Ford: When Experience Walked Out and AI Walked In

In this episode, CD breaks down Ford's recent admission: after replacing hundreds of veteran engineers with AI-driven quality control, the company had to rehire 350 of them. Drawing on 25+ years of leadership experience inside Fortune 500 companies and the federal government, CD unpacks the real story behind Ford's AI stumble — and it's not a technology failure, it's a leadership one.

Contact: cdeerw71@gmail.com

In this episode:

• Why Ford brought back 350 veteran engineers after trusting AI to handle quality control
• The critical difference between information and judgment — and why AI can have one without the other
• Three leadership mistakes that led to Ford's stumble:

1. Confusing information with judgment
2. Treating expertise like data instead of a capability
3. Failing to intentionally design the human-AI partnership

• Three leadership fixes every organization needs:

1. Protect expertise before you automate around it
2. Don't assume AI is objective — question its assumptions
3. Design the human-AI partnership instead of hoping it happens naturally

• A personal story from CD's career estimating software costs, and how it shaped his view on blindly trusting data-driven tools

Key takeaway: Technology doesn't create organizational intelligence — leaders do. You can copy a process, but you can't copy a person.

CD Rogers-Wright

cdeerw71@gmail.com

Transcript

SPEAKER_00 0:01

Real situations, honest analysis, and a fix leaders can actually use. Today's question: What happens when organizations automate expertise before they understand it? Welcome back to CD's Take, Situations and Solutions. I'm Cidi. I've spent over 25 years leading inside Fortune 500 corporations and the US federal government. I've led teams, navigated crises, and watched organizations make decisions that made them better and decisions that hurt them. I'm currently completing a doctorate in organization and management because I want to understand not just what happens in organizations, but why. So, all of that experience is what you'll get from every episode. Real situations, honest analysis, and a fix leaders can actually use. Fellow fixers, today we're talking about the Ford Corporation. Last week, Ford admitted something many organizations are quietly discovering. They rehired the people they thought they didn't need anymore. Some had retired, some had gone to suppliers. Ford brought them back. Why, you ask? Because the AI that was expected to manage quality control on its own could not. And here's the interesting part. Ford is not alone. Across industries, companies like Klana, IBM, Salesforce, they all rushed to replace experienced people with AI, only to discover they had also replaced judgment, institutional knowledge, and decades of hard-earned expertise. As always, I'm going to tell the leadership story, not the story about AI failing. So today's story is going to be about understanding the difference between information and expertise, between automation and judgment, and then between efficiency and wisdom. So let's go back to today's question. What happens when organizations automate expertise before they understand it? Let's get into it. So here's the situation. Over the last three years, Ford quietly brought back 350 veteran engineers. Last week, two things happened at the same time. Ford topped the JD Power initial quality study for the first time in 16 years. And for the first time, Ford's leadership publicly explained why. Ford's vice president of Vehicle Hardware Engineering, Mr. Charles Poon, put it plainly, and this is me quoting him. Mistakenly, we thought that by introducing artificial intelligence and ingesting the design requirements that we had, that would produce a high-quality product. Let that sit for a second. That's a company admitting on the record that it believed feeding data into a system was the same thing as transferring decades of engineering judgment. But it was not. And here's why. The AI that was introduced never had decades of engineering judgment to learn from. What it did have was design requirements, specifications, and historical data. But it lacked the lived experience of the engineers who had worked through multiple product cycles. That experience turned out to matter more than Ford Al Hotler later explained that the company had been relying on a find and fix approach. That is, they were catching the problems after they showed up on the plant floor instead of preventing them before they happened. The fix wasn't more AI, it was people. So today, those veteran engineers lead design reviews, identifying potential failures before a single part is built. And it worked. But Ford is not the real story. Ford is just our case study for today because this isn't happening only at Ford, Klana. They publicly embraced an AI-first customer service strategy, only to later acknowledge that customers still valued human interaction more than expected and began bringing people back into the experience. Across industries, organizations are recalibrating their assumption about what AI can do and what still requires human judgment. Some analysts are calling it the AI boomerang. Organizations rush to replace expertise with automation only to discover that expertise is much harder to automate than they imagined. So the real question isn't why Ford's AI struggled. The real question is why so many organizations made the same bet at the same time and what they misunderstood about expertise in the first place. That's what we need to break down. All right, let's break this down. Because what happened at Ford looks like a technology story on the surface. AI wasn't good enough yet. Simple. Except it's not. What happened at Ford wasn't primarily a technology failure, it was a leadership failure. There are three things going on here, and none of them are really about the technology. And let's break them down. Point one, they confused information with judgment. Ford had design requirements, they had specs, they had engineering documents, they had historical data, etc. etc. What it didn't have though, what AI was never really given, was the thing veteran engineers carried every day, not facts, judgment. The instincts that come from seeing hundreds of small failures over a career, knowing before anything breaks, where the next one is likely to happen. Researchers call this tacit knowledge. It's the knowledge that's difficult to write down. It's because it's developed through experience, through observation, coaching, failure, and repetition. And here's what decades of organizational research tells us. Knowledge loss is rarely a technology problem. It's a leadership and organizational design problem. You cannot upload judgment. You can document information, you can archive specifications, you can preserve procedures, but judgment lives in conversations. It develops through mentoring, it grows through experience. And research consistently shows that technology preserves expertise best when it supports dialogue, not when it simply stores information alone. Email, for example, which is so widely used, has repeatedly been identified as one of the poorest tools for transferring tactic knowledge. Static documents aren't any better either. So what Ford did was it ingested design requirements. That's the equivalent of an email. Information moved, but judgment did not. And again, information is not judgment. Point two, they treated expertise like data instead of a capability. This is the part that should make every leader uncomfortable, honestly. The problem wasn't that Ford built a bad AI system. The problem was that Ford treated expertise as something that could simply be uploaded into one. It cannot. Expertise is in a database. It's a capability. A capability that develops over years through your mistakes, through coaching, through observation, experimentation. By the time Ford realized what was missing, many of the people who held that capability had already retired or moved on. And here's why that gap went unnoticed. Research on automation bias tells us something important. Organizations do not simply overtrust AI after it's implemented. That bias often begins much, much earlier. Leaders assume the technology is objective because it's data-driven. But objective based on what? Or built on whose experience, whose judgment, whose assumptions. Those questions often never get asked. Ford didn't ask them early enough. By the time they did, the people who could have answered them had already worked out the door. This actually reminded me of something from my own career. Years ago, I led a team of software cost estimators. We used one of the industry's leading commercial estimating tools. It was sophisticated, data-driven, and highly respected in the industry. But we kept finding that its recommendations did not fit the systems we were estimating. Eventually we discovered why. Much of the historical data behind the model came from large aerospace and defense programs. The issue is we weren't estimating spacecraft. We were estimating enterprise government systems. The tool was not broken. It was incredibly intelligent. About the wrong experience, though, in our context, that information changed how we used it. We stopped treating the model as the answer. It became just one input into expert judgment. And I think that's exactly the lesson Ford learned. Point three, they failed to design the partnership. This is the part almost nobody is talking about. There's now a substantial body of research examining when humans and AI actually outperform either one working alone. The findings are surprising. On average, simply combining people and AI doesn't automatically produce better results. In many decision-making tasks, it actually performs worse. Why? Because complementarity doesn't happen by accident. It has to be designed. Someone has to decide what should AI do, what should people do, who makes the final judgment, who challenges the recommendation, who teaches the system. Quality control isn't simply an information problem, it's a judgment problem. Ford didn't build a partnership between AI and experienced engineers. Initially, it built a replacement. Two completely different leadership decisions, right? Technology doesn't create organizational intelligence. Leaders do. Because leaders decide how human judgment and technology work together. Ford eventually recognize that. The veteran engineers were not brought back to compete with AI. They were brought back to teach the organization what AI couldn't yet know. And that's a fundamentally different model. So there are three leadership mistakes we can learn here. They confused information with judgment, they treated expertise like data instead of a capability, and they assumed partnership instead of designing it. So what is the fix? Alright, fellow fixers, here's the fix. And before I give it to you, I have to say the Ford Story isn't really about AI. It's about leadership. It's about what happens when organizations mistake information for expertise and technology for judgment. The good news is those are leadership decisions, which means leaders can make different ones. So here are three leadership decisions that are proposed that need to be made. Let's go. Fix one. Protect expertise before you automate around it. Before you build, before you deploy, before you trust any system to handle something that used to require a person's experience, ask one question. Whose judgment is this built on? If the answer is nobody's, if it's just specs and data and documents, you don't have expertise, you have information. The research is clear on what actually transfers judgment. Judgment is transferred through mentoring, shadowing, conversation. So it looks like your senior people working alongside the people who replace them. If your most experienced people are heading towards the door through retirement, attrition, buyout, and you haven't sat down with them, recorded their thinking or reasoning, how they make decisions, you're not protecting institutional knowledge. Fix two, don't assume AI is objective. Question assumptions before you trust the answers. So Ford trusted a system because it felt neutral. Data in, answer out. But every AI system is built on assumptions. Someone somewhere decided what data mattered, what good looked like, what history the system will learn from. So before you trust an AI recommendation, before you lean on AI for something that really matters, like a hiring decision, a quality call, a strategic recommendation, ask a better question. What expertise is the system actually drawing from? Not what it claims to do, but what was it built on? And then build in real verification, not a rubber stamp, actual scrutiny, especially in the moments where judgment matters most. Because confidence is not the same as competence, neither is a system. Fix three: don't assume complementarity. Design the partnership. One of the biggest misconceptions about AI is that people and technology naturally work better together. The research says otherwise. Complementarity doesn't happen automatically. It has to be designed. Leaders have to decide what should AI do, what should people do, who makes the final judgment, who challenges the recommendation, who teaches the system. AI is exceptional at speed, pattern recognition, and scale. Human beings are exceptional at judgment, context, wisdom, and knowing when something simply doesn't look right. Leaders should get the division of labor right intentionally. Let AI handle what it actually is good at and then let people do what they excel at. Ford eventually recognized that. The veteran engineers weren't brought back to compete with AI. They were brought back to teach the organization what AI couldn't yet know. So there you have it. Three leadership decisions. Protect expertise before you automate it, question the assumptions before you trust the answers, and design the partnership. Because information is not judgment. Expertise is a capability, it's not a database. And technology does not create organizational intelligence. Leaders do. That's always the fix. So, what happens when you automate expertise before you understand it? You get the Ford story. Billions in cost, recalls, a quiet, expensive scramble to bring back the very people they thought they did not need. But underneath the leadership lesson, though, there's something deeper. Technology can store information. It can move fast, it can find patterns and scale further than any human ever could. What technology cannot do is to replicate the thing that made those 350 engineers irreplaceable in the first place. That's judgment, wisdom, and discernment. For me, that points to something bigger than organizational design. I personally believe humans were created intentionally in the image of God with the capacity for wisdom, discernment, and judgment. Scripture says we are fearfully and wonderfully made. That's more than a statement of faith. For me, it's a leadership principle. Because if human beings were designed with those capacities, then our goal as leaders is to steward and develop them. You can copy a process, you cannot copy a person. That is true on a factory floor in Michigan, it's true in a hospital, a courtroom, a boardroom, wherever real judgment is required. The human in the room is not a placeholder for a system that isn't ready yet. They're the point. So, fellow fixers, protect expertise before you automate it. Question the assumptions before you trust the answers. Design the partnership instead of hoping for one. Because technology does not create organizational intelligence. Leaders do. And that's always the fix. I'm CD. This has been CD's Take Situations and Solutions. If you'd like to continue the conversation, my information is in the show notes. If today's episode resonated with you, please share it with the leader who needs it. So, fellow fixers, I'll see you in the next episode. Thank you for listening.