Is AI deciding the route or the destination of work?

AI’s authority in choosing the route or the destination of work (image generated with Gemini)

When you use an app like Google Maps or Waze in the car, you hand it one decision and keep another for yourself. The destination is still yours. The route becomes its. You agree to turn right onto a street you would never pick on your own, because you trust that, to get where you want to go, it knows more about the roads and the traffic than you do.

This silent split between the destination and the route is one of the most useful ideas for understanding what artificial intelligence is doing inside companies.

Two kinds of authority

Every time someone tells someone else what to do at work, they are exercising one of two kinds of authority.

There is authority over the “what”, which defines the goal, the expected result, the direction. And there is authority over the “how”, which defines the method, the step by step, the way of doing things.

Authority over the “what” almost always helps, because it aligns everyone’s effort around a common result. Authority over the “how” is more delicate, and its effect depends entirely on who knows more about the concrete situation.

Who knows more: the method or the person doing the work?

When a method is imposed on whoever holds the specialized knowledge of the case, it tends to get in the way. This is the problem Friedrich Hayek pointed out decades ago: whoever decides the step by step from up top often knows less than whoever has their hands on the work.

But the opposite also exists. In domains where the method concentrates learning the person doing the work does not have, or where individual variation costs lives, standardizing the “how” is what keeps people safe. Examples: the checklist in aviation, the surgical protocol, the infection control routine, the compliance rule, and so on. In these cases, abandoning the method in the name of autonomy would be irresponsible.

The useful distinction, then, is this: authority over the method helps when it carries knowledge the person doing the work does not have, and hurts when it smothers knowledge only that person has. The same standardization that saves lives in the operating room destroys judgment in knowledge work. What changes from one case to the other is not the rule, it is who knows more.

Where AI is automating

Now put this distinction together with what artificial intelligence tools do day to day.

Much of what gets automated today is not the decision about the “what”. The goal is still defined by people. What AI takes over, most of the time, is authority over the “how”, exercised with a granularity no boss has ever had. It suggests the next sentence of the email, the order in which to handle tickets, the call script, the design of the piece, the snippet of code. Step by step, choice by choice, the method of the work is being written by a system.

This is neither good nor bad in itself. It is exactly the same double-edged sword as before, now sharpened and distributed into every task.

Where AI’s method carries knowledge the person does not have, the gain is real. A script that embeds best practices in customer service helps a beginner start off better. An automatic check that prevents a compliance error does what the checklist does in aviation.

Where AI’s method smothers knowledge only the person doing the work has, the cost is silent. The salesperson who has known that client for ten years starts following the script the system recommends and loses the conversation that would have closed the deal. The doctor who notices something outside the protocol hesitates to contradict the machine. The person who knew the right shortcut unlearns it, because they stopped deciding.

The authority AI is taking over

The most common mistake I see is treating AI adoption as a single yes-or-no decision. Either the company embraces the tool, or it distrusts it.

The better question is a different one, and it comes precisely from this distinction:

In each task you are about to automate, is AI taking authority over the “what” or over the “how”? And, if it is over the “how”, does the method it imposes carry knowledge the person does not have, or is it about to suppress knowledge only that person has?

This question cannot be answered for the whole company at once. It is answered task by task, and the answer changes depending on who does the work. For the newcomer who has not yet developed judgment, AI’s script is an important foundation. For the specialist who spent twenty years developing that judgment, the same script can be a leash.

Adopting AI well is, to a large extent, knowing how to tell these two cases apart before standardizing everyone’s method the same way.

Food for thought

• In the tasks where you already use AI, is it deciding the destination or the route? And had you noticed the difference?
• Where does the script the tool suggests carry knowledge your team does not have yet, and where is it starting to suppress knowledge that only the people doing the work have?
• If a year from now the most experienced people on your team only follow the step by step the system proposes, what will your company have stopped learning?