The Test of AI in India Is Whether a Small Workshop Can Use It

The ORF Perspective on AI Diffusion

Soumya Bhowmick, a fellow at the Centre for New Economic Diplomacy at the Observer Research Foundation, published an Expert Speak article on 12 February 2026 on democratising AI for MSMEs. It was timed ahead of the AI Impact Summit scheduled for 16 to 20 February 2026 in New Delhi. It is a policy argument, and we separate its claims from its cited facts.

The facts it cites

The author cited official estimates that MSMEs contribute about 30.1% of GDP and 45.73% of exports, and said the formalisation drive had expanded the recorded base to roughly 65 million units. These are the author's citations; we have not verified them, and the number of recorded units differs from the 7.83 crore Udyam registrations reported to Parliament in March 2026, probably because the two measures count different things.

“AI diffusion stalls if tools remain English-first and metro-first.”

The argument

• The real question is "no longer whether MSMEs should adopt AI", but whether AI can be made accessible in ways that raise productivity and advance inclusion. • India "won't be judged by how fast it builds frontier AI", the author wrote; success should be measured by diffusion. • Adoption is constrained more by practical barriers than by awareness: limited knowledge of suitable tools, affordability and budget. Policy should focus on market-shaping, lowering first-mover risk and making trustworthy tools easier to find. • Plug-and-play tools embedded in existing workflows, such as customer support, demand forecasting, inventory optimisation and computer-vision quality checks, are the most realistic route for most MSMEs. • "Frugal AI", meaning smaller, application-specific models that work within limits on cost, compute, power and connectivity, is presented as a design principle, which the author linked to the Economic Survey 2025-26. • Public digital infrastructure helps: the IndiaAI Mission's shared compute and BHASHINI's language layer, which enable Indian-language interfaces. • "AI diffusion stalls if tools remain English-first and metro-first." • Inclusion is a test. If AI concentrates among firms that already have capital and managerial capacity, productivity and market-access gaps will widen.

A critical reading: our analysis

The essay's strength is its focus on the user. A small firm does not buy "AI"; it buys a solution to a problem such as late deliveries, missed leads or wasted stock. The emphasis on tools that fit into existing workflows is realistic, because most small owners have little time to change how they work. There are limits. The essay is an argument, not a study of outcomes. It does not offer evidence on how much productivity gain small firms achieve from the tools it lists, nor on the cost. It does not say who should pay for the market-shaping it recommends. It mentions language infrastructure, but we cannot judge from it how good the available tools are in practice.

What an owner can take from it

1. Start with a problem you can measure. Which of your costs or delays is largest? Ask whether a tool could reduce it. 2. Prefer tools that fit your current process. If a tool requires you to rebuild how you work, the adoption cost will probably be higher than the benefit. 3. Test in the language your staff use. A tool that your team cannot comfortably use will sit idle. 4. Run a small pilot and keep a control. Try the tool on one product line or one team for four to eight weeks, and compare with the rest. 5. Count the whole cost. Include subscription fees, training time, set-up and the cost of checking outputs. 6. Keep a human in the loop. For quality, safety or legal decisions, a person should review what the tool suggests.

Inclusion in practice

The author's warning that gains may concentrate in already-capable firms applies to any technology. Local associations and cluster bodies can help by sharing tools, running joint trials and negotiating group prices. Shared infrastructure can lower the entry cost for firms too small to buy alone.

Women-led and rural firms

The same essay noted that women-owned units are 20.5% of Udyam registrations but a smaller share of employment and investment (see our People & Culture piece PC3). Tools that work in local languages and on a basic phone are likely to matter most to firms with fewer resources.

A caution on hype

An essay timed to a summit will emphasise opportunity. A careful owner will balance it by asking for demonstrations, references from similar firms and a clear exit option. Editorial analysis based on ORF (12 Feb 2026). Our analysis and guidance are labelled as such.