Everyone is trying to make content faster.

AI can write the first draft. It can summarize the interview. It can generate twenty hooks before you've finished your coffee.

But talk to almost anyone running a content operation and you'll hear a different problem.

They don't have enough time.

Not because writing takes too long.

Because everything around the writing does.

Someone needs to book the guest.

Someone needs to remember that the guest still hasn't sent a headshot.

The editor needs the raw files.

The producer needs to know whether the episode is actually approved.

The newsletter needs a draft.

The social team needs clips.

Someone needs to notice that Thursday's episode is supposed to publish tomorrow and three things are still missing.

Content teams have gotten very good at adding tools to this problem.

What they haven't added is capacity.

The invisible job behind every piece of content

Look at a finished podcast episode and you see an hour of audio.

Look behind it and you see an operation.

There is outreach, scheduling, intake, research, recording, editing, approvals, transcription, writing, distribution, promotion, and reporting.

And between all of those stages are smaller actions.

Send this.

Check that.

Move this file.

Remind this person.

Update this status.

Find that link.

Ask again.

Those little actions are easy to ignore because none of them feels substantial enough to be called a job.

Together, they are a job.

Sometimes several.

This is one reason content teams can feel overloaded even when the actual creative workload hasn't changed very much.

The work around the work keeps growing.

Project management solved a different problem

For the last decade, the answer to operational complexity has mostly been better organization.

Put the work on a board.

Assign an owner.

Give it a due date.

Create a template.

Add an automation.

This was a big improvement over email threads and spreadsheets.

But there is a limit to what organization can solve.

A task manager can show you that the guest intake form is overdue.

It can't make the guest fill it out.

A calendar can show you that an episode publishes Friday.

It can't prepare the transcript.

A project board can show you that the newsletter hasn't been started.

Someone still has to start it.

The software is tracking the shortage of capacity.

It isn't necessarily creating more of it.

That distinction is becoming much more important.

AI started in content at the wrong end of the workflow

Most content teams first encountered generative AI through a blank text box.

Write a blog post.

Give me ten titles.

Turn this transcript into a LinkedIn post.

That was useful, but it also created a strange workflow.

A person had to find the source material, open the AI tool, explain the task, review the response, copy the result somewhere else, update the project, and tell the next person that it was ready.

AI did one part of the task.

The human remained the operating system.

That's starting to change.

The more interesting use of AI isn't generating another piece of text.

It's giving software responsibility for a part of the operation.

An agent is more useful when it has a job

We've been thinking about this a lot while building Northflow.

There is a meaningful difference between asking AI to do something and giving an agent an ongoing responsibility.

Imagine a content team has a podcast guest coming up.

In the first model, someone on the team opens an AI assistant and asks it to write an outreach email.

Helpful.

But the producer still sends the email, tracks the response, follows up, collects the guest's information, attaches it to the episode, and tells everyone when the guest is confirmed.

In the second model, guest recruitment is a job.

An agent has responsibility for that part of the workflow.

The human decides who should be invited and remains in control of the process. But the repetitive operational work underneath that decision can start moving somewhere else.

That is a much bigger change than getting better copy from a chatbot.

It changes where the work lives.

The best work to give an agent is usually boring

There is a temptation to judge AI by the most impressive thing it can create.

Can it write an entire article?

Can it make the video?

Can it replace the strategist?

Those questions get attention.

But they may not be the most valuable questions for a content team.

A better one is:

What work does your team keep doing that nobody actually wants to spend their time doing?

Following up with someone for the third time.

Moving information from a form into a project.

Checking whether a file arrived.

Turning a finished recording into a transcript.

Preparing a first pass of supporting content.

Keeping track of what is ready and what isn't.

These tasks aren't glamorous.

That's exactly why they matter.

Every hour spent coordinating the process is an hour your producer, strategist, host, editor, or creator isn't spending on something that actually needs their judgment.

Humans are still the hard part

This doesn't mean content operations can run unattended.

If anything, agentic workflows make human judgment more obvious.

Someone still has to decide which guests are worth pursuing.

Someone needs to know whether an episode is good.

Someone has to recognize when a draft technically follows the instructions but completely misses the point.

Someone needs taste.

Someone needs context.

Someone needs to know when the plan should change.

This is also what teams actually using agents are discovering. Every, for example, has written about moving away from the idea that every employee simply needs a personal agent. Their experience was that agents need ownership, maintenance, and humans responsible for making sure the work is good.

That's an important distinction.

The interesting future isn't autonomous companies where software quietly does everything.

It's teams getting much better at deciding what deserves human attention.

Content operations is becoming a capacity problem

This changes the way we think about software.

The old question was:

Where should we manage this work?

The new question is:

How much of this work should a person still have to carry?

Those are very different product problems.

A useful content operations system still needs the basics. Teams need to see the work, understand what is late, keep assets attached to the right project, manage guests, coordinate schedules, and know what is publishing next.

Northflow does those things today across episodes, tasks, projects, files, forms, guest workflows, calendars, spaces, and templates.
But we're increasingly interested in what happens after the dashboard.

Once the system understands the work, can it help carry some of it?

That's where agents become much more useful.

Not as an AI feature sitting next to the workflow.

As part of the workflow itself.

Start by looking for operational drag

You don't need to redesign your entire content operation around agents tomorrow.

Start smaller.

Look at one week of work and pay attention to the things your team repeatedly has to push forward.

Where are people chasing information?

Where does somebody have to copy something from one system to another?

Where does work stop until someone remembers to check on it?

Where are highly skilled people spending time on low-judgment tasks?

Those are your capacity leaks.

And they are probably more important than finding another tool that makes the board look nicer.

The teams that figure this out won't necessarily be the ones generating the most content with AI.

They'll be the ones that learn how to divide work differently.

Humans deciding.

Humans creating.

Humans applying judgment.

And agents carrying more of the repetitive operational load around them.

Because the biggest opportunity for AI in content may not be helping us create more.

It may be giving us more room to do the work that was worth creating in the first place.