Personal growth · · ⏱ 8 min read

Framing displacement

The AI/work debate polarised between panic and denial. Environmental bias, task versus role, and the competencies that emerge in the gap it leaves.

The public debate on AI and work has polarised between two bad readings: panic (“half of jobs will disappear in five years”) and denial (“it’s just another tool”). Neither survives contact with the data. There is a third possible reading — displacement exists, it happens at real speeds, it admits reading with judgement — but it needs something almost no one is doing: understanding the biases that distort how change is perceived, and looking at what competencies actually emerge in the gap it leaves behind.

A figure looks through a spyglass towards a horizon with silhouettes of people at dawn
Focus is not a matter of calm. It is a decision about where to look.

Two bad readings and one possible

The panic reading turns every study, every headline and every demo into evidence of an imminent avalanche. Sources that speak of “40% of jobs will be affected” get read as “40% will disappear”, which is not the same. The time scale gets compressed: what may take twenty years to settle is told as if it were a matter of months.

The denial reading does the opposite. It reduces every new capability to “assistant”, “tool”, “doesn’t change the role”. It ignores that earlier tools — email, spreadsheets, search engines — did change entire professions when their time came, even if in the moment they seemed to be just that: tools.

Neither is useful for making decisions. And between them there is a third reading, far less noisy: professional displacement exists, is measurable, and happens at speeds measured in decades, not months. That reading doesn’t sell clicks. But it is the one that allows action with judgement.

The technical-environment bias

There is a reasoning pattern that explains a good part of the distortion. It can be called technical-environment bias, and it works like this:

  1. In my professional environment, everyone handles a given skill fluently.
  2. Therefore that skill is common.
  3. Therefore an AI that automates it is replacing something common.
  4. Therefore many will be left out.

The error is between step 1 and step 2. This bias operates with particular force in organisations whose business core is IT knowledge: software development firms, technology integrators, IT departments of large corporations, tech-specialised media. All share a structural property: they live surrounded by people who know how to write code. Their teams, their clients, their competitors, their LinkedIn. In that environment, “writing code” feels like a base skill, almost commodity. And from that reading, any automation sounds like sweeping the profession away.

But the environment is not reality. Outside the glass of that bubble there are:

  • Companies that have been trying for years to hire technical profiles and cannot find them.
  • Entire departments where no one knows what a pull request is.
  • Whole sectors with manual processes that have never seen code.
  • Regions where the technical talent shortage is structural, not cyclical.

This is what can be called the density fallacy: generalising the density of a skill in one’s own environment to the whole market. When that fallacy operates in the background, any advance in automation reads as a threat because it is assumed — wrongly — that there is a saturated mass of people doing what the machine now does.

Reality is more asymmetric. In many cases the machine is not replacing a saturated skill — it is filling a gap that has gone unfilled for years.

What actually gets displaced, and at what speed

To frame displacement well, a fine distinction is required that most headlines don’t make: task automation is not the same as role elimination.

A task is a bounded fragment of work: drafting an email, reviewing a standard contract, searching for information in technical documentation. A role is a whole professional position: lawyer, engineer, journalist, accountant. Roles are made up of dozens of heterogeneous tasks, and only some of them are AI-automatable at current levels.

When the public conversation says “lawyers are being replaced”, what usually happens is that some tasks of legal work are being automated — standard-contract review, standard case-law search — while others — negotiation, procedural strategy, interpretation of nuance — remain deeply human. The role does not disappear: it recomposes. And that recomposition happens over years, not quarters.

Time scales matter. Automation of calculation with spreadsheets began in the 1980s. The effective elimination of the accountant as a profession has not occurred — the role transformed, gained layers, and still exists forty years later. Deep changes in the labour market are measured in professional generations, not in hype cycles.

And there is something important to acknowledge: a good part of current uncertainty comes from pretending certainties no one has. Nobody knows precisely what percentage of which tasks of which roles will be automated in what timeframe. The studies that circulate are estimates with enormous margins. Naming that uncertainty, instead of replacing it with dramatic predictions, is part of framing it well.

The competencies that emerge

Facing displacement, there are two distinct attitudes. One is to hide one’s head. The other is to ask what capacities gain value when others get automated.

Three emerge clearly when part of the work is done by a mixed human-agent system:

Design of mixed systems. Knowing how to write code or manage projects is no longer enough. The work of designing how humans and agents share tasks, synchronise and validate each other appears. It’s a new competency combining systems architecture, process management and ethical judgement. Very few people have it yet.

Auditing distributed decision chains. When several agents cooperate to produce an output, the ability to reconstruct why the system decided what it decided becomes a specific professional skill. It requires understanding agent-to-agent communication protocols, emergent failure patterns and traceability structures. It is what in a few years will be called “AI systems auditor” and today has barely established formation.

Governance of reconstructible responsibility. When human signature no longer covers what the person produced task by task, but what a system they designed produced under their structural supervision, a new layer of professional responsibility appears. Translating that responsibility into legal, contractual and organisational frameworks is work still to be done, and whoever masters it positions themselves in a niche where supply is nearly nonexistent.

These three competencies are not consolation. They are a practical map of emerging niches where supply is today far smaller than foreseeable future demand. Those who begin to train in them now are not learning to survive displacement — they are positioning themselves in the gap displacement opens.

The responsibility of framing

So far, the analysis. But framing displacement well is also a professional responsibility, and it’s worth naming.

Reacting like a headless chicken protects no one. Not the person paralysed by panic, not the organisation that avoids the conversation out of commercial fear, not the public debate degraded into dramatic headlines. Panic produces decisions worse than the problem it fears: mis-targeted training, reactive labour policies, organisational resistance that delays necessary adaptations.

Framing with judgement is a professional skill in itself. And like any skill, it has three requirements worth naming as a triad: data over intuition (concrete numbers order better than feelings), long scales over urgencies (structural changes are measured in years, not quarters), and differentiation over generalisation (task is not role, activity is not profession, automation is not elimination).

All three must be present. Neither alone is enough.

And here something uncomfortable but necessary appears. Organisations that feel threatened by the debate tend to be precisely the ones that would benefit most from participating in it with judgement. Hiding does not protect them — it leaves them out of the conversation where it is decided what training is prioritised, what profiles get hired, what capacities get developed. The best way to miss the train is not to board it with judgement: it is to pretend the train doesn’t exist.

Open questions

  • If panic and denial are equally incapable of framing the change, what training do professionals need to develop the third reading with judgement?
  • When an organisation frames displacement from the bias of its own environment, who has the responsibility to point out the distortion before it becomes internal policy?
  • If emerging competencies have nearly no supply today, why are educational systems and IT organisations not rushing to train in them?
  • Are we discussing what needs to be discussed, or are we amplifying headlines while real changes pass under the radar?

The questions don’t have closed answers. But one final idea is worth stating: professional displacement is not stopped by ignoring it, nor is it solved by dramatising it. It is framed with judgement, anticipated with training, and navigated with the capacities that emerge in the gap it leaves behind. Those who train in those capacities now will be where supply is scarce when demand matures. Those who paralyse waiting for certainties nobody will give will keep reacting to headlines while change happens elsewhere.

References