In Part 2 of this Feed & Grain Chat, (watch Part 1 now) Rob Huston, co-founder of Innovative Grain, LLC, explores the tangible ROI of artificial intelligence (AI) in agribusiness, advising companies to start small, identify internal champions and expand adoption as quick wins emerge.
He emphasizes that AI is most powerful when paired with human expertise, as seasoned merchandisers and managers provide the context large language models (LLM) need to deliver meaningful insights.
Looking ahead, Huston predicts rapid progress in agent-based AI systems that can automate complex tasks — freeing professionals to focus on higher-value decisions and relationships.
Transcript of interview with Rob Huston, Co-founder, Innovative Grain, LLC
Elise Schafer, editor, Feed & Grain: Hi, everyone, and welcome to Feed & Grain Chat. I'm your host, Elise Schafer, editor of Feed & Grain. This edition of Feed & Grain Chat is brought to you by WATT Global Media and FeedandGrain.com. FeedandGrain.com is your source for the latest news, product and equipment information for the grain handling and feed manufacturing industries.
Today, we have part two of my insightful conversation with Rob Huston, co-founder of Innovative Grain. This time, he delves into the ROI of AI and shares the vital role of human experience in getting the best results from AI models.
Now, for feed mills or elevators looking to justify investment in AI-driven tools, what kind of gains or ROI metrics should they realistically expect?
Rob Huston, co-founder, Innovative Grain, LLC: I think it's important to start small and look for the low-hanging fruit — try to find those quick wins. If you start small and you find some champions within your organization that are early adopters, and if you work with them, you listen to them, you provide them training, and you help them kind of work out the kinks within a platform, then two things will generally happen. That group of people that you're looking for, the early adopters, they're normally larger than what you originally thought.
And then once you help them work through the kinks and you listen through them and they find success, then they tell the others, and adoption will occur naturally. From there, you can decide to expand on your models and offer it up to other departments, offer it up to other people within the organization.
A lot of these enterprise models have a cost per user. So, if you start out with your small group first rather than just paying for all users to begin with, some of them may not use it because they're skeptical, but allow those champions within the organization to find those quick wins and then to tell others about the efficiencies they've achieved that will enable you to expand it across the organization and keep those returns fairly quickly.
I would say you can expect three times growth for an ROI anywhere between six and 18 months
And that's depending on the organization, but it used to be ROI for three times ROI, it took years. But with AI, you can do it in a much shorter time frame.
Schafer: A lot of grain industry veterans have deep expertise built over decades. How do you help them see AI as a tool to amplify their knowledge rather than something that will replace them? And what role does the human expertise play in making AI actually work?
Huston: Great question. Really important because a lot of people, part of their fear is they feel like AI can take over their jobs one day. And we're a long ways from that, what they call AGI [Artificial General Intelligence].
But I think it's really important to provide that training and show the experienced merchandisers and bookkeepers and managers that their decades of experience or their years of experience can provide a lot of context to the prompt. And that's what's needed.
And these large language models, they don't actually think. They're just a tool. And so you've really got to provide a lot of context to these prompts to make the large language models perform.
And you've got to have accessible data relative to the question you're asking about in order for the LLM to properly access that data, research it. Now, the great thing is the LLMs can research years' worth of internal data and massive amounts of it to do what humans can't possibly do on their own.
So, you put the two together and the years of experience, that context around the prompts to the LLM, and then you get back some really good responses. And although younger people that might be hired straight from college are full of energy and excitement and ready to try the latest and greatest technology, they don't have that years of experience and the contextual background that someone with experience has, so they can't really provide that context to AI in order to get those best answers.
Really, I think a combination of people that are early adopters and then working with people that have that professional experience that can add to it, I think that's where the magic occurs. And then once they see that, then they're like, "Wow, this is a really powerful tool."
Schafer: Yeah, definitely. Now, what excites you most about where AI is headed in the next five years for the grain industry? And how can today's businesses position themselves to take advantage of this?
Huston: Well, there's a lot to talk about there. One of the things that I'm excited about the most is, I'm not going to get too technical here, but MCP, or Model Context Protocol, being able to access different systems through the AI, and then even say, Microsoft Copilot, for example, they started with ChatGPT, but now within Microsoft Copilot, you can now also access Claude.
There's a lot of advancements in these LLMs, and there's always an advancement being announced each week, it seems. So being able to keep up with that technology and enable users to expand into these areas that have advanced is just keeping up with it. And I think the more that continues, the better the models they improve, the more efficiencies we're going to find.
To answer your question specifically, agents that you hear a lot about are really what I'm excited about. Because when you think about efficiencies in particular, you can ask an agent to go do something. And it may take 20 or 30 minutes to go through all the steps necessary because it's accessing other tools, as well. It's not just a tool itself. It's accessing the internet. It's accessing another database. It's accessing some of the context that was provided to it in previous conversations.
While it's doing that, you can continue to focus on more value-added solutions, building relationships with producers, looking at larger, more value-added transactions as opposed to some of the smaller transactions that AI can do.
I think being able to leverage agents, that's kind of the coming excitement and almost all large language models have these days. Learning to deploy those agents to complement your job — not replace your job — but to complement that, that's really exciting. And I think over the next five years, we're going to see a lot of advancements in that.
Schafer: There's certainly lots for us to look forward to, Rob. Thank you so much for sharing your insights with us today.
Huston: Thanks for having me. It was a pleasure and happy to help anytime.
Schafer: Of course! Now that's all for today's Feed & Grain Chat. If you'd like to see more videos like this, subscribe to our YouTube channel, sign up for the Industry Watch daily eNewsletter, or go to FeedandGrain.com and search for videos. Thank you again for watching, and we hope to see you next time!