ChatGPT WORK Begs the Question: What Work Can I Give to AI?

For the last few years, most conversations about generative AI have started with some version of the same question: What should I ask it?

So, we learned to prompt.

We built prompt libraries. We experimented with custom GPTs. We asked AI to draft emails, summarize documents, brainstorm campaign ideas and rewrite copy.

All useful. I do these things, too.

But I think we’re entering a much more interesting phase of AI adoption.

The question is shifting from “What should I ask AI?” to “What work can I give AI?”

That may sound like a subtle difference. It isn’t.

We’re moving from prompting to working

ChatGPT Work is a good example of where this is headed.

Instead of sitting in a chat window and completing one AI-assisted task at a time, Work can take on more substantial, multi-step assignments. It can research, analyze information, work across files and connected applications, navigate web-based processes and produce finished outputs.

In other words, AI doesn’t necessarily have to sit on the sidelines waiting for us to feed it the next prompt.

We can begin giving it a defined role inside the workflow.

For marketing and communications teams, I think that’s far more consequential than the latest announcement that a model is 17% better at some benchmark most of us have never heard of.

Because this is about the actual work.

Think about a product announcement

I’ve worked in marketing and communications for 30 years, so I know what these processes actually look like.

They rarely involve one person creating one piece of content.

A product announcement might require a press release. Executive talking points. An employee announcement. Customer communications. Social content. Media research. FAQs. Maybe sales materials. Probably a tracker.

And, of course, reviews and approvals.

Lots of reviews and approvals.

Today, a team might already use AI throughout that process.

• Someone asks ChatGPT to draft the press release.
• Someone else uses it for social copy.
• Another person develops executive talking points.
• Internal communications works on the employee announcement.
• Marketing drafts the customer email.
• Someone researches reporters.
• Someone else updates the project tracker.
• AI may touch nearly every part of the process.

But the people are still doing all the orchestration.

They’re finding the information. Moving it from system to system. Re-explaining the context. Managing the handoffs. Checking whether version six of the executive talking points still says the same thing as version four of the press release.

That’s the part I think is about to change.

Imagine giving AI a broader assignment:

Review the approved product information and existing brand materials. Research the market context. Prepare first drafts of the press release, executive talking points, employee communication, customer email and social content. Check the materials for inconsistencies. Flag any claims that need substantiation. Then organize everything for the appropriate people to review.

Now we’re doing something different.

That’s not just content generation.

That’s workflow execution.

And that’s where AI gets much more interesting to me.

We have spent a lot of time teaching people to use AI

We needed to. But that isn’t enough anymore.

Organizations now need to understand how AI fits into the work people already do.

That’s why at Human Driven AI, we spend so much time looking at workflows.

Before we add AI, we map the work.

What starts the process? What information is needed? What systems are involved? Where are people losing time? Where are the unnecessary handoffs? Where does human judgment really matter?

Then we look at where AI belongs.

And equally important: where it doesn’t.

Our Human + AI Workflow Blueprints™ are designed around this idea because simply inserting AI into an existing process doesn’t magically make it a good process.

Sometimes all you’ve done is automate the mess.

Faster isn’t necessarily better if you’re doing the wrong thing faster.

AI needs a job description

This is one of the biggest changes I think organizations need to make in how they approach AI.

Stop telling employees to “use AI more.”

What does that even mean?

Instead, define what you want AI to do.

• Maybe AI gathers the information.
• It could conduct the initial research.
• Perhaps it compares information against established criteria.
• Or, it creates the first draft.
• Let’s say it checks five pieces of content to make sure they all say the same thing.
• Maybe it updates the tracker.
• Then, it monitors for a change and initiates the next step.
• Next, a person reviews the analysis, applies judgment, makes the decision or approves the final work.

Now AI has a role.

And so does the human.

That’s a much more useful operating model than giving everyone an AI license and hoping something productive happens.

It’s also a much better foundation for governance.

Once you know what AI is actually doing, you can make informed decisions about what information it should access, what it can do independently, where human approval is required and how its work should be evaluated.

The tech is moving faster than our operating models

ChatGPT Work is just one example.

Across ChatGPT, Claude and Gemini, we’re seeing the same general direction: connected systems, reusable skills, agents, scheduled work and AI that can take action rather than simply generate an answer.

The models will keep getting better.

I’m less convinced that’s going to be the biggest challenge for organizations.

Work design is.

As we say at Human Driven AI, Technology is easy to buy. Changing how people work is harder.

And that’s why I don’t think the organizations that get the greatest value from AI will necessarily be the ones with the most AI tools, the longest prompt libraries or the fanciest agents.

They’ll be the organizations that understand their own work.

So here’s what I’d do

Don’t start with your entire marketing department.

And please don’t start by announcing a massive “AI transformation.”

Pick one workflow.

Choose something your team does regularly. Ideally, something that takes too much time, involves too many manual handoffs or causes everyone to groan when it appears on the calendar again.

Map how it actually works today.

Not how the process document says it works.

How it really works.

Then look at each part and determine:

• What does a human need to own?
• Which parts could AI own?
• How could AI prepare for human review?
• What systems and information would AI need?
• Where is human judgment essential?
• How would actually make this workflow better?

Then test it.

You may discover AI can take on much more of the process than you expected.

You may also find places where adding AI makes absolutely no sense.

Good.

Both answers are valuable.

Because the goal isn’t to put AI everywhere.

The goal is to build a better way of working.

Tools like ChatGPT Work are giving us the ability to do that.

Now the harder, and much more interesting, work is figuring out what we should do with them.

As usual, we can help with this.

---

Human Driven AI helps organizations move from scattered AI experimentation to shared, scalable practice. We build the governance foundations, redesign workflows around human and AI strengths, and deliver custom training programs that teach your teams how to put those systems into practice.

Whether you need an executive Lunch-and-Learn, a hands-on offsite workshop, an enterprise AI transformation program or a strategy to strengthen your brand’s visibility through GEO, we help you turn AI capability into better ways of working.

Ready to move from using AI to working differently with it? Let’s talk.

Read more