AI in Manufacturing: Smarter Factories, Faster Decisions, Real Results

Episode 2 of India Automation Hub’s Engineering Meets AI series — the short read for those who missed it live.

We sat down with Mr. Anil Sagar, founder of Renata AI, for a candid, hype-free look at where AI actually earns its keep on the factory floor. An IIT Delhi engineer and IIM Calcutta MBA with two decades in telecom (Nokia, Ericsson, Bharti) before industrial AI, Sagar worked across automobile, pharma, metals and more — and across plants of every size.

His Core message: 

AI on the shop floor is no longer a science project. The hard part isn’t the model — it’s data discipline, deployment sequence, and the culture to make it stick.

The shift

AI in inspection isn’t new — what changed is the economics. Cheaper compute and more powerful deep learning have moved the advantage from costly, rule-bound hardware to flexible software on simpler cameras. Solutions are getting cheaper and more capable at once. Adoption in India is top-down: large OEMs led and pulled suppliers along; SMEs lag on ROI.

Where AI is landing

• Vision AI — defects and beyond. One pre-dispatch line replaced 18 inspectors across three shifts with cameras and a robotic arm — ROI inside a year, with those workers moved to higher-value roles. The bigger prize: existing CCTV, repurposed for counting, productivity, SOP and safety compliance, and fire/leak/theft detection.
• Traceability. Fast becoming table stakes for recalls and compliance — operation sequencing, product data tied to QR/barcode with timestamps, integration across dozens of machines, rolled out tier by tier.
• Safety. Mostly a behaviour problem against known norms — vision flags missing PPE, restricted-zone intrusion and early smoke, while IoT sensors make it predictive.
• LLMs / knowledge management. A centralised, version-controlled repository plus an on-premise LLM becomes a “digital twin” of company knowledge: design reuse, voice troubleshooting, sales and training — running locally, never calling the internet, already live in Hindi.

What executives asked

• Security: Manufacturers default to on-premise; proprietary data stays in-house, no compromise on capability.
• Existing CCTV: Usually kept — if the camera sees it, the feed routes to a new PC for AI. No rip-and-replace.
• Jobs: Augmentation more than replacement. You can’t fight the technology — upskill and move with it.

Takeaways

• The barrier is data and culture, not the model.  Digitalise first; chase easy wins to build belief.
• Software is the differentiator now — accuracy up, cost down.
• Your cameras are an underused AI asset.
• Traceability is becoming non-negotiable — and a source of brand value.
• On-premise is the default, no longer a trade-off.
• Treat AI as augmentation, and start now.

Thanks to the speaker and  to everyone who joined and asked sharp questions. The full recording will be available soon . 

Our next Webinar is on the way — watch this space. See you on the next one.