AI in Job Profiling and Evaluation – Part 3: AI Does Not Create Expertise
– It Exposes It
In the first two articles of this three-part series, Belinda Oregan, Executive Consultant and Industrial Psychologist at 21st Century, explored why strong source data, a documented methodology and defensible governance are essential for effective job profiling and evaluation. In this final instalment, she turns the spotlight on the professional at the centre of the process. AI may be changing the tools of job evaluation, but is it changing the need for expertise, or making genuine expertise more visible than ever?
AI Does Not Create Expertise, It Exposes It
Which leads to the biggest misconception of all: that AI creates expertise. It does not. It exposes whether the expertise was there in the first place. Someone who could not evaluate jobs before AI still cannot evaluate them simply because the final report looks professional. Poor judgement in polished language is still poor judgement.
An experienced practitioner uses the same tools very differently: to increase productivity, not to outsource thinking. They may use AI to draft profiles faster, compare similar roles, flag inconsistencies, summarise interview notes and strengthen documentation. Before any grade is accepted, however, they test the output against organisational context, the selected methodology, comparable roles, governance principles and professional judgement. The good ones do, anyway. The expertise never lived in the AI; it stayed with the practitioner. AI became another tool in the hands of someone who already knew what good looked like.
The Honest Counter-Argument
I must be fair to the other side, because the honest version of ‘this time is different’ is more serious than it sounds. Generative AI can produce language, analyse information, compare options and construct explanations in ways that resemble parts of professional reasoning. That capability is materially different from many earlier forms of workplace automation. It is therefore possible that AI will automate or reshape a larger share of cognitive work than previous technologies did. The scale and speed remain uncertain, but the ILO’s task-level evidence supports expecting substantial job transformation rather than dismissing the impact as hype (Gmyrek et al., 2025). Some researchers also debate whether increasingly capable systems could ever possess genuine understanding or consciousness. That debate is not settled, and this article does not need to settle it. The practical question is simpler: would a board, a trade union or an employee accept ‘the AI said so’ as accountability? I doubt it.
Why do I not believe that settles job evaluation? Because judgement here is not simply clever reasoning over words. It is accountability for a decision, informed by one organisation’s context and made defensible to the people it affects. A model can generate a flawless-looking rationale, but it does not automatically know the context that was never written down, identify what the profile omitted or stand before a Remuneration Committee or trade union and answer for the outcome. Its output must still be checked against the organisation’s evidence, rules and methodology. Whether these systems are genuinely ‘thinking’ is less important than the fact that the employer remains responsible for the employment decision. Technology may assist that responsibility; it does not transfer it.
The Real Lesson
Maybe that is the real lesson. I do not believe AI will simply replace experienced job evaluators – or capable professionals generally. It will make many of them dramatically better and faster. But it will also allow inexperienced practitioners to produce weak evaluations more quickly, more confidently and in documents that appear more credible than ever. That may be the greatest risk of all: not simply that AI gets the answer wrong, but that it can make the wrong answer look unquestionably right. Worse still, it may arrive at the ‘right’ answer without the credible reasoning, evidence and audit trail needed to defend the decision. The inexperienced person who put the profile through AI will certainly not be able to supply what is missing.
So use AI – but do not mistake it for a substitute for deep knowledge of the work. Those who evaluated jobs well before AI, and who possess genuine expertise, will use it to become even better. Those who lacked the expertise, or performed the work poorly before, are building a house of cards; the first serious trade-union or board challenge may bring it down. Because when the board or the union asks, ‘How did you get to this grade?’, ‘The AI said so’ will never be a sufficient answer.
