With artificial intelligence (AI) growing at a faster rate than any other technology in the past, grain elevator operators should prepare now to ensure their facilities reap the benefits while protecting their data.
Over the next five years, AI growth rates are expected to be 29% year over year, compared to the 6% to 7% growth rate for the technology market in general, said David Smit, operational technology architect at Interstates, Inc., during the Grain Elevator and Processing Society’s (GEAPS) Exchange 2025 event in Kansas City, Missouri, US, in February. One report said by 2026, 80% of organizations will be using AI in some form.
“That’s why it’s important,” Smit said. “It’s not a matter of are we going to be able to adopt these technologies; these technologies are coming, and we must be ready for them. We’re not going to be able to do what we’ve done the last 15 years. We’re going to have to make some changes to support this or we’re going to have issues.”
There are three basic types of AI, Smit said, including machine learning, which has been around for several decades. With this system, machines learn from your data and improve over time. Generative AI, such as ChatGPT, will generate content based on what you say or what you ask. The Holy Grail of AI, Smit said, is agentic AI, which can make decisions and perform actions without direct human intervention.
To prepare for AI technologies, a facility must start a digital transformation journey, if it hasn’t already. Smit suggested facilities complete a self-evaluation to determine where they are on the journey. Next, infrastructure readiness will be key along with security applications.
“As we’re adding technologies to our facilities, we start to have more security vulnerabilities,” he said. “Fifty years ago, all our networks were isolated. Now as we start to apply these new technologies, our networks are becoming exposed, and cybersecurity is becoming a real issue.”
Smit also discussed the capabilities of AI for predictive maintenance and operational efficiency gains and concluded with a discussion on what data governance is and its importance.
“We’re going to move away from gut feel toward data-driven decisions,” Smit said. “Soon, data-driven decisions are going to develop what AI-driven decisions will be in the future.”
Digital transformation
Digital transformation means turning current practices into a digital process so that data is available for other things, Smit said. There are five stages in digital transformation, he said, which is the starting point for any AI system.
The first step is standard reporting, which doesn’t include a lot of automation and is paper based so there’s not a lot of data intelligence. Next is descriptive analytics where data is being collected and stored in a central location.
“We have that as a reference going forward and we may do some number crunching around that,” Smit said. “That’s all manual based. We’re not putting any intelligence around that.”
The third stage is diagnostic analytics, which includes starting to automate informed decision making and problem solving. For example, some of the data collection, such as bin levels, is automated and the information is stored in a central database. From that, decisions can be made automatically, such as when it’s time to start filling a new bin.
Sensing new opportunities with AI: Supplier: Grain industry on ‘cusp of an intelligent revolution’
Predictive is the next level, Smit said, which is where predictions can be made on what’s going to happen next based on the data. For example, predicting when equipment might fail based on hours of usage, and replacing it before that happens.
“I would say there’s probably somebody in here doing all of these phases today,” Smit said. “There’s one final stage called prescriptive analytics, and this is probably where nobody is really doing anything today yet, mostly because it’s new. This is where AI really starts to take place.”
In all other stages, operators were making decisions based on the data they had. In prescriptive analytics, the system is making the decision for you, he said. This is the level needed for such things as autonomous or manless elevators, he said.
“My challenge to you is do you understand what data you have today,” Smit said. “What are you collecting and what do you need to be collecting. It’s really important to start collecting that data today so that maybe in two or three years, when you’re ready for these solutions, you have the data you can apply.”
Artificial intelligence in agriculture is changing the way farmers farm
Infrastructure readiness
To create these systems, the network infrastructure must be in place, Smit said, including digital devices. As facilities go through lifecycle planning, they should consider how to add smart technology, like sensors. Connections to the Cloud most likely will be needed.
Start reviewing what systems already are connected to the Cloud and build policies and procedures around that. Assess the reliability of internet service and determine if a second provider is needed as a backup in case the first fails.
“Start to look at scalable hardware and software,” Smit said. “If you’re going to replace a PC, investigate if you should invest in a virtual solution that’s going to prepare you for the future.”
Smit said to talk to vendors about what data they are storing and how it’s being used. For example, is your facility’s data being used to train your competitor’s systems?
“Continue to educate yourself on the emerging technologies,” he said, and make sure to train employees. “As we move from an elevator with just buttons and switches to more of a digital environment, we need to make sure that we’re training our people.”
Smit suggested making a digital maturity road map that includes what already is being done as well as future goals.
“It’s good to identify what things you’ve done already, what things you need to do today and what things you need to do tomorrow,” he said. “What’s your end goal? Do you want to be the facility that has no people? Or maybe you want to start with grain quality, so you can automate the process to get great samples, analyze those samples and do grading?”
David Smit, Interstates, Inc., shares how grain facilities can prepare for artificial intelligence during the Grain Elevator and Processing Society’s 2025 GEAPS Exchange in Kansas City, Missouri, US.
| Credit: ©SOSLAND PUBLISHING CO.
Cybersecurity
Security implications are significant given the amount of data involved in AI systems. Cyber threats are on the rise and adversaries are starting to use AI as a tool to attack, Smit said.
“We need to consider how we can use AI as a defense mechanism,” he said.
For example, automated responses are possible with AI, so if a network is attacked after hours, it will notify the appropriate person and provide the pertinent information.
Network segmentation is needed to provide security measures, Smit said, including a firewall and micro segmentation. For example, don’t allow devices to talk to each other if it’s unnecessary.
Monitoring tools are available that can listen to all the network chatter and show what’s communicating with what and how to secure it, he said. Understand your assets and threat vectors.
“One of those biggest risks is your employees,” Smit said. “Talk to your employees about this and make sure that they understand the cyber implications and follow your company guidelines.”
Predictive maintenance/efficiency gains
AI can help a facility become more proactive instead of reactive when it comes to maintenance. Unscheduled downtime happens because it’s not known when something is going to fail. But with predictive maintenance and the use of historical data, AI can anticipate when equipment is likely to fail, Smit said.
“We know when things are going to fail after 100 hours because it’s happened 20 times in the past five years,” he said. “We can plan for those outages and build our maintenance schedule around those.”
AI could also play a role in operational efficiencies by automating processes. Smit suggested mapping operational workflows and looking at high-impact areas where AI could be beneficial.
“High impact doesn’t necessarily mean the biggest thing,” he said. “Look at what it means to you. Maybe you have a lot of manual processes today that cause inefficiencies. Maybe we don’t know how to optimize where we store our products or maybe you have a transportation problem because you can’t do grain inspections fast enough.
“With AI we’re automating that process and doing a real-time allocation of resources.”
For example, given the temperature and humidity, AI could determine that a fan should run on a particular bin at a certain speed and set amount of time.
“We’ve optimized that,” Smit said. “We’re not letting the bin fan run all weekend long because the operator forgot to shut it off on Friday.”
He suggested starting small rather than trying to solve all the problems at once. Think of something that takes 30 seconds to do but is done multiple times a day and by more than one person. For example, use AI to create a meeting agenda template that can be used repeatedly.
“It doesn’t take long to do a meeting agenda, but every time we do that, we have to think about it,” Smit said, but asking ChatGPT to generate a template makes it easier.
Data governance
Smit said one of the most critical elements in the move to AI systems is data governance. Less than 30% of organizations have processes and policies in place for AI, he said. At the same time, 29% of the data being added to AI contains personally identifiable information, such as Social Security numbers.
“If you, or your organization, haven’t put these data policies in place, go back and tell them you have to do this now,” Smit said. “Employees are going to start using AI, so making sure that we have clear expectations what you can and cannot use AI for is important.”
Other aspects of data governance include classification and management, including ownership and accountability; quality and integrity such as data standards, quality metrics and validation; and security and privacy, including access control, data protection, privacy compliance and incident management. Data retention and disposal are also important, Smit said.
“How long do I keep this data for? Is it six months or six years?” he said. “If you give bad data to ChatGPT, you’re going to get bad outputs. Uniformly collecting data across your entire organization is important.”
AI is not going away, so now is the time to prepare, Smit said.
“I think we’re at this decision point where maybe you don’t like this, but we have to be willing to embrace this technology because our people will,” he said.
