Kshitij Thakur, CEO & Co- Founder, Agrograde explains how AI and Intelligent Automation will break India’s post-harvest bottlenecks in an exclusive interview with AgroSpectrum. He also argues that post-harvest management will Define India’s next Agri-tech revolution while emphasising the role of intelligent packhouses in the fresh produce economy in India. Edited Excerpts:

Why is post-harvest management emerging as the next big growth opportunity for Indian agriculture?

The post-harvest supply chain in fresh produce is completely broken with no critical infrastructure and extremely low transparency. The trust deficit is very high in these supply chains. We lose around 40 per cent of the produce grown globally, and most of that isn’t happening on the farm, but it’s happening in grading, sorting, handling, transport, and storage, the stages nobody really talks about. That’s exactly why post-harvest is the next frontier. We’ve squeezed most of the easy gains out of yield improvement, but the bigger, faster win now is in how we handle what we’ve already grown. In crops like onions, for instance, a significant share of storage losses is often decided even before the produce enters storage. If damaged or diseased produce is not identified and removed early, or if produce isn’t graded and sorted properly, it can increase spoilage. It shows that better quality control before storage can have a much bigger impact than many people realise.

Where are the biggest inefficiencies in the potato and onion value chain specifically?

Potatoes and onions are unique because they are bulk, high-volume, low-margin crops handled almost entirely manually at the aggregation and packhouse level. Some of the biggest inefficiencies in these crops are the current subjective quality standards and the manual grading and sorting practices. Subjective quality standards lead to a trust deficit in the supply chain. Quality-driven trust gets replaced by person-driven trust, and hence we see a lot of middlemen in these supply chains. This single bottleneck results in losses everywhere. Price discovery, marketability, market linkage, and post-harvest losses all are dependent on quality assessment, grading, and sorting. For example, ungraded onions will sell at Rs 1-2/kg less than a graded lot and will also have higher storage losses. A buyer from Delhi will never be able to purchase from a farmer or trader based in Nashik. Almost 20 per cent of the consignments face quality-related disputes. On a storage level, onions see almost 35-40 per cent losses.

The technologies that work in developed supply chains abroad fail in Indian conditions and were not designed for Indian varieties; hence, these supply chains never got any technological solutions that could solve the age-old problems of losses and quality consistency.

That’s exactly where our solutions at Agrograde are focused. Whether it’s the Vector Series for potatoes or the Grade Plus Series for onions, we’ve designed grading solutions specifically for Indian produce and operating conditions, helping packhouses and traders improve grading consistency, minimise wastage, and ensure better-quality produce moves further down the supply chain.

What’s the economic impact of poor post-harvest handling across the value chain?

It affects everyone down the line. Farmers lose the most directly, as a poorly graded, poorly stored batch means lower realisation at the very first point of sale. Traders and packhouses often absorb rejection costs and rework. Processors get inconsistent raw material, which affects yield and product quality downstream. Exporters probably feel it hardest of all, because export markets are unforgiving on quality and consistency. Improper grading sorting leads to unfair price realisation and repetitive grading sorting processes. It also leads to high storage losses. On an average a farmer gets 10 per cent less price for his/her crop because of this.

What are the limits of manual grading, and how does AI actually improve it?

Manual grading is based on subjective quality assessment. Grade A for someone might not mean the same for someone else. There’s also a high chance of letting rotten produce get packed and dispatched along with good quality produce, which results in high transit losses. Manual processes are inconsistent and not scalable. The average age of manpower working in these activities has crossed 45 years in certain supply chains with no new workforce joining. This has now led to existential level challenges in the supply chains. Right now, farm labour wages are barely at or even below minimum standards. The current economics just don’t support higher wages in manual farm tasks. That is why automating critical activities is not just a good to have but a must-have now.

At Agrograde, we’ve built our AI grading systems specifically for Indian produce and operating conditions, helping businesses make faster, more reliable grading decisions while reducing dependence on manual inspection and improving quality consistency across every batch.

Can AI-generated quality data become central to traceability, pricing, and supply chain transparency?

It’s already becoming necessary. Once you’re grading with a machine instead of a human eye, you’re also generating a data trail for every batch: size distribution, defect rates, and so on. That data is what lets a trader price a lot fairly instead of eyeballing it and lets an exporter certify quality to a buyer in Dubai without a dispute later. As India’s agricultural exports continue to grow, buyers are increasingly looking for products with consistent quality backed by proper documentation. This means machine-based quality checks are no longer just an added advantage, but rather they’re becoming essential for doing business.

Why has automation adoption in post-harvest agriculture lagged manufacturing?

In agriculture, most of the post-harvest practices happen at places that have a lot of dust and challenging environments. Moreover, the variation in shape, size, varieties, and crops makes the activities difficult to be automated. The practices are decentralised, volumes are low, the unit economics are challenging, and the operating conditions are harsh in agriculture. Technologies that work in clean, closed environments would fail in open-air conditions while handling agricultural produce. Any kind of automation in agriculture has to be designed considering the high variations and challenging conditions. Hence, machines that work in foreign supply chains fail miserably in Indian packhouses. Those were not designed for India.

The costs of these machines is also another barrier. These machines were designed for enterprises handling 100MT per hour and were extremely expensive. The scale and the costs do not work for India. For example, the potatoes in India are not washed before storing or packing; this makes any existing optical sorter fail. Indian onions have a loose skin as compared to onions we see in Europe or other countries; hence, the technologies that worked there end up damaging the onions.

We had to study the physical characteristics of the crops to design suitable solutions that also justify the ROI. With technological advancements and the quality of talent India has, we now have technologies at par with global manufacturers and some even better than them.

Is there a mindset shift among packhouses, exporters, processors, and cold storage operators toward automation—and what’s driving it?

Definitely, and it’s been quite visible over the last two to three years. A lot of it is being forced by economics rather than choice. Labor is harder to find at harvest peak, and wages have been rising steadily, so a machine that runs a consistent shift without fatigue starts looking like a necessity rather than a luxury. Export buyers are also a big push factor: once a buyer in the Gulf or Southeast Asia asks for size-graded, defect-free batches with documentation, a packhouse either invests in consistent grading or loses that order to someone who has. Government support has helped too, with schemes like PMKSY and the Agriculture Infrastructure Fund having sanctioned close to Rs 67,717 crore in loans for exactly this kind of post-harvest infrastructure, which brings the capital cost within reach for mid-sized operators, not just the big players. Once businesses see a neighbouring packhouse getting better prices and more export orders through automation, adoption tends to spread quickly. That’s the momentum we’re seeing at Agrograde as AI-based grading sorting moves from being an innovation to becoming a business necessity for post-harvest operations. 

Which customer segment is adopting AI-enabled grading and sorting fastest, and why?

We are seeing adoption at aggregation points of fresh produce, which includes packhouses, traders, exporters, cold storage operators, and food processors. We are seeing strong interest from the traders, FPOs, and cold storages that supply quick commerce, modern trade, and exports, as quality is extremely critical. For example, Agrograde’s Vector series turns grading and sorting of potatoes autonomous by identifying external defects using AI accurately at speeds as high as 20 mt/hr.

How do you see intelligent packhouses evolving over the next five years?

Packhouses would follow the same evolution as modern warehouses. Current warehouses are now almost completely autonomous and smart, enabling them to receive, check, store, track, and dispatch at amazing speed. Critical activities are now getting automated, and eventually the entire packhouse will be optimised and designed to run autonomously to handle maximum volumes, eliminate handling losses, and make the process trackable and transparent. Right from inward, to vision- and sensor-based quality assessment, cleaning/washing, grading, sorting, weighing, packing, to dispatch, everything will be one connected system. Imagine packhouses turning into mini mandis, where the primary processing and transparent price discovery, as well as market linkage, are done via technology.

At Agrograde, that’s the vision we’re building towards helping packhouses evolve into connected, data-driven hubs where every batch generates insights that improve quality, storage, pricing, and traceability across the supply chain.

Beyond grading and sorting, which parts of the supply chain are ripe for AI-led transformation?

It’s not just about deploying software but building essential hardware and making machines smarter using AI. I think technologies that enable market linkages near the farm gate or at collection centres with technology-enabled price discovery would create a lot of value in the supply chain. The current supply chain is quite opaque; anything that removes friction and guesswork from this using technology would be beneficial. Farm activities, too, are quite operationally heavy and time critical. During peak seasons, skilled manpower is scarce. Automating activities like yield estimation, harvesting, plucking, and de-weeding need to be automated.

Dipti Barve

dipti.barve@mmactiv.com