Shubham Trivedi
Dr. Prabhat Bhuddha Dev
The Ladder and the Trap: AI's Promise and Peril for Indian MSMEs
AI is no longer a distant dream . It’s a lived reality, and I’d be lying (and a little foolish) to pretend I didn’t use it to help write this very article. Companies even tangentially connected to AI have seen their valuations multiply, countries with strategic positions in the semiconductor supply chain have gained significant economic influence, while others are racing to host the data centers instead. Whatever else is true, AI is reshaping the global value chain.
Globally, MSMEs represent roughly 90% of all businesses and contribute more than half of total employment. In India’s case, where manufacturing is increasingly being positioned for global supply-chain integration, the MSME sector’s ability to adopt new technology isn’t just a growth story , it’s a national priority. After all, it’s the second-largest employer after agriculture. In emerging markets, seven out of every ten formal jobs trace back to a small or medium enterprise.
It’s easy to talk about AI in the abstract: large language models, automation, intelligence at scale. But for most Indian MSMEs, the reality is more grounded , a textile unit run by two brothers in Surat, a precision-parts manufacturer outside Pune managed by a husband-and-wife team, a logistics operator in Ludhiana doing everything from procurement to billing himself. This year’s global theme, “Human-Centered Entrepreneurship in an AI-Driven Future,” couldn’t be more fitting. As AI reshapes how we produce, trade, and work, the question isn’t whether small businesses can keep up . Rather, it is how we make sure technology serves them, instead of leaving them behind. For these businesses, AI isn’t a boardroom strategy. It’s a potential lifeline.
AI, Up Close: Three Founders, Three Lessons
We talked to three MSME owners in the Karur textile cluster about how they are actually using AI day to day, and what came back was less “AI will change everything” and more a set of specific lessons about how to do this well , and how not to.
Mr. X, ABC International runs a small AI toolkit rather than a single assistant: ChatGPT on his phone for quick technical questions, Copilot for supplier and buyer correspondence, and premium versions of Gemini and Claude for market research and supply chain planning. AI now drafts his emails, helps lay out the factory floor, tracks machinery wear, and builds his team’s production-tracking templates.
However, the approach has limits. Mr X estimates roughly 90% of what he gets back is accurate, which means 1 in 10 isn’t. Asked to source Indian suppliers for specialty nylon, AI found two genuinely usable leads out of three. Sourcing manpower has been worse, often surfacing contacts that are outdated or unresponsive. The lesson isn’t to stop using AI for sourcing , it’s to never treat its output as a final answer rather than a strong first draft.
What works: matching the tool to the task instead of forcing one platform to do everything. Mr X’s deliberate split, a different assistant for technical queries, communication, research, and logistics is part of the reason his accuracy rate holds up as well as it does.
Mr. Y, Managing Director of DEF Textiles, is minimal and deliberate by comparison , he barely touches AI for routine tasks like email. His focus is AI-based visual inspection for fabric and finished-product defects, trained on a database of defect images.
What works: digitise before you automate. Mr Y is explicit that this only works once production records are fully digitised first , automation follows digitisation, it doesn’t substitute for it. Businesses that skip this step usually end up automating chaos, not fixing it.
Mr. Z, Managing Director of The GHI Impex, uses AI mainly as a creative and relationship tool , designing presentations, generating product concepts, and drafting correspondence that keeps overseas buyer relationships strong.
What works: don’t take the hardest step alone. To move AI from the design stage into the factory floor, he’s working with a local tech partner, Objectways, rather than attempting that leap in-house. And on the jobs question, his answer is the sharpest of the three: the real safeguard against AI-driven displacement isn’t policy alone , it’s upskilling. Workers stay irreplaceable for as long as they keep upgrading alongside the tools they use.
Three founders, three tools, three lessons but with the same underlying instinct: AI earns its place by removing friction, not by replacing judgment.
The View From the Factory Floor
Not every reaction to AI in these clusters comes from the boardroom, and it’s worth sitting with the one that doesn’t. Wearable tech that tracks workers’ movements and patterns , introduced for tracking and training models and robots has started showing up on cluster floors. Alongside it has come a harder, more personal anxiety: the sense that the data collected from a worker’s own body and routine could end up training the very system that eventually replaces them. Some have put it starkly, that workers are, in effect, digging their own graves.
That concern deserves a real answer, not a reassurance. MSME owners are genuinely optimistic that AI won’t cost jobs, and history offers some support for that. Computers, ATMs, and barcode scanners all triggered similar fears that didn’t fully play out. But that comparison has a limit: in those earlier shifts, workers weren’t usually the ones directly supplying the training data used to automate their own tasks. That distinction is exactly why this fear feels more personal, and dismissing it as “the same old automation panic” misses what’s actually different this time.
AI not costing jobs in a cluster like Karur’s isn’t an inevitability , it’s a choice the cluster’s leadership makes, in how transparently it treats the people whose movements have quietly become data points. That choice is the actual test of whether “human-centered entrepreneurship” is a theme for a banner or a practice on the floor.
Conclusion: Optimism With Eyes Open
India’s MSMEs are entering the AI era with something many advanced economies have already lost genuine enthusiasm, unencumbered by cynicism. For owners like Baskar, Perumal, and Madhan Kumar, that enthusiasm is earned: AI is one of the few technologies in recent memory that doesn’t demand large capital to access. It can be seen as our “Latecomer Advantage” , picked up on a phone, on a free tier, and it compounds in value the more a business learns to use it well. That’s a genuine ladder, and it would be a mistake to undersell it.
But a ladder for the owner isn’t automatically a ladder for everyone the owner employs. The same ease of access that lets a founder experiment with a new AI tool on his own initiative doesn’t extend, in the same way, to the worker on the line whose movements that same AI may be trained on. He or she may not have the device, the data plan, the spare hour after a shift, or most importantly, a say in how decisions about their own job get made. The optimism this piece has documented is real, but so far it belongs almost entirely to the people with the capital, education, and authority to choose how AI gets used. The people whose routines are quietly becoming its raw material mostly don’t get a choice at all.
That asymmetry is exactly where “human-centered entrepreneurship” stops being a slogan and starts being a test of what a business is actually willing to do. A responsible MSME and the larger companies, banks, and platforms it works with , owes its workforce more than reassurance. At minimum, that means:
- Extending AI literacy and tool access beyond the owner’s desk, rather than assuming it stops there by default.
- Treating any efficiency or productivity gain that comes from worker data as something to share back , through real wage and skill progression, rather than treating the gain as a default windfall.
- Bringing workers into the transition plan before new tracking or monitoring tech arrives on the floor, not after.
- Being plain about what’s being tracked, why, and where that data goes , because vague reassurance is exactly what turns caution into the fear that their own labour is training its replacement.
These aren’t philanthropic asks , they’re conditions for sustainable growth. The care that goes into verifying a supplier lead, sequencing automation correctly, or documenting an AI-generated design for IP purposes has to extend to the people standing on the floor while all of that gets built around them. The tools are here. The opportunity is real, for the owner who picks them up, and for the country, if it chooses to make that opportunity wider than the boardroom. Whether AI ends up being the ladder this sector climbs together, or the trap that only some of its members escape, depends entirely on whether that choice gets made on purpose.
Changes are all incorporated , the opening is toned down, geopolitical references removed, the “pitfall/right way” labels made consistent across all three founders, the second “digging their own graves” replaced, and the conclusion reframed to be constructive rather than accusatory.
“NOTE: The views expressed here are those of the authors and do not necessarily represent or reflect the views of CRB.”
