The Technology Changed. Management Didn’t.

The Amazon story about an AI project quietly spending nearly two million dollars generated predictable headlines. Most of them focused on artificial intelligence.

I found myself thinking about management.

Part of the story involved something employees jokingly called “tokenmaxxing.” Managers wanted AI adoption, so token consumption became a visible measure of progress. Employees responded exactly as people have always responded to measurable targets. They consumed more tokens. The metric improved. Whether the business improved was another question entirely.

None of this surprised me.

During my career I watched organizations repeatedly create measurements because they were easy to collect. Before long, those measurements quietly became the objective. Teams worked hard to improve the numbers on the dashboard while actual progress toward the original goal slowed, stopped, or even moved backward.

The project looked healthy.

The objective quietly disappeared.

That isn’t an AI problem.

It’s a management problem.

Long before AI, managers learned that successful organizations begin with clearly defined objectives. Those objectives should be specific enough that everyone understands them, measurable enough that progress can be assessed, achievable enough to motivate rather than discourage, relevant to the organization’s purpose, and tied to an expected timeframe. Most of us recognize those ideas today, but the underlying principles are much older.

The point wasn’t to create clever measurements.

The point was to make sure the measurements actually reflected progress toward the objective.

Somewhere along the way, organizations often forget the difference.

Measure lines of code and programmers write more code.

Measure meetings and calendars become full.

Measure reports and reports become longer.

Measure AI token consumption and employees find reasons to consume more AI tokens.

People optimize what management measures, not necessarily what management intended.

That is exactly what happened with Amazon. Management wanted AI adoption. Token usage became the visible measure. Employees optimized token usage because that was the signal the organization chose to send. The company got more of what it measured—but not necessarily more of what it actually wanted and it consumed a lot of cash doing it..

As I read the story, I found myself thinking about Peter Drucker.

Not because Drucker understood artificial intelligence.

He didn’t and couldn’t.

He understood organizations.

He spent decades asking questions that still matter today.

What are we trying to accomplish?

How will we know if we’ve succeeded?

What results actually matter?

Those questions have not become obsolete simply because some of today’s workers happen to be software.

In fact, they have become more important.

Artificial intelligence is remarkably efficient.

It will happily optimize almost anything we ask it to optimize.

The problem is that organizations sometimes ask it to optimize the wrong thing.

One of the advantages of living through several waves of technology is recognizing that this pattern repeats itself.

I watched it happen with mainframes.

Then personal computers.

Then the Internet.

Then cloud computing.

Now AI.

Each new technology arrived with predictions that everything had changed. Some things had.

Many hadn’t.

Organizations still need clear objectives.

They still need governance.

They still need accountability.

They still need meaningful measurements instead of convenient ones.

Today’s executives rarely reject the ideas of Peter Drucker or other great management thinkers. More often, they simply never encounter them. Management, like technology, has fashions. New frameworks replace old ones. New terminology replaces familiar concepts. It becomes easy to assume that decades-old thinking no longer applies.

Yet much of that earlier work was never about technology.

It was about people.

It was about organizations.

It was about incentives.

It was about deciding what success actually looks like before deciding how to measure it.

Artificial intelligence may be the most important technology of our generation.

That doesn’t mean it requires us to abandon everything we have learned about good management.

If anything, it reminds us why those lessons were worth learning in the first place.

If anything, it has reminded us why they mattered in the first place.

Read original documentabput Amazon here:  https://exploringchatgpt.substack.com/p/aaaand-its-gone?utm_source=post-email-title&publication_id=1272495&post_id=209135814&utm_campaign=email-post-title&isFreemail=false&r=jusf8&triedRedirect=true&utm_medium=email