Artificial intelligence and agriculture: how can AI impact vertical farming? What we know so far

IGS’ Chief Technology Officer, Dave Scott, and Head of Data, Emily Seward, speak about how we use AI and automation, and what the future holds for artificial intelligence in agriculture.

Recent years have seen the topic of artificial intelligence (AI) rise to the forefront of conversations across virtually every industry. Agriculture is no different. As vertical farming technology providers, we live and breathe innovation, and have done since our inception in 2013. AI (and when looking more broadly, automation) is cut from the same cloth.

But what impact will artificial intelligence in agriculture have on vertical farming and agriculture in the years to come? AI is all too often overpromised, yet underdelivered. We believe that AI in farming has the potential to make a massive difference to vertical farming, while automation is already leading the charge.  

Think streamlined practices and efficiencies, diversified revenue streams, and increased productivity. All of these are ways this technology could significantly impact agriculture with AI.

It’s important to note we’re still very much at the start of the journey, and it could be years before we see large-scale AI making a difference in our industry. To help get an idea of what the road ahead looks like, we’ve roped in IGS’ Chief Technology Officer, Dave Scott, alongside Head of Data, Emily Seward, to speak about what the future holds.

Applications of AI in agriculture

Dave Scott, Chief Technology Officer: “When it comes to the use of AI in agriculture on a bigger scale, the most challenging problem we’ve got to solve (technically speaking) is addressing the number of variables we can adjust. There are so many – take light pulsing and phasing, and gas composition within the environment, for instance.

“People tend to follow the path of best result. For instance, if we do a higher/lower temperature, then this crop works better. But there are hundreds of variables. You’ve then got to factor in what actually happens to crops in the natural world: plants will go through different periods of stress and rest which can distinctly impact the end product.  

“An example of these variables would be lighting. Indoor farms traditionally work on a binary system (e.g. from zero colour, to full colour). That’s not how nature does it – it will work gradually across the course of a day. You’ve then got how you deliver the light – do you give the crop a stable, steady light, or do you allow it to pulse and flicker? This is just one element, but when you break it down, the variables are huge.

“A single person cannot look at all these elements in a trial environment, so looking at such a vast dataset is where machine learning (ML) and AI come into play."

"We can analyse a near-infinite number of variables to look for trends that we wouldn’t even have considered had we not accessed that data. After identifying the trends, it can then be used to carry out the task at scale.”

Emily Seward, Head of Data: “For us, we want to use artificial intelligence in agriculture to improve the grower experience by saving them time, effort and resources. For example, we’ve trained an AI crop segmentation model to analyse images and identify which bits of the image it thinks are crop. It is a type of computer vision model, the goal being to separate the image into regions we care about that contain crops and regions we can ignore, such as the background of the image.

“We’ve had to do this over time by feeding it the right information, ensuring it is learning to identify crops and not just labelling everything that is green as a plant. The more images we use to train, the better the model becomes and the more we can use it for crop monitoring.  Already, our crop monitoring applications can do things that humans can’t do at a glance – for example, detecting small changes, which to the human eye are difficult to spot, but we can train an algorithm to look at.  

“Plants are incredibly complex. It’s not always a case of A + B = C. Instead, you've got a whole alphabet of inputs, and the outcome can depend on how you combine them. AI (and automation) can help deal with these incredibly complex, multifaceted inputs and turn them into something we can actually use.”

Using AI and automation to address common pain points

Dave Scott: “For me, the key areas we can use AI and ML in agriculture are in improving efficiencies and working at scale. Scientific experiments traditionally work in terms of a planning process, execution, and finally evaluation. You’ve then got to repeat the process to experiment on different outcomes, which can be time and resource intensive.

“By using AI and ML, we can remove these stalling points and use machines to gather a huge amount of repeatable – this is key – data."

"We do this with our Growth Towers through the industrial lifts which operate between towers. These can determine with absolute precision the same relative location for every tray and take many different measurements which can be used to help grow crops.

“We use hardware to gather repeatable datasets over a large period, with touch points that allow the AI and MI to make iterative improvements to the way crops are grown. This is only worth doing if you have a large dataset and can gather actionable insights through it.”

Emily Seward: “Having the dataset is key, yes. AI isn’t worth all that much unless you’ve got the solid data infrastructure to back it up. Even without AI, growers get so much benefit from having this infrastructure and the operational oversight it delivers.

“On a more granular level, AI has the ability to significantly speed up the crop recipe development process. Traditionally, we’ll change one or two variables at a time to be confident about the impact that they have on the recipe. With AI, there’s the potential (although not currently the realisation) to quickly create tailored recipes, taking input from crops IGS has grown at our research centre in the past, and using that knowledge to reduce the human, nutrient, and energy inputs required during the experimentation process.

“You’ve then also got benefits for the recipe outputs by tailoring recipes to improve speed, growth, nutritional content or even height. This minimises development time and allows for a more bespoke growing process, which can help to give the customer a competitive edge.

“I’d caveat this by saying that a characteristic like taste is highly subjective and varies greatly. If it is difficult to quantify a desired output like taste and automatically collect data on it, it is harder to optimise for. It all comes back to having the data infrastructure to back up the AI – it can improve the observability of crops and, in turn, the quality.”

Diversifying revenue streams through AI

Emily Seward: “AI has huge potential, so it can be taken in many directions to help diversify revenue. It could be used to regulate energy use, for instance. This way, growers could effectively work alongside the grid to even out usage, selling electricity back to the grid. This effectively acts as another revenue stream, made simpler through AI.

“You can also use it to speed up the research phase and help growers expand the portfolio of crops that they are growing. For instance, you might start out growing lettuce, but by using AI to determine optimal recipes at a fast rate, you can diversify output and start to grow basil, too. Or diversify your lettuce production to cater to a wider range of end customers.”

How we use AI and automation in IGS technology

Dave Scott: “We can also use automation to help alleviate any labour concerns. This is standard-practice in the industry and has been for the last ten or so years – it's just being talked about a lot more now along with the rise of AI. Both AI and automation complement each other, but they are entirely separate.

“Let’s start by looking at tasks which traditionally are quite manual and labour intensive, such as cleaning the trays and harvesting crops. We’ve designed our technology so that this can either stay manual, be semi-automated, or fully-automated. Full automation allows crops to be delivered to production lines, loaded and unloaded, then the trays to be cleaned.

“Again, this comes back to optimising crops for elements that your everyday supermarket doesn’t really consider at the moment – such as elevated nutrition or extended shelf life."

"Once this becomes more mainstream and the quantities become more stable, that’s when automation comes into its own and the whole process can be streamlined.

“You don’t want to automate for a market you don’t fully understand. Automation using IGS technology helps growers to scale effectively, making efficient use of resources while upping production levels.”  

Emily Seward: “We want to use AI to take away a lot of the boring, menial work which is required day-to-day on a farm. For example, we can use AI to warn a grower if their crop is getting too tall without them needing to manually inspect thousands of images. It gives growers peace of mind that crops are growing as expected, and you can then streamline the flow of information to create efficiencies. This way AI allows people to do the same jobs more efficiently, freeing up time for other areas of work.”

About IGS

Intelligent Growth Solutions (IGS) is a Scottish-headquartered vertical farming technology provider. IGS has combined crop science, engineering and software (amongst other disciplines) for over a decade, in the process deploying technology across multiple continents.

In 2023, IGS announced a deal with ReFarm at COP28 in Dubai to build the world’s first ‘GigaFarm’. This ‘GigaFarm’ is set to replace 1% of the UAE’s total food imports, helping the region to reduce food emissions through vertical farming.  

About the interviewees

Dave Scott, Chief Technology Officer

Dave Scott is the co-founder and Chief Technology Officer at IGS. Dave’s role draws on his significant engineering experience in industrial automation and electrical control systems. He has developed breakthrough technologies for indoor growing environments, addressing the critical economic and operational issues of power reduction and increased automation in this sector.

Emily Seward, Head of Data

Emily Seward is Head of Data at IGS. She leads a multi-disciplinary data applications team comprised of data scientists, data engineers and visualisation experts. The team uses advanced data analytical techniques (including artificial intelligence and machine learning) to build applications designed to enhance the experience of growing in an IGS Growth Tower for the business’ customers.