Start With Two or Three: A Practical Reading of an AI Report for Small Manufacturers

The EY-FTCCI Report: Scope and Context

EY and the Federation of Telangana Chambers of Commerce and Industry, FTCCI, produced a report titled "AI-force Multiplier" for its GROWTH[X] 2026 event. The document did not state a publication date, but carries a 2026 copyright notice and cites sources up to February 2026. It focuses on Telangana, and its findings should not be assumed to apply unchanged elsewhere. The report said MSMEs contribute about 30% of India's GDP and, in one place, about 36% of manufacturing output, while the introduction gave "over 45%"; the report is internally inconsistent on the latter, and we do not choose between them.

The use cases it lists for manufacturing

• Predictive maintenance: analysing vibration, temperature, pressure and power data to catch wear before breakdowns. • Computer-vision quality inspection: detecting surface defects, dimensional errors and assembly faults in real time, with root-cause analysis. • Intelligent production planning: demand-driven, constraint-based scheduling that reschedules around disruptions. • Supply-chain and inventory optimisation: demand forecasting, inventory balancing and supplier performance analytics. • Energy optimisation: machine-level monitoring, wastage detection and peak-demand management. • Workforce safety: computer vision for protective-equipment detection and unsafe-behaviour alerts. • AI-assisted design and virtual prototyping: fewer physical prototypes and faster time to market. • For traditional MSMEs: AI-enabled accounting and cash-flow forecasting, automated procurement, seasonal demand prediction and digital twins for small production lines.

“AI will not replace jobs but will augment the workforce.”

Barriers and recommendations

The report named high implementation cost, a lack of skilled talent, limited digital infrastructure, low awareness, privacy concerns, fragmented data and insufficient ecosystem support. It also noted thin margins, equipment downtime and inconsistent quality as operating issues. Its recommendations included starting small with a readiness diagnostic and piloting two or three high-impact use cases, setting minimum AI governance early, scaling through sector playbooks and shared cluster infrastructure, building data-protection controls from the start, strengthening cybersecurity and investing in workforce upskilling. It framed AI as augmenting, not replacing, staff, quoting FTCCI's Bala Prasad Peddigari: "AI will not replace jobs but will augment the workforce". The report proposed a phased roadmap: readiness and quick wins in 0 to 6 months, scaling proven use cases in 6 to 18 months, integration for customer-centric growth in 18 to 36 months, guardrailed semi-autonomy in 3 to 5 years and convergence with other sectors in 5 to 7 years.

How to choose: our framework

Rank each use case on three questions. How much does the problem cost you each month? How easily can you get the data? How quickly could you see a result? For many small factories, a plausible first pick is whichever problem is already costing the most and where data exist in some form. • If breakdowns hurt, start with simple condition monitoring on the one or two machines that stop production most. • If rejection is high, photograph defects consistently and check whether a vision tool can identify them. Even before software, a standard photo routine often improves quality. • If the power bill is large, begin with sub-metering your biggest machines and look for waste at idle times. • If delivery dates slip, map the bottleneck and test a simple scheduling tool before an advanced one.

An invented example

A small foundry loses several hours a month to unplanned furnace stops. It records stop times for two months, installs low-cost temperature and vibration sensors on the worst-affected equipment and reviews alerts weekly. Even if the first alerts are imperfect, the log shows patterns that the maintenance lead had not seen. The example and numbers are invented to show the sequence: measure, sense, review.

Cautions

The report is from a consultancy and a chamber, both of which have an interest in promoting adoption. It did not provide evidence of results from small firms in our excerpt, and we have not seen cost figures. We cannot say what payback to expect. Ask any vendor for references from firms of your size and a clear statement of what data and staff time you must supply.

Shared infrastructure

The recommendation to share resources across a cluster deserves attention. Common inspection centres or maintenance services can bring tools within reach of firms that cannot justify them alone. Industry associations can negotiate with vendors on behalf of members. Editorial analysis based on the EY-FTCCI report "AI-force Multiplier" (2026, undated). Our framework and example are labelled as such.