Why we're betting Puro AI Labs on the Indian factory floor

We just took on our first manufacturing client.

A yarn-spinning mill: five factories across India, a multi-generational family business, real operating scale. We're building two things for them at once: a website that finally reads as the institution the operation actually is, and a custom ERP with AI simulations and forecasting baked into the core, not bolted on as a separate analytics tool.

The website is live. The ERP is in flight. It's the first of many engagements we want to take in this category, and it's worth saying out loud why.

The most software-starved market in India is the factory floor

A lot of the AI conversation in India right now is about the obvious markets: direct-to-consumer brands, agencies, services businesses, content creators. Fair. Those are the easiest to reach and the easiest to demo for. But they're also the markets that already have ten flavors of software to pick from. The marginal value of better tooling there is real, but it's not transformative.

The factory floor is a different story.

India has thousands of family-run manufacturing businesses (yarn, fabric, food processing, chemicals, auto parts, packaging, agri) that have compounded value over decades. Two, three, sometimes four generations of operating knowledge. Real businesses with real customers, real margins, and real complexity. The kind of operation where the founder's son and the floor supervisor have a working relationship measured in years, not months.

These businesses run, almost without exception, on a working set of spreadsheets, a half-deployed ERP nobody fully trusts, and a great deal of tribal knowledge. Not because the operators are unsophisticated. They're the opposite: they understand their operation in a way no piece of software has ever modeled correctly. The problem is that the software market has never built for them. Enterprise ERPs are designed around large-corporate Western operating models. SaaS is designed around the predictable workflows of services businesses. Almost nothing in either category fits how a multi-generational Indian factory actually runs.

So the operators do what rational operators do. They keep the spreadsheets, they trust the tribal knowledge, and they get on with the business.

Why ERP has been particularly bad for them

ERP is the worst category here. Off-the-shelf ERPs make three assumptions that don't hold for a family-run Indian manufacturer:

They assume the data model fits. It doesn't. Yarn counts, fabric grades, lot tracking across multiple factories, supplier agreements that include side deals, the way a particular customer's orders get prioritized when capacity tightens: none of it slots cleanly into a templated module.

They assume the workflow fits. It doesn't. Decisions span the floor and the front office in ways the system never sees. The supervisor knows that when supplier X delays a batch, you call Y first because of a relationship that goes back fifteen years. The system doesn't know that, and the operator stops trusting the system the first time it suggests calling Z.

They assume forecasting is a separate analytics product. It is, and that's the bug. The ops team doesn't want to leave their operational dashboard, open a separate BI tool, run a query, copy the numbers back. They want to ask "what if our biggest customer increases their order by 30% next quarter, can we hit it without buying more capacity?" inside the same system that's running the operation today.

What you end up with, in 95% of these companies, is a partial ERP rollout that the team ignores, plus the spreadsheets that actually run the business. Software that nobody wanted, paid for, and forgot.

What AI changes, and what it doesn't

The temptation is to say AI fixes ERP. It doesn't, by itself. A bad ERP with an AI chatbot bolted onto it is still a bad ERP.

What AI changes is the cost and shape of building an ERP that actually fits.

A custom ERP for a 5-factory yarn-spinning mill, built around their specific data model and their specific workflows, used to cost north of a crore and take two years. Today, with the right team, you can build it in months. That's the cost-curve change we wrote about last week. The new thing, the thing that turns a custom ERP into a real edge, is what you do when you bake AI into the core of the system instead of bolting it on as a feature.

For Surge Spinners, that means:

• The operations team can ask forecasting questions in plain English, inside the same screen that shows them today's production. "If our biggest customer increases their order 30% next quarter, where does the bottleneck land first: capacity, shift planning, or supplier lead time?" The system runs the simulation and shows the answer next to the dashboard, not in a separate tool.
• Scenario planning becomes a daily activity, not a quarterly exercise. The operations head can run "what if" queries across the five factories and get answers in seconds.
• Anomalies on the floor get flagged in language the supervisor uses, not in dashboard widgets nobody reads.

That's the thing that's hard for an off-the-shelf ERP to retrofit and easy for a custom system to build native.

Why this is where we want to spend the next five years

Puro AI Labs has done strong work in services so far: coaches, agencies, creators, surgeons, founders. We'll keep doing that. But the most interesting market we've seen, measured by size, by need, and by how poorly served it's been historically, is Indian family-run manufacturing.

The reason it's been underserved is structural. Enterprise software vendors don't sell to it because the average ticket size is too small. Dev shops don't sell to it because the relationships, the trust required, and the operating-knowledge depth needed are too high for a typical engagement. So the market has just sat there, run by spreadsheets, growing on operating skill alone.

That's exactly the kind of mismatch we like.

If you're a family business sitting on real operating complexity and you're still running on spreadsheets, if you've been told your business is "too custom" for software, or "too small" for serious ERP, or your last ERP rollout was a write-off, let's talk.

It's the kind of work that quietly compounds, for the operators we build it for, and for us as a studio. We're a few months in, and we already think it's the most important category we'll work in this decade.

Get in touch if you want to be one of the first.