The AI Boomerang

There's an interesting trend emerging in the AI economy.

Companies invest in artificial intelligence to automate work, eliminate positions, and improve productivity.

Then something unexpected happens.

The technology performs the tasks it was designed to perform. But the organization discovers that completing those tasks isn't quite the same as doing the job.

So it starts bringing people back.

Some are calling it the AI Boomerang.

And I think it exposes a fundamental misunderstanding about what makes experienced professionals valuable.

When the Automation Doesn't Know What It Doesn't Know

Consider Ford.

Earlier this year, the company acknowledged that its increasing reliance on automated quality-control systems hadn't produced the results it expected.

Ford subsequently hired or promoted approximately 350 experienced engineers to strengthen its quality processes.

Some were returning employees. Others came from suppliers or moved into new responsibilities.

These weren't simply people being hired to perform the same repetitive inspections AI was supposed to automate.

They brought technical judgment, experience recognizing defects, knowledge of manufacturing processes, and the ability to anticipate failures before they reached the production line.

Ford wasn't abandoning artificial intelligence.

It was discovering the importance of combining technology with experienced engineering expertise.

The lesson isn't that AI failed. It's that automation without sufficient expertise can fail to deliver the outcome the business actually needs.

And Ford isn't the only organization reconsidering its assumptions.

The Return of the Human Expert

Recent Associated Press reporting describes organizations that have eliminated positions during AI-related restructuring, only to discover that their automated systems still require substantial human oversight.

Klarna, for example, resumed recruiting human customer-service representatives after acknowledging that its emphasis on AI-driven cost reductions had gone too far.

Research involving hiring managers has identified similar situations across finance, technology, healthcare, and other industries.

The underlying issue is straightforward.

A job description doesn't capture everything a person actually contributes to an organization.

An experienced employee doesn't simply complete assigned tasks.

They recognize unusual circumstances.

They know which problems deserve attention.

They understand why a process exists, not just how it works.

They recognize when information is incomplete, when an exception matters, and when the technically correct answer is operationally wrong.

And much of that knowledge is never documented.

It's acquired through years of experience, mistakes, relationships, and exposure to situations that don't fit neatly into a procedure.

That is extraordinarily difficult to capture in an automation project.

Even the Robots Need Teachers

Here's another example that caught my attention.

China's robotics industry is investing billions in developing humanoid machines capable of performing physical work.

But making a robot walk, lift an object, or demonstrate an impressive movement is different from teaching it to operate reliably in an unpredictable workplace.

Reuters recently reported that robotics companies are hiring people to demonstrate physical tasks while wearing motion-capture equipment.

One company, Spirit AI, has assembled a team of approximately 1,000 people to collect the human demonstration data needed to train its robotic systems.

Think about that for a moment.

We're employing humans to teach machines how to perform human work.

And the challenge doesn't end once the robot learns a movement.

Someone still needs to understand the environment, evaluate performance, identify dangerous exceptions, troubleshoot failures, and decide whether the machine is performing the task correctly.

The technology may change who performs the work.

It doesn't necessarily eliminate the need for people who understand the work.

There's a Difference Between Automating Tasks and Replacing Expertise

I think this distinction is being overlooked in many conversations about AI.

A job is rarely one task.

It's a collection of activities, decisions, relationships, responsibilities, and accumulated knowledge.

Some of those elements are excellent candidates for automation.

Others require judgment, accountability, or deep contextual understanding.

For example, AI can summarize candidate resumes, identify skills, prepare outreach, and coordinate scheduling.

Those are valuable capabilities.

But deciding whether someone can succeed in a complicated technical role may require understanding the engineering environment, the hiring manager's actual needs, the team's capabilities, and the candidate's experience beyond a list of keywords.

And if the hiring process itself is poorly designed, AI may simply make that poor process operate faster.

Automating an ineffective process doesn't make it effective. It makes it faster.

The same principle applies in manufacturing, engineering, finance, logistics, and virtually every other profession.

💡 Brett's Perspective: AI Needs Experts, Not Just Engineers

I've spent a considerable amount of time building and experimenting with AI systems, connecting applications, developing workflows, and figuring out where AI agents can add genuine value.

And the more I work with these systems, the more convinced I become of something I've written about before.

AI needs experts, not just engineers.

That's not a criticism of engineers.

We absolutely need people who can develop models, build infrastructure, write software, and design reliable systems.

But we also need experienced professionals who understand the industries and processes those systems are supposed to improve.

A manufacturing engineer understands why a production process fails.

A supply-chain professional recognizes how a small disruption can affect an entire operation.

A recruiter understands that identifying a candidate and evaluating that person's potential are not the same activity.

And an experienced business leader knows that the most efficient decision isn't always the best decision.

The people closest to the work should have a meaningful role in deciding how AI is introduced, what it should accomplish, and how success is measured.

That doesn't mean every professional needs to become a software developer.

It means more professionals need to become active participants in designing the technology-enabled future of their work.

The Workforce Strategy Problem

There's another implication I think organizations need to consider.

If companies eliminate experienced employees before understanding the expertise those employees contribute, rebuilding that knowledge can be expensive.

And sometimes it may not be possible.

Recruiting replacements takes time.

Training requires experienced mentors.

Institutional knowledge doesn't automatically return when a requisition opens.

We should also be thinking about the next generation of professionals.

If AI handles much of the entry-level work that historically helped employees develop judgment and experience, how will organizations produce tomorrow's technical experts, managers, and leaders?

That is a workforce-planning question.

Not simply an automation question.

And it's one I believe Talent Acquisition should be helping business leaders answer before those decisions are made.

The Next Competitive Advantage

I don't believe the AI Boomerang means organizations should stop automating work.

Quite the opposite.

AI is becoming too capable and too useful to ignore.

But organizations need to become more deliberate about how they combine technology with human capabilities.

Before eliminating a role, leadership should understand which responsibilities can genuinely be automated, which require oversight, and which depend on expertise that may not be obvious.

Before deploying an agent, someone should understand the process well enough to establish boundaries, evaluate outcomes, and recognize failure.

And before celebrating productivity improvements, organizations should make sure they're measuring business results rather than simply the volume of automated activity.

Because the next competitive advantage may not belong to the company that automates the most jobs.

It may belong to the company that understands its people and processes well enough to automate the right work.

One Final Thought

We're entering a period when more professionals will be able to build useful AI systems without traditional software engineering backgrounds.

I believe that's a tremendous opportunity.

But knowing how to connect an AI agent to a workflow isn't the same as understanding how that workflow should operate.

The technology is becoming more accessible.

The expertise required to use it responsibly is becoming more valuable.

And perhaps that's the real lesson of the AI Boomerang.

You can automate a task.

You can redesign a process.

You can even eliminate a position.

But if you eliminate the knowledge that made the work successful, don't be surprised when it comes back around.

Brett Dyess | Brett's Field Notes

Exploring Talent Intelligence, artificial intelligence, workforce strategy, and the changing relationship between people and technology.