Size a role the same way every time, with a written reason for every call. A skill that follows your method, checks what is missing, and assesses the role, never the person.
Job sizing goes wrong in two ways. The first is inconsistency: different people, on different days, give different answers for similar roles. The second is sizing the person instead of the role, so a strong performer's job drifts up and a struggling one's drifts down.
This skill fixes the process, not the judgement. It makes the assistant check the role description first, assess only the work, rate each factor against your level descriptions, quote the words behind each call, and flag the calls that could go either way. A person still decides the final size.
For your team, the gain is consistent answers whoever does the sizing, and a written trail you can show when someone asks why a role sits where it does.
A skill is a written set of instructions, saved as a file, that an AI assistant follows every time it does a particular job. It turns "a good first draft, if you ask the right way" into "the same method, every time, whoever is asking".
Two saved conversations. In the first, an invented role is attached and sized. In the second, someone pushes back by pointing to the person in the role. Press Play to watch, or Show all to read.
These are real responses, saved and replayed. Nothing you do here is sent to an AI.
Five steps from a role description to a result you can defend. Press Play, or step through at your own pace.
Make sure every Banksia employee is paid correctly and on time, in line with our enterprise agreement and the relevant Modern Awards.
Start with the role description. The skill works from the words on the page. The highlighted phrases are the kind of wording it will quote as evidence.
The gate. If something essential is missing, the skill stops and lists what it needs instead of guessing. A silent team or budget does not stop it, but it is flagged and the result is provisional.
Factor by factor. Each call has a level, a one-sentence reason and the exact words that support it. If there are no supporting words, the call is treated as an assumption.
This one matters: at Level 3 the overall level would drop from 4 to 3.
Borderline calls. The skill names the two levels, says what would tip the call, and says whether tipping it would change the overall result.
Provisional because the budget was not stated and was assumed to be none. Medium because two borderline calls could each lower the overall level.
A result you can defend. The level, the rule used, the confidence and the reason, every call tied to the role's own words. A person still decides the final size.
The role and every rating in this walkthrough are invented, for illustration only.
You can build your own version with any capable AI assistant your employer allows, for example Microsoft Copilot, Claude or ChatGPT. The skill on this page uses an illustrative five-factor method written for this site. The point is to put your own method in its place.
Use only material you are allowed to share. Do not paste a licensed job evaluation method you do not have the rights to share. If your method belongs to someone else, write down your own plain-English version of how you apply it, or ask your supplier what you may load into an AI tool.
Most methods live in people's heads and in slide decks. Get the assistant to interview you, one question at a time, and write it up.
Now ask for the skill file itself. This prompt lists the behaviours that make the difference: the gate, the role-not-person rule, quoted evidence, borderline flags, confidence and a consistency check.
A skill is only as good as its results on roles where you already trust the answer. Test before anyone else uses it.
Once the skill gives answers you trust, make it easy for the team to use it.
Microsoft's documentation says agent instructions can be up to 8,000 characters, and a skill file up to 20,000 characters (Microsoft Learn, checked November 2026). We have not yet tested this skill inside Copilot.
The sample SKILL.md on this page is about 7,550 characters long, counted by us, so it is close to the first of those limits. If you add your own level descriptions it will probably need trimming to fit an agent's instructions.
Role descriptions are about jobs, not people, but they often carry names. Conversations about sizing also drift into the jobholder. Remove names before you start, and keep the conversation on the work.
Role descriptions without names are usually low risk. Your organisation's sizing method may be licensed or confidential, so check before you paste it anywhere. Use tools your employer has approved.
Privacy rules for employee records differ between private employers and government agencies, so check with your privacy officer. Start with the OAIC's guidance on the Australian Privacy Principles and on employment and privacy. General information, not legal advice.
To use it: add SKILL.md to your approved assistant, attach a role description with the names removed, and ask it to size the role. Then replace the factors and levels with your own method.
What next: Back to all use cases