Perdue Farms has been selling eggs since 1920, and chicken since 1925.
Today, the Maryland-based company is a fourth-generation family-owned business with revenues of $9 billion. And Chief Information Officer Mark Booth is the executive responsible for leading it into the AI era — ideally without breaking more than a few eggs.
We spoke with Booth about the company’s AI journey, including how he works with outside partners, how he prioritizes projects, and what future concerns are keeping him up at night. Booth joined the company in 2020, after stints as the CIO of Whirlpool Corp. and Reyes Holdings, a maker of beer, wine, and spirits.
• • •
Can you tell me how the company has moved from the early days of using AI and machine learning, to how you’ve been utilizing newer AI tools such as generative AI and agentic AI?
We used a lot of machine learning in our plants to understand products. The initial work focused on applying machine learning and computer vision technologies…to improve quality and process verification.
If you think about products coming down the assembly line, we take the chicken and we turn it into grocery store food. We use machine learning in our plants to understand product attributes, such as is the package labeled right? Are there four drumsticks in there instead of three? Is the label readable when it gets to Walmart and places like that?
We moved from machine learning in the plant, and we began actively evaluating and adopting generative AI technologies following the broader commercial emergence of tools such as ChatGPT and Microsoft Copilot. Our initial efforts focused on exploring practical business applications, establishing governance, and identifying opportunities where the technology could improve productivity and decision-making. Our first formal use case is in the procurement area, which is now in production.
What changed when generative AI and agentic AI entered the market?
We looked at our total IT ecosystem. We are moving from an old version of SAP to SAP Rise. We are refactoring SAP as AI comes along, and what I mean by that is when we started this project, we called it business process reengineering. It’s about business processes first, it’s about business strategy first, and it’s about keeping employees in the loop with everything we do. So as we redesign our business processes, we are moving from old SAP to new SAP tools with AI capabilities, and as part of that, [our new] AI use cases need to really differentiate in some way what Perdue is doing.
Can you give me a use case using generative AI that has helped you differentiate yourself in the marketplace?
The first couple we did were in the procurement space. Working with some of our partners we created a spend control tower… The system tells us, for example, “this price is too high,” or “here’s the comparison of X, Y, and Z companies that can give you these products at a reduced price.” The generative AI…will present us with the information, but it’s our employees that actually make the decisions to complete that loop. So generative AI does not make those transactions for us, but [it] speeds up the work. We get in milliseconds what used to take days. It took us several days to understand the marketplace for these various items that we buy… So that’s one recent use case.
Our position is we are not replacing people with AI… It’s about augmenting our employees and doing things better.
You said the last 18 months you have seen a lot of changes, especially as you are moving from the old SAP system to the new AI-embedded SAP software. We’ve seen the adoption of generative AI result in worker layoffs as well as changes in what workers are being asked to do. Can you tell me how your company is going through change management as a result of the AI adoption you’ve implemented in recent years?
Our position is we are not replacing people with AI. And a little personal story: my daughter, Ashley, asks me about AI all the time. She’s an engineer. My philosophy is [that] you will not be replaced by AI, but you need to embrace AI to make yourself more efficient and more productive, and as long as you do that you will be just fine in the workplace.
I believe that, and Perdue believes that. It’s about augmenting our employees and doing things better.
We think of AI on three different levels. The first level is personal AI… We are a Microsoft shop, and we use Microsoft 365 Copilot to support a variety of everyday work activities, including drafting emails, summarizing meetings, developing presentations, creating first drafts of documents, conducting research, organizing information, and improving written communications. The tool helps accelerate routine tasks, so employees can spend more time focused on higher-value work. It is important to note that each employee must sign our AI use policy before [being] given access. We are very clear that the employee is responsible for content produced.
Our second level of AI is within these SAP packages that we have bought, such as SAP Rise, which is a cloud-based business transformation and ERP offering that provides the foundation for accessing newer SAP innovations, including AI-enabled capabilities. Perdue Farms is currently evaluating and implementing AI-enabled capabilities within our SAP environment as part of our broader digital transformation efforts. We see opportunities for tools like AI-enabled SAP Joule to help streamline workflows, surface information more quickly, reduce manual effort, and support employees in completing routine tasks more efficiently.
…For example, if there is an accounting process that used to take 12 human steps, we can use AI to boil that down into two steps. Now, as an employee at Perdue I’m spending my time using AI to analyze things, not to create spreadsheets…
The third level is custom AI. We use cases like the procurement example to implement a very formal process of change management with the team. We train employees on how to use the software, how it’s going to enable you in your workflow, and things of that nature. So each use case gets its own training and change management processes…
How involved is Perdue Farms leadership in AI adoption across the company?
I will credit Kevin McAdams, our Chief Executive Officer & President, for encouraging the leadership team to talk about AI quite a bit. We had the top 150 leaders in the company spend time discussing AI – what does it mean, and how do we prepare employees for AI implementation? We also have national town halls that Kevin does, which reach all the employees, and we talk about AI there as well. So we are very open about it.
How do you prioritize AI initiatives across the enterprise? What criteria do you use to decide where to deploy AI?
First of all, does the project align with the business strategy? How does it fit in holistically, from an ecosystem perspective? If we can solve it with SAP, we will do that, and then if…there are AI use cases that differentiate Perdue, those projects will get prioritized by business case need, such as this enhances market share, or this is about saving money, or this is about quality.
We stack those up and say, “OK, which ones are the most meaningful in differentiation and add the greatest value to Perdue?” We do that at the management committee. We have what we call a transformation management office, which is mostly made up of the management committee, I sit on that as well, and all the strategy and use cases are determined at that level. When it comes to differentiation, it’s an enterprise decision by a management committee as to which AI use cases [we] move forward with…
How do you choose your AI partners? What do you consider when you choose a partner, and how do you find them?
…We have several different partners we work with. We are working with PWC, we are working with Wipro, and there’s others, but it depends on the use case. Chickens are a hard thing — the deconstruction of chicken and turning it into grocery store food, as well as taking on the agri-business side and turning soybeans into oil, and all that kind of stuff. Frankly, we look for partners that understand our business and understand AI. We are also using Lemongrass.
This is going to sound a little harsh, but we don’t want to be reliant forever on partners.
We have a very formal process. We RFP people that we think know whatever the subject may be. Frankly, I use Gartner quite a bit. We are a Gartner subscriber, and Gartner has the Magic Quadrant that looks at leaders and laggards in the industry, and it’s a somewhat independent view of who can supply what and where they are in the tech industry…
We want to make sure that when a partner comes and does something, when they are done, they are done and we can carry on without them… This is going to sound a little harsh, but we don’t want to be reliant forever on partners. That’s probably a bad thing to say if our partners read it, but they know that.
How do you measure return-on-investment? And what are your key performance indicators?
We are very formal on ROI. Let’s talk about this whole SAP move. We did a business case, and within the business case are…lots of different KPIs. There are measurements around costs, savings, productivity, efficiency, and so on, because that information has to go to our board. And you know there are some ROIs that are very positive and some ROIs that are like, “Look, we need to do this, because it is foundational.” That’s sort of the SAP upgrade; it’s foundational. We need to move from the old SAP to the new AI-powered SAP, even though it may not return millions of dollars in savings.
We can talk about the procurement project. Because of the efficiencies from a process perspective, where we said we will do this AI project, create the spend control tower, and it’s going to cost an extra amount to do, and the payback is a year-and-a-half or something like that. The payback was pretty quick. Each project, whether it’s a giant SAP move or a procurement one or a transportation system project…goes through that very formal process. It gets presented at this transformation management office, and gets approved, because we have limited dollars to invest in things like this.
The ones that get approved bring the greatest productivity or efficiency or dollars. There’s any number of KPI, but it’s not always the one that returns the greatest ROI.
What platforms or tools are essential to your AI stack?
Let me give you an example. Dataiku transforms data from SAP into a data lake, which is Snowflake, and that sits on Microsoft Azure. We do have a technical stack of tools that we use to do custom AI development. …We are constantly looking at what are the right tools, and what are the right LLMs. We are not just using OpenAI or just using Claude. Different LLMs have their strengths, and different use cases need those different strengths.
What trends or AI developments are you watching?
…One is just the technical side —how fast things are changing… I would say the flip side is we are watching what AI can do to us. We are a very secure organization and we spend quite a bit of time on cyber security… It’s not just about use cases and business value, but how do we keep Perdue safe? The bad guys are always knocking on the door and now they are using AI to knock on the door even faster…
How do you see the future of AI, and the functionality of AI evolving, within your company?
… We will use AI to differentiate ourselves from our competitors, and that differentiation can mean efficiencies, it can mean innovation, it can mean any number of things.
We’ve implemented agentic AI, but we have not implemented so much agentic AI that we are worried about orchestration at this time. So part of our strategy is as we implement agentic AI, how do we make sure we orchestrate it right? You need a layer above the agents that understands what SAP is doing, what Blue Yonder is doing, and what is Salesforce doing. What is the custom AI doing? And can we orchestrate that all into an ecosystem?
Keeping track of the agents, making sure that the agents work with each other, and orchestrating the agents is…going to be one of the biggest challenges.
What has been the biggest challenge in your AI journey?
We don’t have a ton of challenges, quite frankly. We’ve taken a very methodical approach from the top down, and it’s really helped us. …Keeping track of the agents, making sure that the agents work with each other, and orchestrating the agents is, I think, going to be one of the biggest challenges. We haven’t faced that yet, but that’s the one that keeps me up at night.
Featured image by Nick Fewings on Unsplash.
You must be logged in to post a comment.