← All articles

David Roon · · 12 min

Rethinking Work in the Age of AI

Judgment, responsibility, and new ways of working together. A reflection on what becomes valuable when AI makes production easier.

The debate about artificial intelligence and work often revolves around one question: how many jobs will disappear? I wonder whether that question distracts us from a deeper transformation.

If AI dramatically lowers the cost of producing things, it could change why we organize ourselves into companies, how we choose the people we work with, and what we are willing to pay for.

People will still have projects, needs, and ambitions. But why dedicate someone full-time to one project if they can now take responsibility for several? Why buy a solution if you can create it yourself? And why try to do everything yourself simply because it has become possible?

My intuition is that value may shift toward two closely related abilities: choosing a direction and taking responsibility for that choice.

When action becomes easy, choosing becomes essential

A job does not exist solely because its tasks are useful. In organizations, some activities also preserve habits, make work visible, or reinforce the importance of a particular role. Work has a theatrical dimension.

Automating a task therefore does not explain, by itself, why a position would disappear. Its real purpose may lie elsewhere. AI could even amplify this activity: more presentations, reports, and projects, now easier to produce, without any clearer sense of their usefulness.

We have already experienced a similar shift with information. When it was less accessible, we had to know how to find it. Once it became abundant, we had to learn how to filter it and protect our attention.

The same could happen with action. We may have so many opportunities to do something that the challenge becomes deciding what deserves to be done.

A problem can be real, and its solution within reach, without it being worth addressing right now. Each proposal may seem reasonable on its own, while their accumulation takes us further away from what we wanted to achieve.

Abundant means make intention more important. We can do an enormous amount of work without ever really choosing a direction.

The value of a point of view

In my experience, AI is particularly good at translating a well-defined problem into a solution. It helps me less with deciding what I actually want to pursue, what I want to preserve, and what I am willing to give up.

I do not know whether this limitation will last. AI can already suggest priorities and argue for an approach. I would not base the future role of humans on an assumed inability of machines to make decisions.

But making a recommendation and standing behind a decision over time are still different things.

Expertise is often presented as having the right answers. Yet several competent people can understand the same situation and choose different directions. Part of what we seek in another person is a way of seeing, a sense of taste, and an approach to trade-offs.

A point of view becomes meaningful through what we prioritize, what we refuse, and the compromises we accept. It gives decisions coherence. It must also be able to evolve when reality pushes back.

Perhaps this will become a growing part of what we buy: “I want to work with you because your way of seeing things leads to outcomes I care about.”

Being distinctive will not be enough. A position can attract attention without being useful. Its value will emerge through the decisions it enables and what it produces over time.

A biodiversity of ideas

AI could lower the cost of giving a conviction a concrete form.

A different way of designing a service or organizing an activity can remain abstract for a long time when trying it requires a team, funding, and approval from many decision-makers. If those obstacles diminish, more people will be able to show what their vision actually produces.

I see the possibility of a kind of biodiversity of ideas: different ways of building, welcoming customers, selling, and collaborating could find their place.

That diversity is not guaranteed. The same tools make imitation and the reproduction of conventions easier. They can multiply polished outputs that look alike and ultimately express very little.

The potential is there, but it requires intention. We will need to recognize when our tools are diluting a distinctive idea into something merely familiar.

A client can know how and still want to delegate

This raises an economic question: will it still be possible to solve a problem once and repeatedly sell the solution when clients can reproduce it at low cost?

A problem can remain important while its solution becomes difficult to sell. Some businesses built around procedural complexity or scarce access to a skill could lose part of their market value.

That does not mean all friction is pointless. Some checks and procedures protect real interests. But we will need to distinguish the value of a service from the value that depends mainly on how difficult it is to do without it.

Above all, there is a fundamental difference between having a tool and wanting to take care of an issue.

Even when an answer is accessible, someone must remember to ask the question, understand the situation, choose an approach, follow through, and respond to the consequences. Access to information does not remove that mental load.

A person can become much more capable with AI and still be limited by their attention. Every new capability can even become another thing to keep track of.

They may therefore be willing to pay to focus elsewhere, knowing that someone is genuinely taking care of a particular area.

Buying responsibility

If I have to remember to contact you, formulate every request, check your progress, and make every decision, you are helping me produce things. But I still carry the responsibility for the subject.

If you notice what needs attention, propose a direction, make the decisions we have agreed you can make, and involve me when necessary, some of that load has actually been transferred.

You cannot delegate a mental load while continuing to carry it.

The client is buying the capacity to take responsibility. They want someone they can rely on. What matters to them is less the effort involved than the reliability of that commitment.

The promise might be: “You entrust me with this part of your project. I keep it moving, make decisions within the boundaries we agree on, and involve you when your input is genuinely needed.”

That requires real decision-making authority and explicit limits. Taking responsibility for an area does not mean guaranteeing every outcome or absorbing every risk. We need to know what remains with the client, what requires approval, and what should trigger an alert.

Delegation becomes a skill on both sides: knowing how to accept responsibility and knowing how to entrust it to someone else.

One person, several projects

Another hypothesis follows: a project may no longer need all the time of the person responsible for it.

Today, a responsibility can fill a working week because carrying it out involves so much production: preparing, analyzing, writing, building, checking, and coordinating. If AI substantially reduces that time, the need for judgment and follow-through may remain without occupying someone full-time.

The load also varies. Some moments require intense attention; others mainly require someone to stay aware and intervene at the right time. A professional might then have both the capacity and the desire to contribute to several projects.

Could they take responsibility for five, six, or even ten projects where one previously absorbed all their time? Those numbers express a question, not a promise.

Production time and mental load do not necessarily fall together. AI can make it easier to recover context without removing the weight of decisions and relationships. Several projects can also become urgent at once.

This is a hypothesis to test: how far can we expand our commitments while remaining dependable?

If the answer is “significantly further than before,” assigning a person exclusively to one company could become less relevant for certain roles. We would need to organize multiple working relationships, each strong enough for the people involved to move forward with confidence.

A subscription as a mutual commitment

This is where a subscription starts to make sense.

Paying per assignment forces the client to make a new purchasing decision with every request. Is this problem worth spending money on? What exactly should I ask for? How much will it cost? That series of decisions maintains part of the load they are trying to delegate.

A subscription can simplify those interactions. But it also serves another purpose: sustaining a relationship with someone you do not need full-time, but do need to be able to count on.

A salary already creates continuity between remuneration and commitment. A subscription could fulfill part of that function in a different, non-exclusive relationship, where a professional contributes to several groups.

The client pays for a real place among the professional’s priorities and commitments. The professional organizes their work so they can understand the context, take responsibility for a defined area, and remain available within agreed boundaries.

A form of reciprocal loyalty develops. The client sustains the relationship; the professional honors their commitments, even when another request seems more interesting or better paid.

This arrangement also allows activity to be uneven. The value of a quiet month may be that the subject remains under watch and someone is ready to act. It cannot be measured solely by visible output.

That requires balance. A professional cannot sell the same unavailable capacity several times over. A client should not feel obliged to generate requests to “get their money’s worth.”

Deciding not to launch something, avoiding a complication, or stopping a project can all be valuable outcomes.

The network becomes a working environment

Taking responsibility does not mean knowing how to do everything.

If a company asks me to improve its sales, my technical skills may help me understand certain obstacles, organize processes, and put tools in place. That does not automatically make me a sales director or an experienced salesperson in its industry.

I need to distinguish what I can take responsibility for, what remains with the client, and what requires another person.

A network expands that capacity. It brings together professionals whose skills we know, along with their judgment, limits, and reliability.

It also becomes the environment through which we gain access to several projects and groups. Shared experiences teach us whom we can entrust with an issue and with whom we can navigate a difficulty.

But introducing three experts to a client does not, by itself, relieve the client’s burden. If they then have to coordinate the experts and resolve their disagreements, the load returns to them. Coordination itself needs an owner.

The network can also provide support when someone runs short of time or reaches the limits of their expertise. That makes individual commitments more dependable, provided context can be passed on and everyone understands who is responsible for what.

A community gives this network continuity through discussion, learning, mutual support, and the sharing of experience. Relationships develop beyond a particular assignment, making subsequent collaborations easier.

Less exclusivity, greater trust

There is an important tension in this arrangement: relationships could become more flexible while requiring a deeper human commitment.

When we buy a task, we can define a deliverable and check the result. When we entrust someone with a responsibility, we accept that they will exercise judgment in situations we have not fully anticipated. We are placing something that matters to us in their care.

The analogy with entrusting someone to look after our children is useful in this specific respect. We explain what matters, choose a person we trust, and then need to be able to leave. If we remain there to watch every action and make every decision, delegation has not really happened.

The point is to direct our attention elsewhere, knowing that someone is taking care of things.

That trust requires mutual understanding. The professional needs to understand the client’s intentions, priorities, and what they want to preserve. They need to act without detailed instructions while recognizing when consultation is necessary.

The client, in turn, must accept that not everything will be done exactly as they would have done it. The discretion they grant is necessary for responsibility to be transferred.

Trust develops through experience: keeping a commitment, flagging a difficulty early, acknowledging a mistake, taking a useful initiative. It can begin within a limited scope and grow gradually.

Visibility still matters, but it should provide reassurance without requiring supervision. Good delegation makes it possible to stop paying daily attention to a subject.

A subscription can support that continuity. It does not create trust on its own.

What companies still provide

These changes do not remove every reason for companies to exist.

A company makes it possible to invest, build a shared culture, train people, and pool certain risks. It can keep a team together during a difficult period or fund learning before it becomes profitable.

If activities move into networks, those functions will need to be preserved or reinvented. Fluidity does not eliminate risk; it can simply shift it onto individuals.

Some professionals may enjoy abundant demand while others face greater instability. Skill alone will not always be enough. Access to relationships and clients will matter considerably.

The position of beginners could be particularly difficult. How do people learn to exercise judgment if entry-level tasks are automated and clients mainly seek people who are already autonomous?

A sustainable network will need to offer opportunities to learn through real situations, with guidance and progressively greater responsibility. It will need to fund that learning and develop forms of mutual support.

Otherwise, this arrangement could work extremely well for established professionals while being much less welcoming to everyone else.

A hypothesis to put into practice

This is the shift in which I would like to find my place, beginning in Switzerland, with people and situations I can know directly.

I would start with an area I can take responsibility for, working with a small number of clients. Around those concrete commitments, I could involve people I know and discover with whom a dependable collaboration develops.

The first questions would be practical. Can the client genuinely let me act? Can I take responsibility without becoming permanently available? Does AI reduce production work enough to make several commitments possible? Does the price make the relationship sustainable for everyone?

The network would grow out of those experiences. Expanding into other countries would make sense when it brought useful relationships, perspectives, or expertise.

I do not yet know how far this model can go. But it seems to raise a more productive question than whether jobs will disappear: how should we organize work when producing becomes easier, but choosing, trusting, and taking responsibility remain demanding?

An essential part of the value of work might then lie in a simple promise: “You can focus on something else. I’ll take care of this.”