Not long ago, the question around automation was straightforward: Can this be done? 

With AI, that’s no longer the constraint.

Much of the work we do—writing, analysis, synthesis, even early-stage decision-making—can now be automated, at least superficially well.

So the question shifts: What should we automate?

At Indeed, we spend a lot of time thinking about those questions. The decisions we make about AI aren’t just about speed or efficiency. They shape how people develop skills, make decisions, and ultimately do their jobs.

The Hidden Cost of Outsourcing Thinking

In the 1980s and 1990s, companies outsourced aggressively to move faster. It worked, until it didn’t.  Some companies lost the ability to innovate and adapt, finding that rebuilding those internal capabilities later was slow and expensive.

AI creates a similar, but more subtle risk. You aren’t just outsourcing production to a vendor; you’re outsourcing parts of your own thinking.

There’s strong evidence that people achieve real understanding by generating ideas themselves. Struggling through a problem, forming a point of view, making tradeoffs. That’s what creates expertise.

Because AI outputs are often good, they create an illusion that the thinking has already happened. Over time, that leads to cognitive atrophy: weaker judgment and more generic work.

Used well, AI clears away noise, surfaces insights faster, and gives teams the space to think more strategically.

The “Can vs. Should” Distinction in Practice

Inside companies, this isn’t theoretical. It shows up quickly once you look at how work is actually done.

At Indeed, we’ve found success by getting explicit about what work requires judgment and what doesn’t.

Take our Legal team. They classified 100% of their work at the point of ticket closure. About 81% of the work required expert judgment. The remaining 19% did not. That 19% is where AI had immediate impact. One workflow that used to take 26 hours now takes about two.

The point is not just the time savings. It’s that the team didn’t try to automate everything. They were deliberate about protecting the work that requires and depends on expertise, while aggressively compressing the work that doesn’t.

That’s the “can vs. should” distinction in practice.

Why This Matters Now 

This intentionality is no longer optional. AI adoption is colliding with a harsh labor reality: a shrinking global workforce. AI is no longer just a lever for growth; it’s a buffer against a talent supply we can no longer take for granted.

To stay competitive, you must ensure your team doesn’t lose their “sharpness” to the machine. You don’t need a complex framework to decide what to automate, but you do need to ask four questions:

  • If I stopped doing this, how hard would it be to rebuild the skill later?
  • Can I reliably tell if the output is correct?
  • Does this benefit from my specific context or judgment?
  • Does this create understanding I’d otherwise lose?

If the answers point toward low risk, automate aggressively. If not, stay closer to the work.

Use AI as a Critic, Not a Crutch

The most effective pattern we’ve seen is to shift where AI shows up in the process.

Don’t let it be the first thinker. Start by drafting the problem, outlining your approach, or forming an initial point of view. Then use AI to pressure test it. Ask what you’re missing, where the logic breaks down, what alternatives you haven’t considered.

This keeps you engaged in the parts of work that develop skill, while using AI where it’s strongest: speed, iteration, and breadth.

It also avoids a common failure mode. When AI provides the initial structure, people tend to anchor on it. Not because it’s always right, but because it’s already there.

The Real Opportunity

AI doesn’t just save time; it reallocates effort. Used well, it compresses low-value execution and creates space for deeper thinking. Used poorly, it removes the exact parts of work that build expertise.

At Indeed, the opportunity is not just faster workflows—it’s better ones. AI won’t make the choice between “can” and “should” for you. That remains the most important human work left.