Rate yourself from 1 to 4 against ten skills employers look for in AI agent work, and for each rating write one piece of evidence. A score without evidence does not count. The gaps that remain tell you what to learn next and which roles fit you now. The template, scale and a filled sample are below.
Why assess yourself before applying or studying?
Because "AI skills" covers very different things. A role reviewing agent decisions in a bank needs judgement, knowledge of the bank's rules and a feel for red lines. A role designing agents needs precise writing, test design and an understanding of limits. Most people are strong in some of these already and weak in others. A self-assessment shows you which, before you spend time on the wrong course.
It also gives you language for interviews. An assessment centre or structured interview will probe the same skills; having rated yourself strictly, with evidence, you will know your strongest examples. If you study alone, pairing the assessment with self-paced learning keeps your plan aimed at real gaps.
How do you score yourself?
Use this four-point scale for every skill:
- Aware: I know what it means but have not done it.
- Practised: I have done it in a course, lab or simulation.
- Applied: I have done it in real work, under supervision or with review.
- Trusted: I do it in real work without review, or I teach others.
Write one piece of evidence for every score of 2 or above. If you cannot, drop the score by one. Be strict; an inflated assessment helps nobody.
The self-assessment template
| Skill | What it means in practice | Score (1 to 4) | Evidence | Next step |
|---|---|---|---|---|
| 1. Reviewing decisions | Approving, amending, rejecting or escalating what an agent proposes, with a reason | |||
| 2. Spotting red lines | Recognising cases that must always go to a person in your sector | |||
| 3. Judging sources | Telling a verified record from a claim or an emailed document | |||
| 4. Understanding limits | Knowing what an agent may do at each depth, and who sets the limits | |||
| 5. Reading a decision record | Finding why an agent acted, which bound decided and who approved | |||
| 6. Escalating well | Passing a case to the right person with what they need to decide | |||
| 7. Spotting patterns | Seeing that many held cases share one cause | |||
| 8. Writing a clear brief or change request | Describing what an agent should do, or what should change, precisely | |||
| 9. Knowing your sector's rules | Data protection, conduct, complaints and any rules specific to your work | |||
| 10. Keeping knowledge current | Noticing when a policy or source is out of date and getting it fixed |
Skills 1 to 6 matter most for reviewer and operator roles. Skills 7, 8 and 10 grow in importance for people who train or own agents; see what skills an AI agent trainer needs and what skills an AI agent owner needs.
How do you turn the scores into a plan?
Look for two things. First, any score of 1 in skills 1 to 3: these are the basics of review and the place to start. Second, skills where you scored 3 or 4 with strong evidence: these are your interview stories and the lines to lead your CV with.
Pick no more than three next steps. A plan with ten will not happen. Revisit the assessment after each course or practicum and update the evidence.
A filled sample
An invented accounts assistant, Daniel, rates himself:
| Skill | Score | Evidence | Next step |
|---|---|---|---|
| 1. Reviewing decisions | 3 | Reviewed held supplier payments in an invoice pilot | Formal practicum for a marked record |
| 2. Spotting red lines | 3 | Refused all bank changes received only by email | — |
| 3. Judging sources | 3 | Same as above; checked orders and deliveries before paying | — |
| 4. Understanding limits | 2 | Introductory course on governed agents | Operator modules on how an agent decides |
| 5. Reading a decision record | 1 | None | Lab on reading decision records |
| 6. Escalating well | 3 | Escalated payments above client limits with documents | — |
| 7. Spotting patterns | 2 | Noticed credit notes billed as invoices in the pilot | Write it up as a proposal |
| 8. Writing a change request | 1 | None | Practise on the pattern above |
| 9. Sector rules | 3 | Daily work under VAT and payment controls | — |
| 10. Keeping knowledge current | 2 | Flagged an outdated supplier list once | — |
Daniel's three next steps: a lab on decision records, the Operator modules on how agents decide, and turning his credit-note pattern into a written change request. His interview stories will be skills 2 and 6.
Where the Academy fits
Pryme Intelligence Academy's tracks line up with this assessment: Foundations covers skills 4 and the basics of 1 to 3; the Operator track covers review, reading the workspace, keeping knowledge current and incidents, with a practicum that turns a self-rated 2 into marked evidence. The AI agent risk assessment template is a useful companion if you work on the employer side.
Questions people ask
How often should I redo the self-assessment?
After each significant step: a course, a practicum, a pilot, a new role. Update the evidence column rather than starting again.
Can I share it with my manager?
Yes. It makes a good basis for a development conversation, because it shows what you can already do and what support you need.
What if I score 1 almost everywhere?
That is a clear starting point, not a problem. Begin with an introductory course on governed agents and come back in a month. The AI skills for job seekers guide shows what employers ask you to prove.