actually impact probably revenue and their KPIs.

Rachit: They experienced the problem firsthand.

Archit: Before that, they were never experiencing that kind of problem because the data was always available to them, even though it will take some time. But when COVID came, they actually realized the data is not available to them.

🔥 ChaiNet's Hot Take: COVID was the forcing function that made industrial digital transformation urgent. When you can't walk to the control room, suddenly remote data access isn't a nice to have. It's survival.

Q: What has been the hardest part of your 9 year journey in building this?

Archit: In the last nine years, we've pivoted a couple of times. But I would say the current version of Faclon is something that's like 4 or 5 years old, post COVID. It's a new version of what as a company we have been doing.

Building business around this particular space, there are two challenges we've faced.

First is obviously the acceptance in the market. Like I said, these people are working for decades in a certain fashion. So it's very difficult to change the mindset of people. It's very difficult even if we deploy the solution. We get a very hard time enabling those values for the customer because the adoption takes time.

There are a lot of challenges: how do you make sure that the people are using what we have provided? So that is the biggest challenge.

But I believe, like I said, we always try and find some or other thing in our deliverables for every user who is out there so that we can enable that stickiness to the platform. That they come to the system for something which is solving a real problem for them. But that's still a challenge. The speed is basically a challenge. How we are going to bring the adoption, that is like the biggest problem.

The second challenge, which is what becomes our USP, is that like I said, these factories and equipment are decades old. If you go into different factories, different plants, they have different legacy systems, different OEMs out there.

So how do you build a system around it which is scalable in a modular fashion that you can quickly integrate those data points and bring everything to a single platform? We're talking about factories which are generating billions of data on a daily basis.

In a factory, if you go, there are 5, 6, 7, 10, 15 different kinds of systems, the data sources. Bringing everything to a single platform and ensuring their IT policies and everything is being taken care of, it's a challenge itself.

🔥 ChaiNet's Hot Take: The two hardest problems in industrial AI: convincing humans to trust machines, and convincing machines from different decades to talk to each other. Neither problem is solved by better algorithms.

Q: You've been doing this for 9 years. Backstage you said the industry is still very nascent in India versus your US customers. Is that a polite way of saying people don't get it in India? What's the blocker? Is it trust? Budget? Experience over computers?

Archit: It is obviously a stage. It's a journey for everyone. Even in fact, when the West started, it was probably 10 years before us. Now they are in a stage where everybody understands that.

So the blocker is obviously unlocking the mindset of the people who are there. Like you rightly said first, people believe that this is some fancy thing which is coming, and right now we don't need fancy things.

But at the end, they also understand. It's not like they don't want to optimize their factory. Nobody, everybody would like to do that. Everybody would like to solve the problems for them. But it's the trust which needs to be built.

I guess it's a journey which everybody is going through. Every enterprise and every solution provider is going through that. The industry gets mature over a period of time. There are, if there will be a lot of successes in the industry, and then that's how, if you see post COVID, people are investing a good amount of money.

Earlier, prior to that, when we used to go, 90% of the time we were like, okay, there was not a second meeting after the first one. Now the numbers are reversed.

So that change is coming. I think it's just a matter of maybe a few years, maybe 3 years, 4 years, 5 years down the line, we'll be at the same level where the West is right now.

Rachit: Amazing. That means there's a change happening and the interesting part is that you're among the people who are leading that change.

Archit: The change is happening. As people will see results, obviously the budget will come. Once they see ROI, once they see the value which is being delivered, everybody would like to invest in something which is going to help them optimize their overall operations. Of course, the results need to be seen first. 100%. Yes.

🔥 ChaiNet's Hot Take: India's industrial AI market is where the US was 10 years ago. The flip side? You're early to a massive wave. The downside? You have to educate the market while building the product.

Q: You're selling to very business oriented people in factories. I'm sure you're not finding them on LinkedIn. How does your sales funnel look like? How do you reach a new customer?

Archit: It's not like our customers are not on LinkedIn. Like I said, in corporate offices you go, there's a separate team, a dedicated team altogether of CDO teams. In fact, all these plant people are also on LinkedIn nowadays.

So LinkedIn is one of the source. But I would say the primary source for us is obviously connecting with these people, building relationships with the people who are out there, and solving some problem for these people.

To give you context, over the last 3 or 4 years, if you take India as a market for Faclon, out of 100 manufacturing related companies from the market cap, 24 are our customers.

But it took us a journey, some 3 years, that we did something or other for a customer, for a plant. After that, they started trusting you, that you have delivered something. Once that happens, then obviously you can talk to the corporate people or you can talk to them in order to scale that horizontally across the plant.

That's how our funnel gets generated. In fact, most of the revenue is from our current customers only. And that's any B2B business, right? You get value from your existing customers.

Rachit: And basically, I think word of mouth would probably be your topmost source.

Archit: Word of mouth. In fact, those people, once they go to a different company, then they bring us. And all these people are very connected. In fact, the industrial people, that's why you see a lot of closed networks or closed location wise, closed places are there where all these factories are set up. People know each other there, and then obviously word of mouth happens for you.

Rachit: Fair. I think that's true for any B2B enterprise software, right? You need to build that trust first before you can actually start selling or upselling your products.

Archit: 100%.

🔥 ChaiNet's Hot Take: Industrial sales runs on handshakes, not LinkedIn ads. You get one customer, solve their problem, and they tell their college roommate who runs a plant 200 kilometers away. That's your funnel.

Q: You help industries meet sustainability goals and use less gas, less water, lower emissions. But at the end of the day, sustainability is a cost. What is the other side of the table looking for as a KPI? Is it save the planet or save 20% on our energy bill?

Archit: Interesting question. Obviously, in the shorter run, the money saved is the primary objective for the people.

But definitely, if you see the enterprises who are creating their plans and objectives to make their products, make their processes, make their factories more sustainable, with all these net zero commitments coming from different people around the world across different industries, not just manufacturing but across industries.

So definitely in the shorter run, saving money is the primary motive for the people to implement this kind of solution. But they always have at the back of the mind that if not today, tomorrow, they want to become sustainable.

Now, if you see all these listed companies, they have to actually show what they are doing in their sustainability in their annual reports. So people are serious about that. But it is a journey. It will take some time.

Money saving is a shorter exercise and you can see results very quickly and very fast, say 6 to 18 months. But making your organization sustainable, changing the way people are working, changing the way people are behaving in the organization, it's a longer journey and probably will take some 10, 20 years to achieve their goal.

But in the shorter run, obviously the money, and the longer run, obviously the sustainability is always there at the back of people.

🔥 ChaiNet's Hot Take: Sustainability sells in boardrooms. Cost savings close deals. Smart vendors lead with ROI, knowing the green metrics will justify the budget when annual reports come around.

Q: For our listeners who are engineers or early founders attracted to the shiny AI side of things, what should people understand about the boring, unglamorous infrastructure before they can get into the AI space?

Archit: If you talk about our space where we're working, obviously the data integration or enabling data from various sources to a single platform is the unsung hero.

Because everybody thinks about what is there at the top, the AI, the results coming. But especially for our industry, the data integration is the answer. That is the boring or the difficult work.

Everybody wants to work on creating this model, that model, working on different tools which are out there nowadays. ChatGPT, this model, that model. Then people want to use Vercel, Lovable, Cursor, all these different things.

But especially for our industry, understanding how to integrate with legacy systems requires a lot of deep understanding of how these systems actually work. That's the first part.

The second is: how can you give context to data? What is the value out of the data which is coming? It's not just the AI which is creating the magic. Ultimately, it is giving some results which has to create value for the people.

But in general, if you see, for people who are looking to start something in AI, my suggestion would be obviously start with something which is very niche or solving one problem at a time for one customer or one set of customers. Then from there, you see and keep iterating, interacting with your customers, and then see what is working for them, what is not, and then improving on top of it.

So find some niche, start with that, and then see where it takes you.

🔥 ChaiNet's Hot Take: Everyone wants to build the AI model. Nobody wants to build the data pipeline. But in industrial AI, the pipeline is 80% of the work and 90% of the value. Unsexy infrastructure wins.

Q: It's been nine years. You've had multiple almost shutting down experiences, multiple pivots. You're fighting the good fight. What keeps you going?

Archit: Honestly, it's the belief what we have, everyone at Faclon. Especially we as founders and the senior management or the core team of Faclon is that what we have built or what we have delivered is just maybe a surface, or we have just scratched the surface which is out there.

That's what our belief is. There is a lot of value which we can bring through the kind of technologies and solutions we are working on.

Secondly, obviously in the last 9 years, there has been a lot of tough times during COVID. In fact, pre COVID also. Nowadays also it comes. It's not like it's a smooth journey. Everybody, every entrepreneur out there or everybody who has started a business or is doing a business has to go through these cycles multiple times over a period of time.

But what is enabling us to keep moving every day is that we always believe that we have come this far. We have spent 9 years, built something. We have customers who are using it. So we have not just come this far to stop here.

That is something which, I believe, everyone in fact in the senior leadership, and then we as founders, is allowing us to work, in fact on those days when things are not going in our favor. And that happens most of the time. In very few days you find something good happening.

But it's a journey. It can change your mood in a couple of hours in a day. Roller coaster things are happening. But yeah, I mean, believing in our product and our solution, and believing that we have spent and everybody has given their time and everybody has given more than time, their heart to it. We've come this far not just to come this far. There is a longer journey where to go.

So I believe that is what is keeping us going through whenever the rough patch comes.

🔥 ChaiNet's Hot Take: Nine years in industrial AI is not a sprint. It's not even a marathon. It's more like climbing Everest in slow motion while everyone asks why you didn't just take the helicopter.

Final Thoughts: The Unsexy Truth About Industrial AI

Archit's closing insight: "We've come this far not just to come this far. There is a longer journey where to go."

The bottom line: Industrial AI is not sexy. It's integrating with systems declared end of life a decade ago. It's convincing 30 year veterans to trust a computer. It's building data pipelines that nobody sees but everyone depends on.

The AI models everyone obsesses over? That's maybe 20% of the work. The other 80% is the boring stuff. Data integration. Change management. Building trust with skeptical operators. Understanding that a cement plant and a steel plant might both be "factories" but they're completely different beasts.

If you're looking to build in industrial AI, here's what matters: find a niche, solve one real problem, show results, build trust, repeat. The companies that win won't have the fanciest models. They'll have the deepest relationships and the most battle tested integrations.

The future of manufacturing isn't about replacing humans with AI. It's about giving those humans superpowers through intelligent systems they can actually trust and use. That takes time. That takes patience. That takes a stomach for the unglamorous work that nobody tweets about.

But it also means you're building something real, something that makes factories run better, uses less energy, creates less waste, and helps experienced operators do their jobs more effectively.

Q: How can people connect with you and continue learning?

Archit: I'm very active on LinkedIn. That's the best place to reach me. You can see what I'm doing professionally and what I'm currently pursuing. If you have questions, I'm open to providing guidance and can chat anytime.

Final words: The hardest part of industrial AI is everything that comes before the AI. Data integration from legacy systems. Building trust with skeptical users. Understanding that technology is easy, but people are hard. If you want to work in this space, embrace the boring work. That's where the real impact happens, one data point at a time.

Related Shorts

Explore short-form videos that expand on the ideas covered in this blog.