Ai In Job Profiling: When Efficiency Becomes A Risk

AI in Job Profiling and Evaluation - Part 1: The Seduction of Cheap Efficiency

In this first article of a three-part series, Belinda Oregan, Executive Consultant and Industrial Psychologist at 21st Century, explores what happens when generative AI moves from assisting job profiling and evaluation to influencing decisions that determine pay, structure and internal equity. Part 1 examines why polished AI output can be so seductive, and why the quality of the job profile still matters more than the quality of the prose.

"This Time Is Different", Again

"This time is different." History has heard those words before. More than once.

When horse-drawn transport gave way to the motor car, stable hands, carriage builders and blacksmiths watched a technology emerge that threatened everything they knew. Many of those jobs did disappear, but mechanics, engineers and transport specialists rose in their place. Work changed, but people did not become obsolete. The Industrial Revolution told a similar story, as did computers, email and the internet. The historical record is not that technology leaves employment untouched; it is that automation both displaces some tasks and complements others, changing the content and distribution of work rather than simply making human contribution disappear (Autor, 2015).

Now it is AI's turn, and once again we are told that this time people really are replaceable. But does the evidence support such a simple conclusion? The International Labour Organization's most recent global task-level analysis found that one in four workers is in an occupation with some exposure to generative AI, but that transformation is more likely than outright replacement because most occupations still require human input (Gmyrek et al., 2025). That is the distinction where many of today's assumptions begin to unravel. Too many organisations have decided that, because AI can produce work that looks professional, it must be able to replace the professional. Do not misunderstand me: I love AI. I love ChatGPT - although I often prefer Claude - and I have no intention of falling for either the fearmongering or the Terminator clips my children keep sending me. I keep up with the developments, build my expertise in using the tools and advise clients on how to deploy them responsibly for scale and significantly greater productivity.

Some organisations are learning that the useful answer is not 'AI or people', but 'which work should AI do, and where must experienced people remain accountable?' Klarna, a digital payments and Buy Now, Pay Later Services company, is a good example of why the story needs nuance. Its Chief Executive acknowledged in 2025 that an excessive focus on cost had contributed to lower-quality customer service and said customers should retain access to a human adviser. Yet Klarna continued to report major AI use: its 2025 annual report states that its AI assistant handled 80% of customer-service chats, based on company data, with no reported decline in customer satisfaction (Daly, 2025; Klarna Group plc, 2026). The motor manufacturer, Ford offers a similar lesson. It hired, promoted or brought back about 350 experienced technical specialists as part of a wider quality-improvement programme, using their institutional knowledge to mentor younger engineers, lead design reviews and improve automated and AI-enabled quality tools. That is not proof that Ford replaced 350 engineers with AI and then reversed the decision; it is evidence that automation was not sufficient without experienced engineering judgement behind it (Mohan, 2026). The lesson is not that AI failed. It is that AI alone was not enough.

The Real Cost of Cheap Efficiency

Why do organisations fall into this trap? They assume AI is cheaper because they compare a monthly subscription with an experienced professional's salary. On that arithmetic, the business case appears obvious - until the real costs are counted. The true cost is not the licence or the credits. It lies in incorrect decisions, rework, complaints and disputes, damaged trust, governance failures, legal exposure and the experienced people eventually required to fix problems that should never have arisen. Efficiency without judgement can become an extraordinarily expensive saving.

Why Job Evaluation Is the Riskiest Thing to Hand Over

Which brings me to the issue at hand: job profiling and job evaluation.

Of everything organisations are beginning to hand over to AI, job evaluation may be among the most misunderstood and the riskiest. It is not simply a writing exercise, like drafting a report or summarising minutes; it is a governance process. Every grading decision reaches far beyond the individual role. It shapes remuneration, career progression, succession, organisational design, recruitment and internal equity and, in South Africa, may affect compliance with the Employment Equity Act and the principle of equal pay or remuneration for work of equal value (Republic of South Africa, 1998). All of this rests on the job profile from which the evaluation begins.

Job evaluation is not only about deciding what a job is worth; it is about determining where that role sits within the wider organisational architecture. Move one job and the effect can ripple through dozens of others. Profile a role in isolation from the structure around it and you may create duplicated work, overlapping decision rights and an unnecessarily senior position on a premium salary that adds little to the strategy. That is precisely why experienced practitioners do not profile or grade a role in a vacuum.

The Profile Is the Source Data

The job profile is the source data, and a grade can only be as sound as the profile on which it is based. A credible profile must accurately reflect why the role exists, how it contributes to the organisation, the operating model, reporting lines, level of decision-making, financial impact, environmental complexity and the role's place within the organisation's grading philosophy. These days, everyone seems to be feeding a few bullet points into AI and producing job descriptions. They look good - but do they explain what the job must actually do to deliver the organisation's strategic intent? An expert job-profile compiler may use AI, but the prompts are informed by detailed interviews, organisational context, years of experience and exposure to many different roles. The result should be a practical, relevant profile, not merely a polished document. This is also consistent with the South African Code of Good Practice, which advises employers to ensure that job profiles or descriptions exist and are current before jobs are evaluated (Department of Labour, 2015).

The First Question to Ask

The first principle is simple: AI can accelerate profiling, but it cannot rescue weak source data or an incoherent role. Before asking what grade the system recommends, ask whether the job has been properly understood, challenged and situated within the organisation.

In Part 2, I take that concern further: what happens when the AI gives you exactly the grade you asked it to defend?

Ends.

Total Words: 1114