How ChatGPT Became My Best Sales Channel

I’m Sudharsan, a product engineer in Bengaluru. I build FactoStack (opens in a new tab) alongside my full-time work on streaming products.

FactoStack started with a pattern I kept seeing across the Indian manufacturing belt: factories running important workflows through WhatsApp groups and Excel sheets. Not because the teams were unsophisticated, but because the available ERP choices were a poor fit. Enterprise tools were expensive and slow to implement; many affordable tools stopped at accounting.

There was room for software that could track production, inventory, and compliance without requiring a consultant. That gap felt real enough to build for.

The quiet launch

The first two weeks after launch were quiet. I tried cold calls, backlinks, and email outreach to manufacturers. Very little came back.

Then demo requests began arriving from an unexpected place. When I asked people how they had found FactoStack, several said ChatGPT. They had asked a specific question—something like “ERP for a small Indian manufacturer”—and FactoStack appeared in the answer.

I had not designed the launch around AI search. The product was simply specific enough to be a useful answer.

A July 2026 snapshot

As of July 2026, FactoStack has its first paying customers and a small but meaningful stream of conversations with manufacturers. I’m intentionally avoiding sharing identifiable customer details or private pipeline numbers. The signal that matters is simpler: people with real operational problems are finding the product, evaluating it, and paying for it.

One early customer came through my network. Another found FactoStack through ChatGPT after evaluating several systems. The second described the product as a strong—but not perfect—fit and joined as a design partner.

That changed how I think about product-market fit. A useful product in the hands of a committed customer can teach you more than a theoretically perfect feature list.

What has worked

AI search rewards specificity

AI search has produced some of the highest-intent conversations so far. My working theory is that language models are good at surfacing niche tools when the user’s question contains clear constraints. “Manufacturing ERP for an Indian MSME” is a much sharper problem than “business software.”

That does not make AI search a magic channel. It means the usual fundamentals—clear positioning, useful pages, consistent terminology, and credible technical detail—now help machines as well as people understand where a product fits.

Relationships still matter

Manufacturing is a relationship business. Conversations through local industry networks move slowly, but the trust they create is durable. AI discovery can start a conversation; it cannot replace implementation confidence or domain understanding.

Design partners beat passive signups

I chose not to optimise for a large pool of free accounts. A customer actively shaping a workflow creates a tighter feedback loop, clearer priorities, and real accountability. At this stage, depth is more useful than vanity volume.

What has not worked

Cold calling has been difficult, and paid search has not yet produced the same quality of intent. Both may improve with practice and better targeting, but they have not been the early engine.

The larger challenge is pacing. Manufacturing sales cycles are long and relationship-driven. Building the product alongside a full-time job makes focus more valuable than channel breadth.

The lesson

The best early distribution insight was not a tactic. It was that a sharply defined product creates its own surface area for discovery.

FactoStack is not “software for every business.” It is software for manufacturers with particular operational constraints. That specificity makes the product easier for a customer to recognise, easier for an industry contact to explain, and—unexpectedly—easier for an AI system to recommend.