Efficiency needs direction.

I remember seeing a post by a growth leader Leah Tharin before the AI slop problem really took hold: “I can't wait until LinkedIn is fully automated.”

She was joking. Mostly.

Because you could already see where things were heading.

AI made it possible to produce a polished LinkedIn post in seconds. The platform itself encouraged the behavior. LinkedIn introduced “Write with AI,” allowing users to give it their ideas and have AI turn them into a draft post.

Then everyone got very, very efficient.

LinkedIn became flooded with content. Perfectly formatted observations. Inspirational leadership lessons. Thought leadership assembled from other thought leadership. Comments that restated the post. Posts that sounded increasingly like every other post.

Eventually, users had enough. And now LinkedIn is responding.

The company says it is reducing distribution of generic AI-generated content, cracking down on automated comments and giving users a way to report content that “seems like AI slop.”

A strange place to land

There is something fascinating about the sequence.

That is a strange place for us all to find themselves.

Efficiency as a mandate

Nearly every team has heard some version of the same mandate. Use AI. Move faster. Do more with less. Increase output. Automate repetitive work. Figure out how your team can become more productive.

Content was an obvious place to experiment. AI could help research an idea, structure an argument, create a first draft, rewrite copy, repurpose an article, generate social posts and help an executive communicate more consistently — which is roughly the case for an AI bench.

The platforms were building AI tools directly into the experience. Using AI wasn't some fringe growth hack. It was becoming part of the workflow.

Now we're entering the next phase. The expectation of efficiency hasn't disappeared. The tolerance for visible AI has.

That's a difficult operating environment.

“This looks like AI” is becoming an accusation

LinkedIn's new reporting mechanism introduces another complication. Users can now flag something because it “seems like AI slop.”

Think about what that means culturally.

We don't have a reliable way for another person scrolling through LinkedIn to know exactly how a piece of writing was created.

Yet we've now created a social mechanism for declaring that something feels AI-generated. That can easily become a weapon.

Don't like someone's post? AI slop. Writing feels too polished? AI slop. Uses a structure we've come to associate with a chatbot? AI slop. Someone's writing style happens to resemble patterns AI learned from human writing? AI slop.

The irony is that AI detectors and human intuition are both imperfect at determining authorship. We risk moving from a legitimate conversation about quality into a much less useful conversation about purity.

The problem with perverse incentives

The real problem is the incentive structure.

When the cost of producing content approaches zero, the supply of content explodes. And when everyone is optimizing for the same platform signals, the content begins converging — the same failure mode as optimizing outbound for volume.

The same hooks. The same structures. The same cadence. The same manufactured vulnerability. The same five lessons. The same comments congratulating the author for sharing such an important perspective.

Eventually, the feed becomes less valuable.

That is the problem LinkedIn is trying to solve. And it should. A professional network depends on people believing there are actual people, ideas and expertise behind what they are reading.

The question is how we get there without punishing the experimentation that brought us here.

Experimentation shouldn't become a scarlet letter

It matters beyond LinkedIn.

We're in the middle of one of the largest changes to knowledge work in decades. Nobody has the operating manual.

We're learning where AI improves work and where it makes work worse. Where automation creates leverage and where human judgment becomes more important. What customers will accept, what employees will accept, what audiences value, and where the boundaries should be.

That requires experimentation.

Some experiments will be bad. Some predictably so. Some will create unintended consequences. Some practices that feel exciting today will look ridiculous three years from now. That's part of technological change.

Organizations need environments where people can learn without feeling that yesterday's encouraged behavior will become tomorrow's reputational offense.

This is why you need a North Star

AI policy alone can't solve this. The technology changes too quickly.

Platform rules change. Social norms change. Customer expectations change. Capabilities change. A list of approved and prohibited AI uses will continually struggle to keep up.

A North Star gives people something more durable.

For LinkedIn, the North Star might have been: does this help professionals learn from credible people and useful expertise?

Then “Write with AI” gets evaluated against that objective. Automation gets evaluated against it. The feed algorithm, the comments, the AI-slop response, all of it.

For a marketing organization, the North Star might be: does this help us create something genuinely useful for the customer and worthy of their trust?

Now the question changes.

The North Star makes many of these decisions surprisingly straightforward.

Efficiency needs direction

AI can make almost any process faster. That doesn't tell you whether the process deserves to be accelerated.

This is becoming one of the central management challenges of AI.

Every one of those capabilities sounds like productivity. Each can also create downstream consequences for someone else.

Efficiency measured at one point in a system can create inefficiency somewhere else. A North Star forces you to consider the whole system.

We are now in an awkward middle

We still have the mandate to become dramatically more productive. AI is becoming embedded in nearly every tool we use. Leadership expects marketing teams to understand it. Boards are asking about it. Budgets increasingly assume some productivity benefit from it.

At the same time, audiences are developing a stronger sensitivity to content that feels automated. Platforms are responding. Trust is becoming part of the equation.

So the marketer's job has evolved again.

We have to understand how to use AI deeply enough that the work doesn't become defined by the tool.

Those decisions require learned experience — the thing no agent can replicate.

We should be learning together

I don't think the answer to AI slop is making people afraid to use AI.

The goal should be responsible experimentation with a clear direction.

The companies that navigate this well won't be the ones that perfectly predict every acceptable use of AI. They'll be the ones that know what they're trying to preserve.

Their customer's trust. The quality of their thinking. The usefulness of their work. The integrity of their relationships. Their point of view. Their North Star.

Tools will change. Platforms will change their minds. The rules will continue moving.

A North Star gives us a shared basis for making decisions as we learn.