Three Thousand Investors and One Unanswered Question

You Can Watch the Full Episode Here 👇

The easy version of Andy Pandharikar's story starts at Cisco, moves to the virtual fitting-room company Fitiquette and its acquisition by Myntra, then arrives at Commerce.AI, a profitable company, building voice and chat agents for contact centers.

The more revealing version begins with a boy in Mumbai who learned by taking toys apart and rebuilding them. When his father's hearing disability made work uncertain, Andy helped support the household by assisting a leather-belt business while studying engineering. He crossed an education system that did not always know what to do with a child who wanted to build more than he wanted to test. Then he landed in America and recognized himself in the people around him.

Andy builds to learn. He wants money because it creates freedom to build. He survived one company's darkest stretch because he believed the thing deserved to exist. I found that instinct deeply compelling. I also left with questions about Commerce.AI's cost advantage, localized models, endless hours, and the responsibility AI companies carry when efficiency becomes somebody else's lost job.

Andy did not describe his arrival in America as alienation. He described recognition.

Growing up, he modernized toys for friends. The academic system around him was optimized for examinations, not curiosity expressed through taking something apart. He assumed that if he kept building and learning, engineering would find him. It almost did not.

Immigrants often arrive with abilities their first country could not fully name. America gave Andy an ecosystem in which the builder identity was immediately legible. At his first startup, people from different backgrounds shared his appetite for making things. He remembered working until three in the morning on his first day and feeling exhilarated.

His larger American lesson came later. India had taught him how to make and sell. America taught him to optimize for scale. The young man who sold belts by cost, margin, and turnaround time began thinking about products, channels, enterprise sales, and repeatability.

Immigration did not replace Andy's operating system. It gave it a new instruction. The Indian builder met the American obsession with scale.

Fitiquette began with a Cisco advertisement showing virtual shopping. Andy asked whether Cisco was building the product. The answer was no. The ad was selling networking infrastructure.

His response was pure founder instinct: perhaps we should build it.

The technology was ambitious. Digitize clothes, understand fit, visualize how a garment would sit on a body, and predict size across brands. But the business model remained unsettled. Should the company sell to consumers or brands? Andy said they kept swinging between the two and never completely solved it.

A brilliant product can still be homeless inside a weak business model.

The final stretch was painful. Andy put in his own money. Resources tightened. The ending was unknowable while he was living it. The acquisition makes the story look inevitable in retrospect, but inevitability is a luxury available only after survival.

He said they got lucky, then explained what founder luck looks like -knocking on a thousand doors until one opens. He also learned that venture capital was not the only way to stay alive. Money bought time and choice. That lesson would shape Commerce.AI and its decision to raise from the crowd.

Commerce.AI began with an unusually favorable start. Andy's work with Walmart created access to a customer, data, users, and a real problem. The company initially helped businesses understand messy, unstructured product and customer data.

The major pivot came during the pandemic. Commerce.AI had recognizable brands and interesting projects, but the work lacked stickiness. Insight sat inside innovation or marketing budgets. Customers could admire it without depending on it.

Contact centers were different. The pain was operational. Better intelligence could improve service, reduce cost, support an agent, complete a task, or trigger a workflow. The value was measurable and recurring.

This was not merely a product pivot. It was a move from interesting intelligence to economically necessary intelligence.

Andy said the decision faced resistance because the existing path was not visibly broken. That is often when a pivot is hardest. Founders must interrupt modest success before it becomes permanent limitation.

He says Commerce.AI has been profitable for two years. He also says revenue is several times larger than when the team itself was several times larger. That detail complicates his later argument about hard work. Progress came not only from adding hours, but from choosing a more valuable market and giving each person more leverage.

Andy described a market crowded with voice-agent companies and competitors that have raised hundreds of millions of dollars. Commerce.AI's counter-position is capital efficiency and lower inference cost. Its Token Trim technology, he said, can reduce token usage significantly.

That is a meaningful wedge in a price-sensitive market where software is compared directly with human labor. But I am not convinced cost alone can be the enduring differentiator.

Model prices decline. Vendors discount. Open models improve. A cost advantage built mainly on today's token economics can narrow quickly.

The more durable part of Andy's answer appeared earlier. Regulated enterprises need explainability, traceability, security, guardrails, and humans in the loop. A contact-center agent becomes valuable when it can act safely, know when to escalate, and leave an audit trail.

That may be Commerce.AI's stronger position. Not simply cheaper AI, but governed orchestration for high-volume customer operations.

Andy's thinking on models has also evolved. Earlier, he advised startups to use existing large models rather than spend extraordinary sums building their own. Today, he sees a role for smaller localized models alongside commercial LLMs and older machine-learning systems.

I agree with the coexistence thesis. Most enterprises will not train a frontier model. The practical architecture is a routed system -a large model for broad reasoning, a smaller model for narrow or price-sensitive tasks, and retrieval, rules, or human review where certainty matters more than fluency.

The real product is not any one model. It is the judgment layer that decides which tool handles which task, at what cost, and with what confidence.

Before the pandemic, Commerce.AI raised roughly $4 million, mostly through crowdfunded rounds. Andy said around 3,000 people invested, including retirees, teachers, firefighters, and doctors. For an enterprise software company, this was unconventional. That was the point.

Most people encounter the AI boom as customers, employees, or workers whose roles may change. Very few get access to ownership before a company becomes public. Andy tried to open that door earlier.

Crowdfunding does not remove startup risk. Retail investors still face illiquidity, limited control, and a long wait that may end with no return. Access alone is not democratization unless communication is equally serious.

Andy understands the burden. Three thousand investors can mean hundreds of emails asking what happened to their money. He said the reporting discipline feels like practice for becoming a public company.

There is a larger symmetry here. If AI concentrates gains among founders, employees, and venture funds while distributing job disruption across society, wider ownership is at least an attempt to broaden participation. It is not a complete answer, but it is more concrete than most statements about inclusive innovation.

The hardest part of our conversation came when I asked what responsibility AI founders have toward workers displaced by the products they build.

Andy was candid.

He argued that each company is optimizing its own world, that the technology is moving quickly, and that if one company does not build it, another will. He also suggested that economics will push products toward social good because technology that harms society will not have economic value.

This is where I disagree.

Markets price value for the customer who pays. They are less reliable at pricing costs carried by everyone else. A contact-center system can create economic value for an employer while eliminating jobs that supported families and communities. Both can be true at once.

The familiar line that AI is "giving time back to employees" often avoids the issue. If the automated task was the employee's job, the company is not returning time. It is removing labor. The worker may eventually move to something more valuable, but that transition is not automatic or guaranteed.

Andy offered a useful example from a business-process outsourcing company in South America. Rising minimum wages made the company less competitive, and AI became a way to remain in business. That is the knot in full. Policy can protect wages and still accelerate automation. Companies can adopt AI to survive and still reduce employment.

No founder can resolve the system alone. But founders can measure labor impact, distinguish augmentation from replacement, avoid euphemisms, and participate seriously in policy discussions. Governments will soon have to confront retraining, wage support, and the stability of communities built around routine knowledge work.

This is not a distant philosophical problem. It is arriving one contact center at a time.

When I asked what value from India most shaped his leadership, Andy answered hard work. He still works late with Europe, then wakes early for South America and the East Coast. He sees hours as the counterweight to the limits of individual intelligence.

I admire the stamina. I do not agree that more hours are the only alternative.

Hard work matters. So do judgment, leverage, recovery, and the ability to stop doing work that no longer creates value. Founders can turn their own tolerance for exhaustion into an organizational virtue and make burnout the admission price for belonging.

Andy's own story offers a more persuasive lesson. Commerce.AI grew revenue with a much smaller team and became profitable after a difficult market pivot. Those outcomes came from smarter architecture and sharper choices, not simply from extending the day.

The best version of hard work is not endless work. It is sustained commitment applied to the right problem.

Near the end, Andy said he now feels more foreign in India than in America.

His friends in India see him as more methodical, scheduled, and less spontaneous. He has adopted American time discipline. He jokes about switching accents. Yet the thing from India he will never let go of is nature, specifically the forested part of Mumbai where he grew up. Even the home he chose in America reflects that memory.

Living between two worlds rarely means preserving a culture intact. We keep fragments: a landscape, a food, a work ethic, a relationship with money. The rest changes so gradually that we notice it only when we return and discover that home now recognizes us as partly foreign.

Andy carries Indian endurance and American scale. He carries the financial caution of someone who once had to help support a household, and the ambition of someone who sees money as freedom to build. He has spent his career making systems more intelligent, but the most human parts of his story remain simple -a father whose circumstances forced responsibility early, a boy who learned by taking things apart, and an immigrant who found his people in a room full of builders.

Clear lessons readers can carry into their own works

Andy did not become a builder in America. America gave him an ecosystem where being a builder was immediately legible. Sometimes the right move is not reinvention. It is relocation to a place that understands the value already inside you.

Commerce.AI moved from insight projects to contact-center workflows because operational pain creates repeatability. An admired product is not the same as a necessary product.

Large models, localized models, older machine learning, rules, retrieval, and human review will coexist. The edge is knowing how to orchestrate them safely and cheaply.

Andy's stamina is real. So is the lesson from his smaller, more productive team. The goal is not to make exhaustion sacred. It is to direct effort toward the highest-value constraint.

Crowdfunding gave ordinary investors access to an AI company's potential value. That is meaningful. It does not erase the displacement question. The industry needs honest language about jobs and serious policy engagement before the social cost arrives at full scale.

The full conversation with Andy Pandharikar, including the near-death stretch at Fitiquette, the pivot that made Commerce.AI profitable, his case for hybrid AI models, and the unresolved question of worker displacement, is on YouTube now.

If you resonate, subscribe there, and forward this to one builder living between two worlds. That is how The Second Whistle gets heard.

Long-form conversations about what people built, and what built them