The Boring Business Revolution: Why This Oracle Veteran Left AI Platforms to Revolutionize Cold Storage

Saurabh Gupta spent 20+ years building AI systems at Oracle, GE, and Michelin before co-founding FroGo. Learn why boring businesses in exciting markets might be the smartest entrepreneurial strategy.

July 20, 2025

5 min read

By Rachit Magon

While LinkedIn overflows with flashy AI rappers and no-code unicorn dreams, the real money is being made by entrepreneurs who've discovered a counterintuitive truth: boring businesses in exciting markets might be the ultimate wealth-creation strategy.

Meet Saurabh Gupta, a 20-year technology veteran who could have built another AI platform. Instead, he chose cold storage. After architecting AI systems at Oracle, GE, and Michelin - including GANs for tire defect detection and locomotive optimization - he co-founded FroGo, quietly revolutionizing India's frozen food supply chain.

His story reveals why the most successful entrepreneurs aren't chasing shiny innovations, but perfecting unsexy fundamentals that actually solve real problems. Here's the masterclass in strategic thinking you didn't know you needed.

Key Takeaways: The Boring Business Blueprint

The Boring Business Advantage:

• Operational businesses are built for endurance, not virality
• Less competition means higher margins and customer loyalty
• Real problems create sustainable business models

Immediate Actions (Next 30 Days):

• Stop chasing flashy consumer apps - identify operational problems instead
• Apply the three-pillar framework: problem size, current solution, scalability
• Look for industries where coordination and process optimization matter

Long-term Strategy (6-24 Months):

• Focus on B2B operational challenges over B2C engagement metrics
• Build systems that orchestrate information flow in real-time
• Use AI to optimize processes, not replace entire business models

Q: Take us back to your time at Michelin. When did you realize you wanted to apply AI technology to something as fundamental as food delivery?

Saurabh: Thanks, Rachit. Cold chain and AI are like poles apart - very different planets altogether. When I was at Michelin, I was fortunate enough to lead AI initiatives. Back then, we were building AI systems for manufacturing, especially supply chain and manufacturing optimization.

Our use cases included voice of customers, demand forecasting, and process optimization. That's when I started setting up a team for research and development to research GANs - generative adversarial networks - to detect defects in tires and scale that process.

Most AI initiatives aren't limited by tech, but by cultural inertia and other factors. Coming from a product mindset, I always believed that whatever business I build will reflect human behavior day in and day out. Product encapsulates many technologies, and AI is one element of it.

When FroGo came to me, it felt like a natural fit. Cold chain is something specific - while we have WMS solutions, warehousing, and delivery, something very specific to cold chain logistics that treats temperature breach as critical is a worthy challenge. Although AI may not solve all cold chain use cases, there are crossroads where AI plays a very relevant role in optimizing cold chain logistics.

🔥 ChaiNet's Hot Take: This is the perfect example of boring business thinking. Instead of building another AI platform, Saurabh identified a massive operational problem where AI could be the multiplier, not the main attraction. Cold chain logistics is a $340 billion global market with genuine coordination challenges - exactly where boring businesses thrive.

Q: You've written five books spanning Oracle databases to psychology. How does human psychology help with building heavy operational businesses?

Saurabh: Authorship came naturally - literature played a big role in our upbringing. When I started writing, there weren't many technology authors coming from India. I grabbed the opportunity to write about Oracle, which I knew best.

Over time, I spent a fraction of my career studying psychology while practicing technology daily. When you start building products, you have to deal with hundreds of users - their mindset, their way of interacting with the product, how they assess and engage with small features you're releasing.

More than introspecting and understanding yourself, you have to be in the shoes of your users. Understanding why a person will take a certain action that gives a certain output - that helps me build products all the time.

Boring businesses with multiple people aren't boring at all - they're absolutely fascinating. But you have to make sure all those fascinating habits get integrated into your product simply. People on the ground have to use it.

Maybe people don't understand what AI features they're seeing. Like when you click a photo on your iPhone, there are AI models working behind the scenes, but they don't appreciate the term "AI" - they just know it makes their lives amazing.

🔥 ChaiNet's Hot Take: This psychological insight is crucial for operational businesses. Consumer apps can rely on dopamine hits and viral mechanics, but boring businesses succeed when they genuinely make complex workflows simpler. Understanding user psychology isn't just nice-to-have - it's the difference between adoption and abandonment.

Q: What patterns do you see when evaluating opportunities? How do you distinguish operational businesses from flashy consumer apps?

Saurabh: Flashy consumer apps anyone can build anytime, but not for long. Unless you retain customers and create stickiness, that's a different world altogether. If I ask you about Web3, most of us don't even know what Web2 is, right? It's all happening as we speak.

Operations-heavy businesses are built for endurance. They're not flashy. If you see my ERP, it's one of the most pathetic ERPs - no colors, nothing. But it works. It solves problems beautifully.

I interpret any venture using three parameters:

First: Problem sizing and framing. Is it worthy enough to be a problem? There could be 100 problems around us, but not every problem needs to be solved with AI, engineering, or a product. It has to be sizable and defined enough that if you and I can define the problem similarly, we've already felt the pain.

Second: Current solution analysis. What's the existing solution? Is it manual operations or a series of steps someone has to go through? What's at stake if it's not solved properly?

Third: Scalability assessment. If we solve the problem, what's the quantum of scale it will achieve over time? Who are the primary beneficiaries? Can this be scaled globally?

In our case, when we act as fulfillment partners for brands listed on quick commerce or with multiple distribution channels, if they don't have fulfillment done on time, what are they losing? What's the size of that loss?

🔥 ChaiNet's Hot Take: This three-pillar framework is gold. While most entrepreneurs ask "what innovative product can I build?", Saurabh asks "what operational problem can I solve where I'm either saving time or money?" It's the difference between building features and building businesses.

Q: What's the opportunity today if someone wants to build an operational business as a startup?

Saurabh: First, find an operational problem. Operational problems are lying in the open. It may be logistics, warehousing - pick any industry. There are no industries except pure SaaS which aren't operations-heavy. Everything is very operations-heavy.

When I say operations, that doesn't only mean building a team of a thousand folks running around. There are multiple systems trying to talk to each other. You have to set coordination within each one and set up the process chain. That's the true nature of operations-heavy business.

We work on the ground with multiple teams and handshakes. Making sure information passes in the correct format at the right time, and the orchestration of that information happens in real-time - that's operations-heavy. That's where most businesses find their own ways of succeeding.

Simple answer: Pick an industry, look at operations-heavy segments within it, see what's happening right now, and identify if you can optimize a portion of it. Can it be optimized with the right success criteria in a limited time?

🔥 ChaiNet's Hot Take: This is brilliant. While everyone's building AI copilots, the real opportunity is in the unsexy coordination problems every industry faces. Information orchestration, process optimization, real-time handoffs - these aren't going away, they're just getting more complex as businesses scale.

Q: What misconceptions do people have about cold chain logistics?

Saurabh: Most misconceptions on the user side: "Is frozen food safe or not?" That's the most common myth I continue to hear, though trends are changing with brand awareness campaigns.

On the operations side, the myth is that if you have a refrigerated vehicle, the task is done. That's not true. Having a laptop doesn't make you a software company, right? Same thing - having a refrigerated vehicle doesn't make you a cold chain logistics provider.

Once you have the cold chain vehicle and stock in hand, you have to ensure temperature adherence stays consistent. When the entire arc is completed, that's fulfillment. The biggest myth is thinking having the vehicle is the end of the story. There's a lot of discipline and handshakes on the ground to fulfill demand on time.

🔥 ChaiNet's Hot Take: This vehicle-laptop analogy is perfect. It illustrates why boring businesses have moats - the complexity isn't in the obvious infrastructure, it's in the orchestration. Everyone sees the refrigerated truck; few understand the temperature monitoring, route optimization, and real-time coordination that makes it actually work.

Q: How should entrepreneurs think about AI - replacement or complement?

Saurabh: This is a big distraction. AI is not replacing humans or jobs. AI is replacing tasks. In large enterprises, you have engines already set up and running - dependent on systems, built using people, dependent on teams. These teams can't disappear overnight.

In large enterprises, safety and security of data matters most. They can never go 100% dependent on AI systems helping them make decisions at scale. AI will replace certain tasks in the process chain, which will make humans more intelligent. It complements intelligence so teams have more time to focus on smarter jobs.

In startups, we have a chance to correct AI systems. What's at scale is smaller compared to enterprises. If my rider is on the road and my map suggests a different route, but this person knows a better route from ground experience, I can learn from that system and make my AI smarter.

When I was at GE, we set up AI systems for locomotives in North America. AI systems would output coefficients used before locomotives start. Once the locomotive is operational, you cannot run an AI system - it's all human intelligence and intervention. Whatever indicators you need must be produced before the engine starts.

For demand forecasting at FroGo, even 80-90% accuracy is fine because it's good enough to indicate how stores should be organized and freezers utilized based on variants, sales outflow, and production capacity.

🔥 ChaiNet's Hot Take: The task-replacement vs. job-replacement distinction is crucial. In operational businesses, AI handles routine decisions (should I accept this delivery? which route is optimal?) while humans handle exceptions, strategy, and complex coordination. It's augmentation, not automation.

Q: In operational businesses, how do you differentiate between needing process optimization versus needing AI?

Saurabh: I'll break this into two parts. When to build an AI system for process optimization, and when to optimize the process itself.

From a product perspective, I always fight to optimize the process first. But running a startup, I don't have that luxury. On-ground processes are built considering available levers - process makers think about the best way without considering product outcomes. They believe the product is there to capture data, upload forms, click buttons.

When I look at productivity of processes or people on the ground, I ask: "What series of steps are you doing right now?" That's where I identify optimization opportunities.

Most of the time, process optimization requests don't come from process makers. They come from users - people in the process. They'll shout first that there's a problem in the system. That's where a third-eye view identifies what portion needs optimization.

Then a fourth-eye view identifies AI opportunities. For example, every time someone needs to record temperature and ask for approval - should I receive this stock at this temperature? The operations head asks four-five questions: volume, time, waiting time, temperature. Most of the time, answers are similar. That's an opportunity - if someone feeds inputs and the system predicts yes/no, that's process optimization.

Process makers won't ask for optimization. It's usually a third-eye view auditing the process or measuring productivity who identifies use cases.

🔥 ChaiNet's Hot Take: This is experience you can't learn from books. The insight that optimization requests come from users, not process designers, is profound. It explains why so many enterprise AI projects fail - they're built by people who don't feel the operational pain daily.

Q: Looking at the next five years, what excites you most about AI in operational businesses?

Saurabh: I'm very excited about advancements in ethical and responsible AI. There have been massive changes in the last couple of years. Ten years ago was just AI/ML - the true sense of artificial intelligence was probably born in the last two years with OpenAI, GenAI, DeepSeek.

There's another angle that got associated with AI that will decide adoption. I'm keen about exploring responsible AI aspects that have molded AI systems. If you see AI image generation, platforms started putting watermarks saying "this is an AI image." When I upload books to Amazon Kindle, they ask for declaration about AI-related content.

This will evolve much wider for AI adoption. Industrial AI will evolve - adoption stands around 18-20% globally. I think with trust building and seeing solutions built using AI, adoption will go higher, optimally 35-40% in the next five years.

I'm not counting AI products like image generation - those are good for democratizing AI so people generate interest, opening opportunities for consumer fancy products.

🔥 ChaiNet's Hot Take: The responsible AI focus is smart business strategy, not just ethics. As AI becomes embedded in operational systems - logistics, manufacturing, food safety - companies need audit trails, explainability, and accountability. The boring businesses that nail responsible AI will have significant competitive advantages.

Q: What's your biggest concern about AI misuse?

Saurabh: My biggest concern is AI floating for free. We don't realize that we are part of the product we're consuming. We share thoughts, content, images, but the ethical or responsible AI layer hasn't hit yet - it will reflect in outcomes, not inputs.

Products available for free are collecting all data without us knowing where it leads. That's a serious threat to AI adoption, partly due to open democratization of AI consumption.

I hope there should be certain norms - the checkboxes we see in long agreements should become much more visible so we understand what we're getting into when we upload images, documents, or ask ChatGPT to do something. That's yet to be regulated. I can only be hopeful it happens sooner rather than later.

🔥 ChaiNet's Hot Take: This data harvesting concern is particularly relevant for operational businesses handling sensitive supply chain, customer, or financial data. Companies using "free" AI tools might unknowingly be training their competitors' models. Smart operational businesses are already thinking about data sovereignty and AI model ownership.

The Boring Business Advantage: Why Unsexy Wins

Saurabh's journey from AI platforms at Fortune 500 companies to cold chain logistics reveals a profound insight: while everyone chases the next shiny innovation, the biggest opportunities lie in solving unglamorous operational problems with excellent execution.

His three-pillar framework for evaluating opportunities cuts through startup noise:

1. Problem Definition: Is this genuinely painful and sizable?
2. Current Solution Analysis: What's broken about how this works today?
3. Scalability Assessment: Can solving this create lasting value at scale?

The cold chain example is perfect - it's not sexy, but India's frozen food market is exploding, and temperature-controlled logistics is genuinely difficult. While competitors build flashy consumer apps, FroGo focuses on the operational excellence that actually moves frozen food from manufacturer to customer without spoilage.

Three Strategic Insights:

1. AI as Process Multiplier: Use AI to optimize tasks within proven business models, not to create entirely new ones
2. Third-Eye Advantage: The best optimization opportunities come from users feeling operational pain, not process designers
3. Endurance Over Virality: Operational businesses compound value through execution excellence, not growth hacking

In a world obsessed with unicorns and viral growth, Saurabh chose the path of building something genuinely useful for an industry that desperately needs it. That might be the most contrarian - and most profitable - strategy of all.

Connect with Saurabh: Find his books on Amazon and follow FroGo's journey on LinkedIn for insights into operational business strategy and responsible AI implementation.

The most successful entrepreneurs aren't chasing innovation for innovation's sake - they're finding boring problems in exciting markets and solving them better than anyone thought possible.