AI on the Farm: What I Learned Speaking at the Great Yorkshire Show
I'll be honest — when I was invited to join a panel discussion at the Great Yorkshire Show, I wasn't entirely sure what to expect. I'd never been before, and it's not the most obvious venue for a conversation about APIs and data dashboards. But the session — exploring how farms can use AI and automation safely to work smarter — turned out to be one of the more grounding conversations I've had about this technology in a while. Grounding in the literal sense: it's hard to get too abstract about artificial intelligence when you're talking to people who need it to work in a muddy field.
I shared the stage with Richard Lanning and Sam Hoste, with David George from Newcastle University facilitating. Here are the things that stuck with me.
The data problem isn't unique to farming — but it's acute there
Most farms are sitting on more data than they realise. Herd records, yield data, weather logs, machinery instruction manuals — the information exists, but it's often trapped in formats that are difficult to actually do anything with. Spreadsheets that were never designed to talk to anything else. PDFs that can't be queried. Paper records that haven't made it onto a computer at all.
One of the points I made during the session is that the first step for a lot of farms isn't AI — it's just getting the data into a usable format. Moving from Excel to Google Sheets is a surprisingly useful step, partly because it makes collaboration easier and partly because it starts people thinking about data as something live rather than something archived. Getting comfortable with the CSV file format — simple, portable, readable by almost everything — opens a lot of doors that stay closed when data lives in proprietary formats.
NotebookLM as a gentle entry point
Several people in the audience were curious about AI tools but understandably cautious about where to start. We mentioned Google's NotebookLM as one of the more accessible and genuinely safe ways to begin experimenting. You upload your own documents — reports, records, research — and ask it questions. The AI only draws on what you've given it, which means you're not worrying about it making things up from the wider internet, and you're not putting sensitive data into a general-purpose chatbot. For someone who wants to understand what AI can actually do with their own information, it's a good first step without a big commitment.
APIs are opening things up
One of the more interesting parts of the conversation was about the growing number of agricultural services and pieces of equipment that now offer API access — ways for software systems to communicate with one another directly. Lely milking robots are a good example: the robot is collecting data about each cow at every milking, and an API means that data doesn't have to stay locked inside the Lely system. It can feed into a farm management dashboard, trigger alerts, be combined with other data sources. The potential is significant, and the fact that major manufacturers are building this kind of connectivity in as standard is genuinely encouraging.
Custom dashboards are more viable than most people think
The barrier to entry for building bespoke data dashboards has dropped considerably in the last few years. This was something I was keen to make concrete for the audience, because "AI-powered dashboard" still sounds like something that requires a large budget and a specialist team. In practice, a custom system that pulls together data from a few different sources — internal farm records, external APIs, environmental data — and presents it clearly is now a realistic project for a single developer working with a farm business directly. The internal dashboards I've built for manufacturing clients over the years came up as an example of what's achievable without an enterprise budget.
The hype problem
Probably the point I felt most strongly about, and one that came up naturally in the discussion: farmers, like most business owners, can't always picture what's actually possible with new technology until they've seen it working in a context that makes sense to them. That puts a responsibility on developers and early adopters to demonstrate things concretely and practically — and to resist the temptation to lead with the most spectacular possibilities rather than the most useful ones. The hype around AI at the moment is significant, and it's creating a lot of noise that makes it harder for people to identify the genuinely useful applications. Showing someone a dashboard that saves them two hours a week of manual data entry is more persuasive than talking about the future of precision agriculture.
The rest of the day
The panel was only part of it. I bumped into Botham's of Whitby and Whitby Distillery among the stallholders. It's always good to see Whitby represented at something as Yorkshire as the Great Yorkshire Show. I also had a catch-up with Kieran and Lisa from Ladycross, who were there with their stand as well.
The highlight of the afternoon, though, had nothing to do with technology. Euan Findlay was giving a live demonstration of traditional cask-making using hand tools — and it was genuinely impressive. Euan is one of only five coopers still working in England, apparently all based in Yorkshire, and his YouTube channel has built a following of over 170,000 people. Watching him work in what was a seriously warm afternoon, shaping and assembling a cask with the kind of skill that takes years to develop, was a good reminder that "working smarter" doesn't always mean working digitally.
Richard and I also got a chance to try out some of the equipment from the LX Foundry project that the Enterprise CUBE team have been building — OBSBOT cameras and RØDE wireless microphones. We attempted to record a short video, but Richard's laptop had other ideas in the heat. We'll find out whether anything salvageable came of it!
It was a good day out. Not my usual environment, but exactly the kind of conversation I find most useful — practical, specific, and honest about where things actually are rather than where they might be in five years, and realistic about the pitfalls, risks and barriers to entry.