Ford just spent three years and billions of dollars learning a lesson the hard way.
The automaker bet big on AI for vehicle design and quality control and jettisoned experienced engineers in the process. The result? A quality crisis. So Ford brought back 350 veteran engineers, the ones it calls “gray beards,” to fix what the AI missed.
The company assumed AI plus adjusted design requirements would produce a high-quality product. It didn’t. The AI was only as good as the information it was trained on, and much of that knowledge had walked out the door.
The fix worked. Ford topped J.D. Power’s mainstream quality ranking for the first time in 16 years, but only after rehiring the experts it had cut. Digital payment service Klarna made a similar correction after an AI assistant tanked customer service. IBM is tripling entry-level hiring in roles it once expected AI to absorb. Different industries, same lesson.
Here’s the pattern: AI can draft the message, summarize the meeting, search the knowledge base, and suggest next steps. But someone still has to know whether its answers are right. Organizations can’t automate that part away.
As AI gets more embedded in daily work, the human role doesn’t disappear. It changes. People spend less time starting from scratch and more time deciding what happens next. That makes judgment more important, not less.
Capacity, not headcount
Nick Unger, founder of AlterityAI, says leaders often reach for AI because they see savings. He urges them to think capacity first: How can AI help people do more valuable work before you touch headcount?
Expertise still matters
AI can generate a polished answer fast. But it doesn’t know your organization’s history: which exception matters, what’s politically sensitive, and what simply will not work. Experienced employees, like Ford’s “gray beards,” do know.
The catch is that those employees may be wary of AI, especially if it feels like their expertise is being extracted or replaced. Unger notes that people who spent careers developing that expertise are naturally reluctant to hand it to a system. Ignore that and you get resistance, which can kill AI projects before they start.
More than a checkpoint
“Human-in-the-loop” too often means someone glances at AI output before it ships. This should start earlier: Who decides which use cases make sense, what data does AI touch, and what are the standards for output? Without that clarity, employees either over-trust AI or avoid it entirely. Ford landed in between and paid the price before course correcting.
It’s worth asking some questions upfront:
- Who owns the final answer?
- Who reviews high-risk work?
- Who decides an output is good enough?
- Who fixes the process when it breaks?
Where communicators come in
For comms, HR and change management teams, the job is to frame AI adoption as a partnership, not a replacement. Help leaders explain where judgment still matters and make space for questions. People engage with AI when they understand its purpose and know their expertise still shapes the ultimate outcome.
Ford didn’t stumble because AI wasn’t helpful. It stumbled because it got rid of the people who knew when AI was wrong. The future of AI at work will belong to the companies that understand the difference between speeding up work and removing the judgment that keeps it from going off course.
The O’Keefe Group helps organizations turn AI strategy into practical, responsible adoption. We bring together communications, change management, knowledge strategy and employee engagement to clarify how AI should be used, where human judgment matters and what people need to work confidently alongside it.
We can help your organization prepare people and content for AI. Let’s talk.
