AI is changing tasks. Job titles tell only half the story.

Early discussion about AI and employment often treated administrative work as the obvious target and technical work as relatively protected. Generative tools have complicated that assumption because they can now draft code, documentation, analysis and written communication across many roles.

Exposure is not the same as displacement

A role is a bundle of tasks. Some may be accelerated, some may be automated, some may become more important and others may not change at all. Research that measures occupational exposure tells us where technology may affect work. It does not prove that a whole job will disappear.

The International Labour Organization's 2025 global index concluded that transformation is the most likely effect for many occupations exposed to generative AI. Clerical occupations remain highly exposed, but exposure has also increased in strongly digitised professional and technical work.

Why development work changed quickly

Software work already happens inside digital tools and produces structured outputs that can be tested. AI can draft boilerplate, explain unfamiliar code, generate tests and help trace an error. That can raise the output of an experienced developer, but it can also alter the entry-level work through which new developers traditionally build understanding.

Why administrative work is not one simple category

Scheduling, transcription, document drafting and data movement may be highly amenable to automation. Administrative roles also contain exceptions, relationships, organisational memory and judgement about what matters. Those parts are harder to see in a process map and costly to lose.

For an SME, the person who appears to be completing routine administration may also be spotting a worried customer, correcting incomplete information, coordinating colleagues and preventing small issues becoming expensive ones.

What an SME should do before changing roles

1. Map tasks rather than making assumptions from job titles.
2. Separate drafting and data movement from decisions and accountability.
3. Measure quality, rework and customer impact, not only time saved.
4. Protect the learning path for less experienced staff.
5. Keep an owner for every automated output and exception.
6. Involve the people doing the work because they understand the hidden steps.

Use AI to improve the role before trying to remove it

In a small business, resilience often comes from capable people who understand several connected parts of the operation. A sound first objective is usually to reduce low-value repetition and improve consistency while keeping the knowledge, relationships and judgement that make the role valuable.

The commercial test is straightforward: does the new method help the business deliver better work with clearer responsibility? If the answer depends on ignoring errors, weakened service or lost knowledge, the apparent saving is not a real one.