V Skill HF-D7.7-003 transversal 9 kompetenser

Evaluating and verifying AI output

Checks AI output for accuracy, completeness and usefulness, verifies claims against sources and decides whether the output is used.

Hur hrmforce mäter detta

The candidate performs a representative work sample under standardised conditions.

Nivå	Beteende på denna nivå
N1 Guided	Checks facts and figures from AI output against a reliable source before passing the result on. works under supervision and follows instruction · routine, one variable at a time · own task
N3 Proficient	Determines per task which checks are needed, spots incorrect or fabricated parts and accounts for the result used. sets own approach and seeks input proactively · several variables, some ambiguity · own team or process
N5 Leading	Designs the organisation's verification policy and determines for which applications human review remains mandatory. sets the standard and the policy · strategic, under high uncertainty · organisation, value chain or profession

N2 och N4 är medvetet inte förankrade. Bedömarna placerar dem mellan förankringarna, i enlighet med konventionen O*NET.

Dessa kompetenser ärver bedömningsvägen och beteendeankarna för detta konstrukt.

T	Kompetens	Definition	Prognos för efterfrågan 2026–2030
V	Verifying AI output AI output verification	Checks claims, figures and references from AI output against an independent source before use.	rising
V	Recognising hallucination Hallucination	Recognises invented facts, sources and quotations in AI output from internal contradictions and untraceable references.	rising
V	Checking AI source citations Citation checking	Opens and reads the cited sources and checks whether they exist and truly say what the model claims.	rising
V	Sampling AI output for accuracy Output sampling	Checks a random sample of large volumes of AI output and thereby estimates the error rate.	rising
V	Editing AI text to own standard Editing AI output	Rewrites AI text to own house style, audience and nuance and removes inflated or vague phrasing.	rising
V	Reviewing generated code Reviewing AI code	Reviews AI written code for behaviour, security, dependencies and readability before it is used.	rising
V	Setting quality criteria in advance Quality criteria	Records for an AI task what the output must satisfy so assessment does not happen on gut feeling.	rising
V	Recording the decision to use AI output AI use record	Records which AI output was used, who approved it and which adjustments were made.	rising
G	Countering over reliance on AI Automation bias	Keeps own professional judgement sharp, tests AI output against experience and voices doubt instead of going along.	rising