Congratulations. You're a Manager Now.

AI is not only automating work. It is quietly turning more workers into supervisors of digital labor.

By DeQuan · Predicai

In one operating run, the AI systems working with me staged more than a hundred pieces of work.

None were sent.

That was intentional.

The systems could research, organize and prepare aggressively. The consequential action still belonged to a person.

In another run, approved copy regenerated several times. Every regeneration created another review cycle.

Then a handoff stalled because a constraint had traveled forward without being properly re-tested.

None of these incidents was dramatic. That is why I find them useful.

Once an AI system starts doing meaningful work, I stop merely “using AI.”

I set an objective. I provide context. I inspect output. I correct mistakes. I decide what is ready to move. I remain accountable for the outcome.

There is already a word for that kind of work.

Management.

Management is spreading beyond managers

BCG's 2026 AI at Work survey contains a remarkable number:

Forty-seven percent of respondents said they now spend more time managing and directing AI than doing the underlying work themselves.

Source: BCG, AI at Work 2026

Think about what that means for an individual contributor.

A salesperson who once conducted all the first-pass account research may now supervise research. A marketer may direct several systems producing content, analysis and campaign work. A software engineer may increasingly review work created by coding agents rather than typing every line manually. An analyst may spend less time gathering information and more time deciding whether an AI synthesis is reliable enough to use.

Their title does not contain the word manager. Their activity increasingly does.

Harvard Business Review put a name to part of this shift in February with an article arguing that companies will need agent managers, people responsible for orchestrating how AI agents perform, learn and operate safely alongside humans.

Source: Harvard Business Review, To Thrive in the AI Era, Companies Need Agent Managers

Gartner has framed the change as a move from a manager's traditional span of control toward a span of complexity. An org chart may still show six human reports while a hidden layer of digital agents operates underneath them.

Source: Gartner, Agentic AI Is Reshaping Manager Roles

And a September Business Horizons paper asks where humans should sit when AI agents can themselves brief and supervise other agents. The authors call the issue managerial altitude.

Source: Business Horizons, Managerial altitude

I am not going to turn that question into a full management framework here.

The more immediate shift is already big enough:

AI may reduce the amount of execution a person performs while increasing the amount of work they have to direct.

A human approval step can still be a badly designed system

The run where more than a hundred items were staged and nothing went out taught me something.

Human oversight is not automatically good simply because it exists.

If AI creates 100 outputs and a person now has to inspect all 100 at the same depth, we may have replaced a production bottleneck with a review bottleneck.

If approved material keeps regenerating and repeatedly requires the same compliance check, cheap generation has created expensive supervision.

The goal should not be to keep humans involved in as many steps as possible.

The goal should be to make human involvement meaningful.

That distinction becomes especially important as organizations adopt the phrase human in the loop.

The phrase sounds reassuring. It tells me almost nothing.

“A human reviewed it” is not enough

A 2026 FAccT study interviewed frontline supervisors working inside algorithmically managed environments.

These supervisors were formally responsible for overseeing algorithmic decisions.

But many lacked the information, discretion or authority needed to meaningfully challenge the systems they were supposedly overseeing.

The researchers describe an asymmetry between formal responsibility and effective control. Supervisors could remain accountable for outcomes they could not fully understand or contest.

Source: Oezdemir et al., Caught in the Loop

That is a much more useful way to think about human oversight.

The important question is not:

Is a human somewhere in the process?

It is:

Can that human actually exercise judgment?

If the person cannot understand why the system did something, challenge the result, stop the action or change what happens next, then “human in the loop” may be more ceremonial than operational.

The skills gap is starting to show up

A September IBM study surveyed 1,500 CHROs and 8,800 employees.

Seventy-one percent of CHROs said the ability to supervise, validate and override AI outputs was the workforce capability they considered most essential.

Only 29% of employees ranked judgment as important in the same way.

At the same time, 60% of employees said they were worried AI was eroding their skills, with critical thinking the capability most commonly cited as declining.

Source: IBM CHRO Study, September 2026

This is survey data, not a behavioral test. But the mismatch is worth taking seriously.

Organizations increasingly need people to become better supervisors of AI precisely as AI becomes good enough to tempt people into checking it less.

That is a management problem.

The manager still matters

Microsoft's 2026 Work Trend Index surveyed 20,000 AI users across ten countries.

The report found that organizational factors such as culture, manager support and talent practices had more than twice the statistical association with reported AI impact as individual factors, 67% versus 32% in Microsoft's model.

Microsoft explicitly labels the result as an association, not a causal effect.

Source: Microsoft Work Trend Index 2026

That caveat matters. The directional lesson is still useful.

Buying better AI does not eliminate the operating environment around the worker.

Someone still has to make it acceptable to experiment. Someone still has to establish expectations. Someone still has to decide what quality means. Someone has to notice when a workflow is generating activity instead of value. Someone has to coach the person whose job is changing underneath them.

Those are managerial functions.

AI for management and AI as management are different things

This distinction matters to me.

I am enthusiastic about AI for management.

Prepare the meeting. Summarize five project streams. Surface the unanswered question. Find the commitment nobody followed up on.

Remove administrative work so the manager can spend more time coaching, deciding, developing people and handling ambiguity.

AI as management raises different questions.

Continuously score the worker. Monitor every interaction. Allocate their work. Control opportunities. Generate evidence used in performance decisions.

A current example shows why the distinction is not academic.

Instructors at Multiverse told The Guardian in late September that an AI-supported teaching-review system made them feel heavily monitored and contributed to stress.

Multiverse said human managers review the system's flags, that AI does not make employment decisions and that performance management remains human-led.

Source: The Guardian, September 28, 2026

I cannot determine from public reporting exactly how the system operates in practice.

The tension itself is useful.

Both things can be true: a human can formally retain the final decision, and employees can still experience the surrounding system as algorithmic management.

So governance cannot end with the sentence:

A human makes the final call.

The way the information was generated, how contestable it is, how much surveillance is involved and whether the human reviewer has meaningful discretion all matter too.

Individual contributors are becoming managers of something

I do not know whether AI ultimately leads to fewer traditional managers across the economy.

There is not enough evidence to make that claim with confidence.

I am much more certain about a nearer-term change.

More people are being asked to direct digital labor.

They are assigning work, supplying context, evaluating outputs, correcting mistakes, escalating exceptions and taking accountability.

That does not make an AI agent an employee.

It does mean that managerial work is escaping the managerial job title.

My own staged-work run showed me the smallest version.

The systems did not need me to type every draft. They needed me to decide what was good enough, what required intervention and whether anything should leave.

The execution moved. The responsibility did not.

That may be one of the defining management shifts of the AI workplace.