One of the biggest questions surrounding artificial intelligence right now is also probably the wrong one:

Where can we add AI?

After attending Ai4 in Las Vegas, I came home thinking we should be asking something different:
‍If we were designing this work today, knowing what AI can do, would we design it the same way?

That’s a much harder question, but it’s also a much more interesting one because it moves the conversation beyond simply automating the work we’ve always done.

‍

Don’t Put New Technology on Old Problems

Most organizations have processes that exist because, at some point, they made sense. A report gets created because someone has always created the report. Information passes through three people because that’s how the approval process developed. Someone manually moves information from one place to another because, until recently, there wasn’t an easy alternative.

Then AI comes along and our first instinct is to automate those steps.

But what if some of those steps shouldn’t exist at all?

One warning at Ai4 stuck with me: AI can make broken processes fail faster. Speed doesn’t create value by itself, and automating ten unnecessary steps still leaves you with ten unnecessary steps.

Before asking how AI can perform a task faster, we should ask why the task exists, what outcome it’s supposed to create and whether there’s now a completely different way to get there.

That’s the difference between simply using AI and actually redesigning work around AI.

‍

Think With AI, Not Just About AI

There’s been a lot of attention paid to prompt writing, and knowing how to communicate with an AI model certainly matters. But one of the bigger mindset shifts discussed at Ai4 was the need to teach people to think with AI, not simply prompt it.

That means understanding AI well enough that it changes how you approach a problem in the first place.

Take a marketing assignment. The obvious use of AI might be to ask, Can you write five social media posts? That’s useful, but it’s only scratching the surface.

We can also ask: What questions is our audience likely to have after seeing this campaign? What assumptions are we making about this customer? Where are the gaps in this strategy? What patterns exist across our past campaigns that we haven’t noticed? What would someone who disagrees with this recommendation say? What information would change our decision?

Now AI isn’t simply completing a task. It’s participating in the thinking that happens around the task.

That shift becomes even more important when we consider what happens to the time AI saves us. If a four-hour task suddenly takes 30 minutes, that’s great. But if the only goal is squeezing more tasks into the remaining three and a half hours, we’ve missed part of the opportunity.

What if some of that time goes toward deeper strategy, more collaboration, better client relationships, education, experimentation or creative thinking?

The future of work isn’t necessarily about removing humans from the process. It’s about redesigning work around where human value matters most.

‍

How Much Autonomy Should AI Have?

Redesigning work around AI also doesn’t mean making everything autonomous.

There is enormous excitement around AI agents, but not every business problem needs an agent. Sometimes AI should create a draft and wait for a human to approve it. Sometimes it can take an action after receiving preapproval. In lower-risk, well-defined situations, it may eventually be appropriate for AI to act independently and report back afterward.

One framework at Ai4 described these levels almost like members of a team:

Intern: AI creates the draft; a human approves it.

Employee: AI can take an action within predetermined approvals.

Manager: AI acts independently within defined boundaries and reports afterward.

The appropriate level depends on the work, the risk and how much judgment is required. The goal shouldn’t be maximum autonomy for its own sake. It should be using the simplest, most cost-effective solution that reliably does the job while keeping human oversight where it adds value.

In other words, don’t build a robot employee because a really good template would solve the problem.

‍

A Better Question for the Future of Work

AI is moving quickly enough that some workflows we built even a few years ago deserve another look—not because they’re bad, but because the assumptions they were built around have changed.

That’s why one of the biggest things I’m bringing home from Ai4 isn’t a particular platform, model or tool. It’s a question I think is worth asking over and over:

Knowing what we know now, what should this work look like?

Not Where can we squeeze AI into the way we’ve always done things?

Instead: If we started over today, what would we keep, what would we change, what would technology handle and where would people matter most?

That’s a much bigger conversation than automation. And it’s probably where the real opportunity begins.

‍

Frequently Asked Questions About AI and the Future of Work

How can businesses identify good opportunities for AI?

Start with the business problem rather than the AI tool. Look for measurable processes that can be improved, repetitive work that consumes employee capacity, large amounts of information that are difficult for people to analyze manually, or decisions where AI could provide useful recommendations.

Should businesses automate existing workflows with AI?

Not automatically. Before automating a workflow, businesses should determine whether the workflow itself still makes sense. AI may create an opportunity to simplify, eliminate or completely redesign steps rather than simply performing the existing process faster.

What does it mean to “think with AI”?

Thinking with AI means using it as a thought partner rather than only a task-completion tool. AI can help challenge assumptions, explore alternatives, identify questions, analyze information and expose potential gaps before humans make a final decision.

Will AI agents replace traditional business workflows?

AI agents may become useful for some workflows, but full autonomy isn’t necessary for every task. The appropriate level of AI autonomy depends on the complexity, risk, cost and need for human judgment.

How should businesses use the time AI saves?

Businesses can reinvest time saved through AI into higher-value activities such as strategic thinking, creativity, customer relationships, employee development, problem-solving and innovation rather than measuring success solely by the number of additional tasks completed.