Stephanie Feraday & John Haupt: aPriori & Our Customers The Transformative Moment

Stephanie Feraday: Take a look at these headlines. Different companies, different industries, different parts of the world. Together, they describe a set of pressures that I suspect everyone in this room has faced, probably even recently. The cost of making a product keeps changing. The economics of where to make it are changing. The competition is changing just as rapidly. So starting with materials and tariffs, for engineering and sourcing teams, a change in the price of aluminum or steel immediately raises practical questions. Can we absorb the increase? Can we pass it on? Should we change the material? Change the design? What should we do in response to that? But each choice affects something else. An alternative material might lower the cost, but it could change performance.

A new supplier might offer a better price, but require requalification, tooling, and time. A change in design might create savings, but only if we understand the ramifications and the implications before we commit. So those decisions become harder when the assumptions keep moving, and they do, day in and day out. A sourcing strategy that looked attractive six months ago could be completely different today. Just look at the Caterpillar headline. Big benefit. They got almost $400 million in tariff savings back in August, but they’re still forecasting billions of dollars in tariff costs for 2026. And the S&P Global forecast, which looks at material prices, is forecasting continuing increases in material prices. So the underpinnings of what we’re doing are just shifting as the sands are.

At the same time, global competitive pressures are continuing to compound what we’re doing. Bloomberg’s headline about the $150 billion China shock quantifies the pressure that Europe is facing. The Volkswagen board took some really serious action. They reduced 50,000 jobs. They’re cutting their product lines in half. They’re shutting down four plants. That’s pretty aggressive. That’s due to competition, tariffs, and technology changes. So all of those things make it much more difficult for us to operate in today’s world. Yet, what do our customers expect? They still expect quality, performance, and delivery, and they expect to do it at a reasonable price. So they’re not giving us any leeway. That really puts tremendous pressure on decisions that are made early in the product development process.

When you’re choosing an architecture, choosing a material, choosing a supplier, those are decisions that impact everything that comes after. And then there’s the question of where to manufacture. Look at the GE Appliances headline. They announced that the company is committing a billion dollars to moving to Kentucky from Mexico. That’s a major strategic change involving facilities, labor, processes, and product plans. The IndustryWeek headline shows us how difficult all of the choices that I’ve just walked through can be. Tariffs are painful. Reshoring is slow. Moving production takes time and investment. Those are major strategic choices, but the business case has to account for labor, productivity, logistics, tooling capacity, and the capabilities of our supply bases.

So what connects all of these headlines? They highlight a growing need to evaluate more alternatives, understand the trade-offs, and make decisions even faster. For many organizations, the challenge they still face, and I was just talking with somebody about this before I walked up here, is that the information needed to make these decisions is spread across teams and systems. That’s why I find the final headline so relevant. With all of these things changing and the data spread across organizations, how do you really design for margin, for profit goals, when it’s all changing? If you can’t control the tariff and you can’t control the strait, redesign the product. It points toward the choices that we can influence.

We can examine different materials, simplify a component, consider another manufacturing process. And if we ask those questions early enough, the answers can really change the outcome. So that’s where AI becomes particularly interesting. Helping teams explore more options, get insights sooner, bring the manufacturing knowledge and the cost knowledge right into the early phases of design while there’s still room to act, and act with speed. AI can really make a difference. So let’s look at speed and economics. Speed changes product economics. The interesting thing is it’s both positive and negative. This is the economic reason to care about speed. Early in a program, teams can still change the geometry, the materials, the process, the region, the supplier strategy.

Later, those choices become constrained by tooling, by contracts, by qualification, and by launch commitments. A perfect analysis delivered after those commitments are made creates limited value. It’s too late to really make a lot of changes. A sufficiently accurate answer delivered while alternatives remain open can change the outcome significantly. This is where AI becomes particularly interesting, helping teams explore more options, get insights sooner, and really bring the knowledge into everyday design decisions that can help achieve the product outcomes. But AI really becomes strategically valuable when it compresses the time between a question and a manufacturing-grounded answer. If you can bring those together, that really makes the difference. So the goal isn’t to create more analysis.

It’s not just to iterate, iterate, iterate. The goal is to move critical knowledge into the part of the product timeline where the teams still have leverage and can use that data to make the decisions quickly. So let’s talk about AI, because that last note really pointed right at it. And I want to be balanced, so let me start with what’s actually working. Nothing on this slide is a vendor claim. This is research from manufacturers and from research organizations in terms of reporting what companies are getting as a result of using AI. I’ve organized them in a way that you can see some examples across the design-to-source-to-make process. GM, and this was in June, did a roof crush analysis.

That was something that used to take anywhere between eight and 40 hours. With AI, it now takes five minutes. And that’s AI together with simulation. That’s just one analysis task. It’s not an entire vehicle program, but think about what it changes. It really gives you the ability within minutes to do multiple iterations. You can start asking questions of every alternative as you’re using AI. When an answer stops taking a week, that’s what you can do. Engineering teams are using AI to evaluate more than three times as many design alternatives in the organizations that are using it as part of their day-to-day process. More alternatives earlier is how you find the problems while they’re still inexpensive to fix.

On the sourcing front, Ivalua surveyed 800 senior procurement decision-makers in March. Seventy-four percent were already using or experimenting with AI in direct material sourcing. That’s great, but what’s happening with that? That’s enabling engineering RFQs to turn around roughly three times faster. So sourcing is using AI, but it’s actually enabling engineering to do more as well. That matters for the decisions we were just talking about. Material changes, supplier changes, regional moves, those all run through sourcing. And sourcing has historically had a big challenge getting to a defensible number. Further downstream, on the manufacturing front, Samsung announced in March that it intends to run all global manufacturing as what it calls AI-driven factories by 2030.

That’s incredible. Digital twin simulation with AI agents for quality, production, and logistics. So the promise isn’t hypothetical. It’s demonstrated in all three areas across the design-to-source-to-make process where you make decisions. So it raises the obvious question: Why hasn’t this become common practice for all major manufacturing companies? We were talking about this at a roundtable yesterday. It’s not yet common practice. Only 9 percent of organizations report a mature, scaled engineering AI program. Many others are piloting or experimenting, so that’s definitely going on in many different pockets. But 74 percent name data preparation and availability for AI as their key barrier. In that same Ivalua research, 44 percent say that their own supply data isn’t ready for AI.

What we’re seeing is organizations working across their systems to try to pull that data together and cleanse it so that it can be used by AI. That’s step one. But a lot of companies are still at that first stage. Procurement leaders are telling us that the constraint isn’t the LLM or the AI model, it’s what the model has to work with. But those numbers can only tell you that it stalls. They can’t tell you why. So let’s listen to what engineers are saying about why it’s stalling. This is industry-wide, and it really comes down to three questions. Can I trust it? How do I know the recommendation is correct? Who’s responsible for the advice if it’s wrong?

That second part of the question is the one that quietly stops adoption. An engineer who acts on a recommendation owns the outcome. It’s not the tool. It’s not the LLM. It’s the engineer. So nobody’s going to stake a release on data that they don’t fully understand and agree with. The second question is: Does it work? And this is really about, “It feels like one more thing I have to do.” My day job is already busy. I’ve got a lot to do to get the product to ship, but this is one more thing I have to do. If the output is noise rather than signal, AI doesn’t reduce the workload, it just adds another queue to manage.

And then the third question that seems to be proliferating is: Does it know my reality, my context for what I’m doing? Advice that ignores the supplier capability is advice that nobody follows. A recommendation that can’t be made in your plant or by your supplier isn’t a recommendation, it’s a distraction. So engineers want an early nudge. They don’t want a late flag after the decision’s already locked in. Junior engineers want an explanation as to why, while senior engineers just want brevity, and they want the answer. But that’s the critical thing. It’s the why. So the conclusion I draw from all of this is that AI adoption was never a capability problem.

It’s a trust problem. Do I trust the data? And trust in our world comes from one thing: data with real context delivered while the decision is still open. That’s a very specific problem, and it happens to be one that we’ve been working on for over 25 years. Large language models bring powerful language and reasoning capabilities. Manufacturing decisions require an additional layer of context. A credible manufacturing answer must understand geometry, materials, process sequence, machines, labor, regional economics, supplier assumptions, and production constraints. There’s a lot. It also needs to show the assumptions that produced the answer so an expert can challenge or potentially refine it. Now, this is interesting from NIST. NIST’s guidance on industrial AI emphasizes that data must reflect real operating conditions and that incomplete or oversimplified data can create risk.

That’s why manufacturing context matters. Fluency can make an answer sound incredibly confident. How many times have you interacted with an AI and thought, “Oh, that’s really expert advice. It’s exactly what I needed,” only to find out later that there was something behind it that wasn’t quite right? Well, domain data and physics in this context really make these decisions ready. Now we see that there are critical gaps between design engineering, sourcing, and manufacturing that are really inhibiting them from working cohesively. Most of us have lived some version of this, probably even in the last quarter. These three roles make decisions about the same parts, the same products, but none of them can see what the others are deciding.

If you look at a design engineer, he selects a tolerance. It takes about 90 seconds. Nothing in the CAD system tells him what that tolerance will cost downstream, whether the suppliers can hold that tolerance, and whether they can actually do it at that price given the characteristics. This is not a knowledge gap. Engineers know perfectly well that tolerances drive cost. It’s a feedback gap. The information arrives after the decision, if it arrives at all. Procurement negotiates without a floor. Without granular should-cost, there’s no reliable way to know whether a quote has reached its fair value or simply stopped near the opening discussion. If you look at industry-wide studies, they put leakage across a development cycle at 3 to 5 percent of annual spend.

It sounds like a small percentage, but it’s a huge number. And what’s more interesting is that, on a large program, that’s a material number that’s invisible because you can’t see money you didn’t negotiate. So there’s a huge number that’s just sitting behind the scenes that should be readily visible. And if you look at manufacturing, last but not least, they surface an issue at first article. That same issue would have cost nothing if it was captured earlier in the design process. Except now there’s tooling, a qualified supplier, and a launch date attached to it. The information needed to prevent all three of these issues existed somewhere in the organization. A little bit of it was in CAD, a little bit was in PLM, a little bit was in ERP, or in the experience of the engineers who have made a part like that before.

It simply wasn’t connected, and it wasn’t present at the moment someone had to make an answer. So my take: This isn’t a people problem. Engineers and buyers know their jobs. They’re experts at what they do. It’s a data problem, and data problems are solvable. So if you take one thing from today, let it be this: What an AI system can be trusted to do depends entirely on what sits underneath it, the context. Three things sit underneath ours. First, physics-based data rather than estimates. This is the ground truth of what it costs to make something in a given process, in a given factory, in a given region, not an average of what similar parts cost to make at some time in the past.

That distinction matters most, especially when you’re in a new factory, in a new region, making a new part. Those are the situations that are most challenging. Now, to give you a sense of the specificity involved, a general-purpose AI model can write fluently about manufacturing. What it can’t do is tell you that moving a particular geometry from a three-axis mill to a five-axis mill in a plant in Vietnam, at a specific volume with a specific material, reduces piece-part cost by $1.43, while adding 12 seconds of cycle time and four-tenths of a kilogram of CO2. Secondly, a deterministic simulation is one where the model returns repeatable results. The same inputs every time return the same outcome.

It doesn’t change. That’s context. And that’s what makes a number defensible in front of a supplier or when you’re with an auditor or in a design review. And when your team disagrees with the result, they can open it up, they can interrogate it, and they can change it if they think the assumption should be different. That’s a very different relationship than being asked to accept an answer on faith. Third is the regional economic data. Look at what’s happened in terms of cost variability between tariffs, material prices, and logistics over the past 12 months. Having current data matters as much as how deep it goes. All of these things produce the outcome on the right, which is the digital factory.

This is the context for the decisions. And this is what many of the customers that we’ve interacted with tell us is one of the most important things in the context of their decision-making. You end up with your own digital factories, your machines, your processes, your tolerances, your overhead, your supplier relationships encoded so the simulation reflects how you actually make things. The context. That’s what it’s all about. Now, a fair question at this point is where we’re going with this. How does it sit relative to the technologies you already own? We’re not replacing your PLM, your ERP, or your procurement systems. We’re providing data that will complement these systems to help you make better-informed decisions.

On the design side, we connect to major CAD and PLM systems today, Dassault, Siemens, PTC, so that the analysis can be triggered by work your engineers are already doing. That’s the automation that we’ve been talking about, rather than a separate request to a separate team. This is how to give teams grounded answers to reason over, rather than having them infer manufacturing cost or manufacturability from their own language. So let’s take a look at what changes if we actually get to this point. And we’re well on our way there. This should help programs launch at or above target cost instead of below. The 3 to 5 percent I talked about a few minutes ago that erodes during the development process, the gap between the target you set at kickoff and the cost you actually ship, becomes visible early enough to act on.

On $500 million of annual spend, think about even if you recover a part of that, say $15 to $25 million. That’s significant. And unlike most cost reduction programs, it doesn’t require renegotiating anything or asking a supplier for a concession. It happens before the commitment is made. Manufacturing issues get resolved at design, when the cost to fix them is zero. The loop that costs six months and several million dollars on a major program, discovering issues at first article inspection, redesigning, requalifying, that goes away. That becomes avoidable. Not every time, but often enough to change a program’s economics, and that’s significant. So every commodity manager in the new world will be able to negotiate with the precision of your best cost engineer.

Most of you have a few people who can build defensible should-cost from first principles. Unfortunately, they don’t have the bandwidth to cover every category, be on every negotiation, so a lot of those negotiations happen without expert support. What I’m talking about is digitally extending that expert’s reach across the portfolio and really leveraging their judgment. Not replacing their judgment, but leveraging the impact. Programs move faster because handoffs aren’t waiting for information that has to be requested, gathered, and reconciled. And your engineers spend their time inventing rather than reconciling cost. Given how difficult these people are to hire and hold onto right now, that matters as much for retention as it does for output.

Now, I know some of you in the room are suppliers. I want to take a minute to talk to you as well. We all depend on suppliers. They’re critical. Should-cost has a reputation problem, and it’s worth naming. Used badly, it’s a blunt instrument. And suppliers, all of you have the expertise to tell immediately when a number’s been constructed to win an argument rather than describe reality. Used well, it can actually do the opposite. When both sides are working from the same physics, the same regional data, the character of the conversation changes completely. It stops being a contest over who has better information and becomes a discussion about where the cost actually sits and what design or process change would remove the cost.

They collaborate on driving a better outcome. Several of our customers have told us the most valuable outcome was discovering which of their own requirements were driving cost without adding value. So three things change in this world for suppliers. They can build a credible, defensible quote on the same basis as their customers, which means less time defending a number. They can be brought in earlier during the design when the process expertise is worth far more than during the negotiation phase. And that matters because every year more and more manufacturing is increasingly sitting in the hands of suppliers outside the four walls of the manufacturer. And lastly, suppliers will receive 3D models that can actually be made as specified.

They won’t have to go back and forth, which removes the rework loops, the change orders, and the margin erosion that makes these relationships adversarial in the first place. So let me close this section by making it concrete, because “intelligence layer” is an abstraction and your next program is not. So rather than take my word on any of this, I want to ask someone who’s been living through a lot of this to come up and join me. John from CNH, please welcome him.

John Haupt: Thank you, Stephanie.

Stephanie Feraday: Just to give you some background on John, we’ve really had the very great luck to work with John for several years, starting when he was in his engineering role. And I’m going to ask him to cover his background, but he’s a very unique species. He’s worked in engineering for a long time, and he’s moved to the world of sourcing, so he’s really covered the gamut. So he can really speak to both the challenges and the opportunities. John, if you wouldn’t mind giving people a little bit about your background.

John Haupt: Yeah. I’m honored to be here today. Thanks for setting up the vision for us, Stephanie. I’ll start at the beginning. I’m a Penn State engineer who grew up on a farm. My dad had a welding business, and I was always attracted to everything off-highway equipment in that space. So I secured my job at New Holland, Pennsylvania, as a field test engineer for a couple of years. I moved into the design space then. I was a skid steer loader design engineer for six years, and then I finished my MBA. I wanted to get into engineering leadership, so I started my career with Ingersoll Rand Road Machinery, and I was a global engineering manager in that space for several years.

Ingersoll Rand was acquired by Volvo Construction Equipment in 2007, and I was in ever-escalating roles in Volvo, and we got up to the final mechanical engineering director position. Then I said, “Look, I want to take a steer toward purchasing,” because we were doing so much collaboration in my roles in engineering. It’s a natural step for me. So I was appointed into the role as Vice President for Purchasing, Region Americas, for Volvo. Then a couple of years later, I got a phone call from my friends back at CNH and they said, “John, can you come back and join us and lead our harvesting engineering group globally?” And I did that for four years.

Then I led the business. I led the harvesting business, a multibillion-dollar-a-year revenue business, for the last five years up until February. In February, I got a new challenge to go global. All products, fully global agriculture, as the Vice President of Advanced Purchasing. So we represent, on all of our 22 business platforms, the whole supply chain, everything from the beginning, from advanced purchasing, all the way to logistics and procurement. So that’s a little bit about me.

Stephanie Feraday: That’s great. And I think you’re new in your role, but you also recently came back from a supplier conference just a couple of weeks ago, right?

John Haupt: I did. We had our global strategic sourcing program supplier conference in Amsterdam. We had about 1,700 people, all suppliers, internal people, engineers, who helped us set up our next wave, our Wave Three of our strategic sourcing program. And you hit on it very well. That’s really about shifting our approach from, “Let’s squeeze these suppliers to get them to capitulate,” to, “We’re moving toward choosing who our partners are.” And that’s a journey that we’ve been on for a couple of years. Even the tools that we’re going to talk about today are really enabling those types of discussions instead of being combative. It’s very much collaborative with our strategic sourcing partners now.

Stephanie Feraday: Fantastic. We’re going to cover a couple of topics. The first we call the transformation mandate, and it’s looking at transformations that are driven by cost pressures, tariffs, et cetera. I think CNH just told investors that you’re cutting on the order of 200 basis points off of ag margins and 600 off of construction. That’s a lot. So what does that pressure actually look like inside of a sourcing organization day to day?

John Haupt: I think a lot of people can imagine what that looks like. It looks probably very similar inside of your own companies. Especially for us right now, we are in the trough. We’re in the downturn. We came off of record highs right after COVID, ’21, ’22, ’23, record revenues. Commodity prices were good. Farmer sentiment was very high. But with all of the external pressures on input costs, our farmer sentiment is not good. And the commodity prices were low. We had oversupply of grain last year. We had a really good harvest, but it put pressure on the commodities. So we have a lot of pressure on us in the procurement space right now to manage this ever-changing world of tariffs.

And we’re trying to respond to that as quickly as we can, which is why, when you emphasize speed, I can’t emphasize speed enough. To get to the right choices. And that’s not easy.

Stephanie Feraday: And so I’m guessing the transformation you just alluded to is part of trying to achieve that. Is that right?

John Haupt: It absolutely is. We’re in the trough right now. Take advantage of being in the trough to fix ourselves and set ourselves up for success for when things really pick up. We’re going to be in a really good position to recover our margins. You’ll hear about it later today. Our margin EBIT is 4 percent.

Stephanie Feraday: Wow.

John Haupt: A normal market would be closer to 12 percent, or at least in that double-digit area. So we have an urgency to drive transformation and partnerships, and that’s what we see moving forward with our suppliers.

Stephanie Feraday: Great. And so, as part of that, you’re looking at standardizing data, setting cost targets earlier in the design. Can you talk a little bit about that?

John Haupt: Absolutely. And that’s been a journey as well. We’ve come through some really significant programs. Just recently, we launched what we call our Next Generation Combine program, the biggest program in CNH history. And we learned a lot through that journey. We’re applying all of those cost target settings, weight target settings, things like that, to our approaches now. aPriori is a huge piece of us being successful in delivering on our program targets and working earlier and upfront. The V-model, shifting left in the way that we work.

Stephanie Feraday: Okay, great. So that actually is a good segue. As you’re thinking about the design-to-source aspect, how are you thinking about aligning early engineering work with sourcing suppliers? And is that seen as a strategic focus or not?

John Haupt: It really is. At the end of last year, we said, “Look, we need a dedicated team to put together our very best approaches for early supplier engagement.” We’ve established that. We think it’s a proprietary way of working now, and we’ve outlined how we manage those parts, how we prioritize those parts, how we drive early supplier discussions, and really partner with those things that are driving the key costs on our products. That was really important as part of our transformation. We created that last year. We’re still driving some pilots on how to better drive early supplier engagement. It was always there, but it wasn’t structured. So we applied our very best practices now, and we’re institutionalizing that across the organization.

Stephanie Feraday: Great. And what does it really take to move that cost upstream in engineering from solely being owned by either procurement, manufacturing, or finance? Because that’s where it sat before.

John Haupt: Absolutely. Part of that transformation was not only the early supplier engagement, but we’ve gone through another evolution of our organization, what we’re calling our platform relaunch. We’ve identified the key business owners of those 22 platforms in agriculture. They have dedicated cost managers. And my team, the Advanced Purchasing team, works hand in hand with those dedicated cost managers. So we have a much faster and more accurate view of where we’re going to land throughout the program. And we can take specific actions in the program before we go to production and find out, “Okay, wow, we had a big spike, and now we’re off our margin targets.” So this platform relaunch was really critical to get dedicated people in those spaces with the right knowledge and the right experience to help us drive the productivity that we need.

Stephanie Feraday: That’s great. And so they’re working upstream with the engineering community as well.

John Haupt: Very closely. Absolutely.

Stephanie Feraday: Great. Awesome. All right, let’s shift to what I think we called “the squeeze.” You alluded to it just a minute ago. For decades, manufacturers have taken more of that squeeze approach to suppliers. “Give me 3 to 5 percent more cost reduction.” Let’s maybe look back. What were the challenges with that approach?

John Haupt: You’ll hear about it in my presentation later, but the relationships were strained. It was a little bit, “Okay, this is your job, Mr. Supplier, Mrs. Supplier, to fix our product costs, and we’ll sit back and wait for you to do that for us.” What we’re finding is it’s much better to partner and get our best design engineers and our best procurement leaders to have open discussions and say, “Let’s sit down and review the technical attributes of the cost buildup.” Part of the strategic sourcing program is transparent costing. Which is not always easy. Let’s put that out there. But the point is, when we get to those levels of relationship and we start opening up, we can have intelligent conversations to say exactly what you said earlier:

“This element of this design is not really adding that much value to the customer that’s using it. So if it’s adding a lot of cost, let’s design it out together.” And we desperately need that input from our suppliers on design for manufacturing and assembly. That’s one of the key things that we’re teaching our buyers to do in technical negotiations. Not use it as a hammer, but as a collaborative tool. Say, “These are the three questions I need to go back and get answers from the design engineers.” Or, if best possible, get the design engineers in front of the manufacturing engineers at the supplier. That’s a strategic initiative for us.

Stephanie Feraday: That’s great. And this wasn’t on the script, but how difficult has it been to get your buyers to shift in that direction?

John Haupt: I would say we’re at the fuzzy front end of that transformation. In some cases, we’ve got buyers who have technical backgrounds and have that interest. In other cases, buyers came up from, “Hey, I’m a business professional. I don’t have that technical background.” Or even the desire. And that’s one of the things you have to address. Trust me, we can always use strong people no matter what their background is. We’ll get them in the right places. But I think that’s a key transformation. In my former life at Volvo, I can remember having a lot of engineers, especially in Brazil, who were in the purchasing team. Almost every single one of my buyers in Brazil had an engineering degree or background.

And they were more successful. So I think having a little bit of that technical background and interest is going to be the way of the future.

Stephanie Feraday: Great. Clearly we’re all talking about AI and how that plays. So maybe let me start with a general question. Can you characterize how your team is using AI and what they’re doing?

John Haupt: Yeah. We had a really good discussion last night with the executive team. AI is, I’m going to say it like this: our senior leadership is paying very close attention to it. We have our roadmap, which is representing probably about 10 percent of our activity within CNH. The other 90 percent is really grassroots, bottom-up. “I can use AI for this. I can develop this. I can do this faster. I can get to that intelligence layer if I do these types of things and use AI to do it.” That’s how we’re approaching it right now. And there are ways to improve, I think, in that space. But that’s what it looks like today.

Stephanie Feraday: Okay, great. Now we’ve been fortunate to work with your team on aiSource when it was in beta. And you’re also using Copilot and I think ChatGPT as well. What was the difference in terms of the data you were getting versus the generic LLM?

John Haupt: I think exactly as you described. We know that aiSource is based in the physical attributes. And it’s pulling directly from those. So that gives us a quantitative trust in that data. Where when you go to Copilot or probably ChatGPT, when we go to Copilot, it’s using the large language models and it’s pulling in that information, but it’s not really our data. It’s not our digital factory. So it’s not always trustworthy. We definitely see a big gap, for example, in India. When we’re looking at parts, what we get out of Copilot in India is not even really close to our models that we use.

Stephanie Feraday: Interesting. Let’s switch to aP Design. For those of you who are new, that’s the engineering product. And you piloted aiSource beta with buyers inside of CNH. So you had the engineering and the sourcing side. What convinced you that this was something you needed to do, looking at both sides?

John Haupt: I really wanted to equip the buyers with this technical data. Instead of the year-over-year, “Give me 1 percent, give me 3 percent,” we have to get to where the outliers are. I want to pay fair pricing for my components. I don’t want to overpay. So we have to use the physics-based technology to get to where we have good and fair costs from our suppliers. If we have that, good. Leave those suppliers alone. Put your focus on where the outliers are. And so that’s really where this collaboration with what we do in design, and how we equip the buyers to also participate in technical negotiations, that’s our path. That’s what we’re doing.

Stephanie Feraday: Maybe just a quick question then on the workflow in terms of when they’re negotiating with suppliers. Do you start with an existing benchmark, or how does all of that work?

John Haupt: Absolutely. We try to do as much benchmarking as we can upfront. We have lots of expert buyers. You’re going to meet Joe later today. He was one of them. We generally know when things are on track and off track. But the benchmarking only takes you so far. And you need to use the digital factories, especially when you know you’re going to change the footprint. Say, “Look, I have to avoid tariffs now, so I have to look at not sourcing things from China, India, but what does it look like internally within North America?” And what are those costs? Because when you add the tariff in on top coming from China, India, the business case doesn’t hold anymore.

So you still want to optimize your cost, your landed cost, into your factories.

Stephanie Feraday: Perfect. Okay, awesome. You mentioned building out and creating a cost culture at CNH. You’ve talked about the design and the sourcing side. So what does that mean across CNH? And what are the signs that you know the cost focus is now really becoming part of the culture?

John Haupt: It’s really interesting because I think the CNH leadership recognized that this was a gap. We had a good focus on quality, a good focus on delivery. We can still grow in both spaces, but cost was really lagging. And that was intentional when we did the platform relaunch. Structure really impacts culture in both ways. So getting the right leaders in place, the advanced purchasing leaders together with the cost managers in every platform, has been really transformational. And I think that’s indicative of the cost culture because the engineers know they’re going to get questions from the cost managers. And the buyers know they’re going to get questions from Advanced Purchasing. So that has been, I think, a really key element of growing our cost culture in CNH.

Stephanie Feraday: Fantastic. Okay, so last question. If someone in this room is standing where CNH stood a few years ago, skeptical, under margin pressure, not sure about AI, not sure about whether automation is more than hype, what’s the one thing that you would tell them?

John Haupt: If you’re new to aPriori or new to this space, I would say, for me, it’s been a great collaboration with the aPriori team. They are industry experts. And we want to leverage that industry expertise. The partnership has been fantastic. We’ve had dedicated individuals coming into CNH, looking at our processes, saying, “Hey, you can do things better if you put aPriori here instead of here.” Get started. Really, just jump in. Start to understand what it can do for you, and get some quick wins. There are absolutely quick wins. If you just start and identify an area where you suspect you have outliers, we’ll help find outliers quickly. Then go exploit that.

And when you start bringing millions back to your senior leaders, they’ll be like, “Yeah, we probably need to invest in this.”

Stephanie Feraday: Fantastic. Great. All right, John, thanks so much for coming up and sharing your experience. Really appreciate it.