Would your role description work if an AI did the matching?

Would your role description hold up if an AI did the matching? For most, the honest answer is no. If you have even asked yourself the question, you are ahead of the field — because AI-assisted matching is already a fact. The real question is how it changes the way we write requirements.

I have had many conversations with managers who took an old job ad, glanced at it, and forwarded it to the TA team. Or who listed every tool and method the team uses — a list that, in the worst case, gets read as a requirement list.

A machine will interpret your text exactly as written. No shared history, no "you know what I mean". Which, it turns out, is a wonderful discipline — because a human candidate reads your ad the same way.

Start with three questions to the hiring team

• What does the team consist of today — personalities and competences?
• What challenge should the person we are hiring help the team solve?
• What is the hard, prioritised requirement list? (No, you cannot have two number ones.)

Then work the requirements down

Here is the method I use to get from wish list to requirement profile:

1. Find the one must-have. Which requirement must a candidate meet for the hiring manager to find the CV worth reading at all? This is usually a single thing. Get the manager to name it.
2. List the genuinely relevant requirements. These are the rest of the "mandatory" ones. Show the manager two candidates who both meet requirement number one, and a prioritisation emerges in the reasoning about why one CV feels more relevant. Note: we are still on hard requirements — merits, not personality.
3. Challenge the nice-to-haves. List everything else and ask: "So if I find someone who has all of the above but none of these — you don't want to meet them?" Work the Christmas wish list down until what remains is honest.

That is the skeleton of a role description that works.

The double win

Here is the part I find genuinely exciting: practising this kind of clarity for machines trains us in the art of deciding — and clearly communicating — what we actually want done. We become better at specifying, full stop.

This is also one of the moments Artemis was built around: when you can see which candidates surface based on how a requirement is phrased, it becomes very concrete which requirements really are mandatory — and which ones were habit.

Next time you start a recruitment, try the thought experiment: imagine your ad will be interpreted by an LLM. Write accordingly. Your human candidates will thank you too.