Since AI arrived, obituaries for experts have multiplied.
They are not written by the deceased. Most are written by someone in the neighboring profession, watching a person who is still working at their desk.
Business leaders say engineers are no longer necessary.
Engineers say designers are no longer necessary.
Implementers say people who review code are no longer necessary.
The funeral has already begun, while the person inside the coffin continues working as usual.
This is a remarkably strange sight.
In The Age Of AI, Obituaries Arrive Early
AI makes many things. It writes code, creates screens, drafts prose, proposes designs, and produces review comments. At the very least, it produces plausible forms of these things in astonishingly little time.
People see the deliverable and imagine the end of a profession.
Code appeared. Then we no longer need engineers.
A screen appeared. Then we no longer need designers.
Review comments appeared. Then we no longer need reviewers.
This reasoning is fast. Almost as fast as AI.
It is also remarkably low-resolution.
It treats the deliverable in front of us as identical to the entire job of the profession that produces it.
Every job has a deliverable on its surface. But the substance of the job is not the deliverable alone. It includes deciding what to make, asking why it should exist, choosing what not to make, imagining how it can fail, judging quality, and accepting responsibility when something happens.
There is a long corridor between AI producing part of a deliverable and the entire function of an expert becoming unnecessary.
We are not running down that corridor. Sometimes we simply pretend we teleported from the entrance to the exit.
"Unnecessary" Is A Heavy Conclusion
Becoming easier, becoming faster, possibly requiring fewer people, and changing roles are all different from becoming unnecessary.
To call something unnecessary is to judge that its function can be removed without damaging anything we need.
To make that judgment, you must at least know what the expert ordinarily does. Not only what they do when things work, but what they do when things fail. Not only what they visibly produce, but what they prevent. Not only when their hands are moving, but where they decide, coordinate, and carry responsibility.
Only after seeing that whole function and showing that it has completely moved elsewhere do we arrive at the entrance of an argument for obsolescence.
A business leader sees the speed of generated code and declares engineers unnecessary. But where did the work of turning ambiguous business needs into systems go? Where did estimating failure, choosing between operational and change costs, and finding causes when incidents happen go?
An engineer sees a generated screen and declares designers unnecessary. But where did the work of deciding information hierarchy, experiential consistency, accessibility, brand, user confusion, and the felt judgment of good and bad go?
An implementer sees AI-generated findings and declares reviewers unnecessary. But where did the work of asking whether a locally correct change breaks the direction of the whole, which risks the organization should take, and who will continue to understand the implementation go?
Work that appears to have vanished may simply have been pushed onto someone else. It moves to users, operators, future implementers, or yourself on the day the incident finally happens.
If you cannot name where the work was transferred, you did not eliminate it. You only removed it from the ledger.
What Only An Expert's Self-Deprecation Contains
An engineer hands part of their work to AI and says, "at this rate, I may no longer be needed." There is real experience inside that statement.
Something that took an hour yesterday now appears in a minute. A skill built over years is reproduced without explanation. Work disappears from the person's hands, and they feel even their judgment moving into the machine.
An expert's self-deprecation contains observation from inside.
What changed from before? Which ability lost value? What still remains? At least with respect to their own work, the person has both a past and a present to compare.
And they are calling themselves unnecessary. The first chair threatened by the declaration is their own.
That difference matters.
Self-deprecation offers oneself as the specimen. An outsider's argument for obsolescence places someone else on the table as the specimen.
Of course, an expert's self-deprecation is not always right. They may see only one part of themselves, or generalize a temporary shock into the end of an entire profession. Sometimes self-deprecation simply begins to feel good, and the person gives up on themselves before the world does.
So when an expert says, "our profession will become unnecessary," that does not automatically make the prediction correct.
But it does contain testimony.
When someone outside the field declares experts unnecessary without understanding their work, there is not even testimony. There is only a deliverable visible from the outside and a conclusion convenient to the speaker.
If an argument for obsolescence is to stand without the expert's own self-deprecation, it needs evidence stronger than self-deprecation can provide.
Arguments For Obsolescence Reveal The Speaker
Words that call someone unnecessary appear to describe that person, but often describe the speaker's desires instead.
A business leader says engineers are unnecessary. Is that truly an observation that the engineering function has disappeared? Or is it a wish to reduce labor costs, make development look faster, and avoid hearing an expert say, "that is not possible"?
An engineer says designers are unnecessary. Is that truly an observation that the design function has disappeared? Or is it a wish to avoid questions about small intentions, avoid revisions to their implementation, and place the entire screen under their own control?
An implementer says reviewers are unnecessary. Is that truly an observation that the review function has disappeared? Or is it a wish to avoid changes, avoid explanations, and merge faster?
I am not saying this is true every time. Sometimes the structure of the work genuinely changes and fewer roles are needed.
But experts often create friction. They ask questions, stop things, expose ambiguity, and attach conditions to easy conclusions. In a room excited by "we can do it," they say, "we can, but it has this cost."
From outside, that friction can look like the expert is slowing the work down.
Sometimes that friction is the work.
Some arguments for obsolescence are made not because expertise is unnecessary, but because people resent having their speed or freedom interrupted by expertise.
They do not want experts to disappear. They want only the expertise that is inconvenient to them to disappear.
AI wraps that desire in beautiful paper labeled "the future."
Expertise That Succeeds Makes Nothing Happen
Much of an expert's work becomes invisible when it succeeds.
When an engineer prevents an outage, nothing appears on the screen. When a designer reduces user confusion, users do not notice that they did not become confused. When a review prevents a future incident, that incident never happens and is difficult to record as an achievement.
There is no screenshot for nothing happening.
By contrast, code, screens, and findings generated by AI are visible. They can be counted, timed, and demonstrated.
Place a visible output beside a failure that never happened, and the output looks more like work. This is a deeply unfair comparison.
The value of expertise lives not only in what was made, but also in what was not made, what was avoided, and what was stopped.
The less you know about a field, the more that space looks empty.
Not seeing something is different from it not existing.
Respect Does Not Mean Unconditional Belief
When I say we should respect experts, it may sound as though I am asking for their sanctification.
I am not.
Experts are wrong, too. They exaggerate to protect their field, hide behind jargon, resist change, defend established interests, and underestimate new tools. Expertise itself can over-optimize someone for the past.
So we do not need to believe experts unconditionally. We can argue with them, evaluate their results, and reduce their role when it truly can be reduced.
Respect is not agreement.
Respect means preserving the possibility that there is work you cannot see.
Do not make declarations about a field while leaving it unknown. Do not assume that reproducing one part of a deliverable means understanding the whole role. Ask at least once what the expert protects, prevents, and decides.
If you still judge the role unnecessary, explain what became unnecessary, what remains, and where judgment and responsibility moved.
Do not force experts to prove the value of their own existence forever. The person repeatedly asking "are you really necessary?" also carries a burden of proof.
The person declaring something unnecessary should be required to have the highest resolution.
Some Work Really Does Become Unnecessary
Of course, some work truly does become unnecessary.
Technology removes processes. Roles merge. A task that once required one specialist becomes something another profession can handle with a tool. AI may make these changes larger and faster.
We must not deny change in the name of respecting experts. Work does not exist only to protect human dignity, and users are not obligated to keep paying for processes that are no longer needed.
But work genuinely becoming unnecessary is different from someone wanting to call it unnecessary as early as possible.
What we need is not a permanent ban on funerals. We need confirmation of death.
Did the function really move? Was quality preserved? Who notices failure? Who decides, and who takes responsibility? Can the system remain without the expert not only through a short demo, but through long operation?
Only after checking these things should we speak about the end.
Measuring a coffin beside someone who is still breathing because you expect them to die one day is not foresight.
Conclusion
An expert's self-deprecation contains experience. It comes from someone who handed work to AI and knows what disappeared from their own hands. It may be wrong, but it is at least testimony from inside.
If someone outside the field wants to declare an arbitrary expert unnecessary, they need understanding and evidence strong enough to exceed that testimony.
Which work disappeared?
Where did judgment move?
Who finds the failures?
Who accepts responsibility?
An argument for obsolescence that cannot answer these questions is not a prediction of the future. It is an introduction that says the speaker cannot see the work of the present.
AI will change the work of many experts. It will take some of it, lighten some of it, and reshape some of it. Entire professions will probably disappear.
That is exactly why we must not end them carelessly.
We do not need to believe experts unconditionally. But we should not pronounce death on expertise we do not understand while continuing not to understand it.
Do not start an expert's funeral without them.
At the very least, ask the person inside the coffin what work they are doing now.