The Power of AI in Supply Chains & Beyond
In this fascinating Q&A session, Virifides’ founder Steve Leng and Prasad Putta, the founder of Vassar Labs, discuss the technology that is, quite literally, writing all the headlines right now: Artificial Intelligence (AI).
As advancements in AI technology continue to gain the interest of businesses across more and more industries, new use cases for AI are increasingly being found in the supply chain, and all the way back to the farm.
Contents:
Harnessing the power of AI: The 90’s and present day
Can AI fix the changing climate? - Applications in the Produce Industry
A Breakthrough Technological Link for Supply Chains?
(Note: The discussion has been edited for length and clarity)
Harnessing the power of AI: The 90’s and present day
Steve: Can I start with perhaps a deceptively simple question. How would you define AI?
Prasad: Well, AI is just a manifestation of how a computer system can recognise pattens and retrieve intelligence from data, just like humans do. But AI can tackle certain types of problem a lot faster than humans can. The output enables you to positively impact business decision making.
Steve: That does sound like a simple definition! But one that makes a lot of practical sense.
So how would you say AI first came to your notice?
Prasad: I first got into the world of AI through from my research on RFID in the late 90s. At that time, the problem was that RFID would generate huge amounts of data, but it was almost impossible to make sense of it using only human logic. You needed an additional intelligence system to decode the information and extract the value.
Later during my 1st job at i2, I was using Genetic Algorithms to deploy scheduling systems for automotive manufacturing. To give an example, when you have to produce both white and red cars on the same line, the pipes used to colour the cars have to be cleaned or you won’t get pure white. Which means a changeover cost in the manufacturing process. At the time, we used a genetic algorithm to optimise the production schedule and minimise the overall cost.
Steve: I suspect this simple but instinctive use of rudimentary AI tools is light years from what is possible today. How would you say these algorithms compare to current AI capabilities?
Prasad: Nowadays, algorithmic optimisation is a relatively easy process. There has been a shift from typical programming technologies to learning models like ChatGPT, which people say can easily pass a law exam. The new models can answer questions in the same way that a subject matter expert would.
Businesses have been quick to pick up on this potential and I’d say we are already at a point where AI is driving many of the decision-making processes of companies across all industries.
Steve: Even from your early days at MIT, I remember you worked on integrating technologies together to build innovative solutions that delivered greater insights than any single technology. Was AI, in a way, the missing piece in the puzzle?
Prasad: I think it’s fair to say that the work we’re doing now at Vassar Labs brings together data from IoT sensors, satellite data, drone data, crowd sourced data into a unified data model and uses AI to drive easy decision making.
For example, the work we are doing in water management in India.
Reservoirs are used for both saving water and for flood control. Typically, in the catchment area of a large reservoir there will be a few thousand small tanks, whose combined storage capacity can be 30-40% of the reservoir. So, to predict the inflows into the reservoir it’s important to know the current storage in these tanks, as the inflow will be a direct function of the rainfall and how empty of full these tanks are. If it rains and the ponds are empty, the rain will fill the ponds first before going into the reservoir. Using AI and Satellite data we predict how much water is currently stored in these tanks and based on the rainfall and hydrology model, predict the net inflows in the reservoir.
Steve: There’s clearly application for AI in these kinds of use cases where physical distance and the difficulty of manually monitoring and analysing this kind of data would make it cost prohibitive to try and deliver the same results. Do you see the same potential for AI in broader sustainability-oriented use cases?
Prasad: Absolutely. To take another example. There is now a lot of interest in harnessing solar and wind energy to reduce carbon emissions. The problem though is that they are not reliable energy sources. However, using data derived from wind patterns, cloud coverage and other weather forecast-based data, you can predict the amount of power that can be produced by solar panels and wind turbines. Combining this with hydro and other energy sources, we can ensure that the grid is stable while also maximising green energy. Sorting through and recognising patterns using this data helps to solve problems that are impossible for human beings to solve by themselves.
Can AI fix the changing climate? - Applications in the Produce Industry
Steve: Philosophically speaking, it sounds like we are now abstracting data from inanimate or non-human entities, such as land, water, etc. Harnessing their ‘intelligence potential’ represents a step-change in what we have been able to do at any kind of scale before.
If I can turn to an industry that Virifides Digital works with, the Produce industry, do you see value in applying AI supported solutions to address the natural and environmental challenges the Produce industry faces?
Prasad: Absolutely. As you know, the outputs of the agriculture-based industries are driven by climate, and where the yield is often impacted availability and by pests. AI can help predict which pest may impact the crop based on the crop life cycle, observed weather, and from this advice the farmer on actions to be taken to protect their crop. Historically, this kind of knowledge was only held locally, in the heads of a few experts, but now this knowledge is being constantly captured by a system so that everyone who needs it can have access to it.
Likewise, if you think of changing climate and weather patterns. These also impact the crop yield, and by association the pricing of the product. With AI, the impact on yield and pricing can be forecasted throughout the growing cycle, not just once the crop has been harvested or at the end of the season. Using this data, you can also identify the area a particular crop is sown, and predict yields in that region for that crop. This can be an important input to both supply and demand planning.
Steve: The potential for AI therefore can also be described as disruptive, creating new opportunities for businesses. If you can predict what the total volumes of, for example, a fruit crop will be throughout the season and then match it to other insights on demand and consumption patterns by country then you would be in a better position to manage your pricing and negotiations with other stakeholders in the value chain.
However, it’s fair to say there are other factors at play that AI would have to account for to improve the accuracy of the output.
Prasad: Yes, that is exactly right. The output is driven by multiple factors e.g., the age of the orchard trees, or the conditions under which the crop is being grown, or market trends.
Take the last point, and the example of Oranges. Here, planting and harvesting decisions must be made on a multiyear basis. These are made based on market, consumption and even variety trends. Plans, or programmes, are developed at the season level, by multiple growers and exporters, based on known data. So, at the individual grower level there is a lot of data that is not known, and worse, within the season a complex set of dynamic factors can make even the best plans obsolete. This can lead to oversupply, shortages, uneconomic prices, and ultimately waste. Greater data capture and data sharing, even at an aggregated, anonymous, level could help generate using AI better predictions and insights to support decision making and action plans. Synthesising the analysis of big data sets related to multiple causal variables is an area we see of great potential value to the Produce Industry, Agriculture, and land use in general, and it’s an area we are focusing a lot of attention and investment.
Steve: From a sustainability perspective, is there a way of associating growth and harvest data, so that at a granular level a consumer would know the conditions under which this orange was grown? Do you see a short to medium term future where all the data that you are talking about can be distilled down to a “consumable” unit?
Prasad: My prediction is that as data becomes more readily accessible to the consumer there will be a shift where people are going to consciously try to buy from the farmer who is growing organically, treating his people in the right way, not using pesticide, and being efficient with water use, etc. All those things matter. People don’t mind paying slightly more if they feel it will help the planet and support sustainable practices.
However, for that, the data that we are providing to the consumer must be trusted. Every participant in the value chain has a responsibility to make visible, and authenticate, sustainable practices because at the end of the day they will be judged on their ‘sustainable performance as much as the products and services they are supplying. Without real trust it’s simply “Marketing” which people will quickly ignore.
Steve: So, trust is key, but can technology deliver the information into the consumers hands?
Prasad: If you buy a sandwich for example, it will tell you the calories and other important information on the packaging. Yet when you look back 20 years, that type of disclosure simply wasn’t there. The same thing will happen with information on sustainability. Similarly, if you fast forward 20 years from now, people and Governments will also expect that level of disclosure from a sustainability angle. It’s just a matter of time. The technology is largely there today. It’s more a case of building adoption and finding the right business cases to justify the investment.
Steve: So, if I can expand further on this point, what you are saying is that it’s as much about a mindset change as anything, and maybe AI gives you the ability to think differently, and more holistically about the whole value chain. To move perhaps from managing events or activities to thinking about optimising outcomes.
Prasad: Exactly. Having AI solutions that build insights and recommendations on top of data collected in real time from multiple sources, can present the information in such a way that a business user can use it more effectively. It makes the right decisions more obvious and the implications for the rest of the value chain more transparent.
A Breakthrough Technological Link for Supply Chains?
Steve: Are there other new opportunities for AI usage in the supply chain? For example, do you see AI being used to anonymise and aggregate data to enable it to have broader industry level usage? In other words, data that is available across the ecosystem that can be used by businesses to make better decisions.
Prasad: Back to our fruit example. Imagine you are a producer’s council. If you know the current inventory of citrus in the UK market, the rate of the consumption and how they both map historically, you can advise your producers about potential excesses of inventory and that creates an opportunity for delaying picking, packing, or shipping or even redirecting the fruit to a different market. Ultimately, it’s in everyone's interest to see the freshest fruit getting into the consumers hands.
Steve: Absolutely, and we are seeing more and more producers question how best to optimise shelf life and plan for unprecedented circumstances.
You’ve given us some great insights and examples about the value of AI in the end-to-end supply chain, are there any other final thoughts you can give us?
Prasad: The point to take away is that the amount of accessible data is consistently growing. In the supply chain, cross-enterprise data sharing is still incomplete, but I believe that will also continue to evolve. It’s also fair to assume that the amount of data available for AI for any corporation will of course continue to grow and that encourages people to use similar technologies to make better day-today decisions for their business. This will impact the overall growth within these companies in a more sustainable way as sustainability continues to underpin investments now and in the coming years.
Thank you to both Steve and Prasad for taking the time to create this Q&A!