How you are graded

Every criterion is a met or not-met question that must cite evidence. Your level on an axis is the highest band whose criteria are all met, counting up from Foundation without gaps.

Rubric version aiq-rubric-2026.11.2-en · Based on AIQ Research, sections 3.7–3.8

Foundation100–399Uses AI within given guidelines and does not leak data

Practitioner400–599Uses AI end to end for routine work and checks the output

Advanced600–799Designs multi-step, multi-tool work and sets quality bars for a team

Expert800–900Builds systems and rules others use, with testing and oversight

1. Delegation & briefing

Level	Criteria
Foundation	Evidence of telling which tasks suit AI and which do not Evidence of writing prompts that state the goal and the output format
Practitioner	Prompts or briefs include context, constraints, examples or quality criteria Iterates on prompts until the output is usable, without supervision
Advanced	The participant designed a reusable template or prompt set that others can useOwn design work only Turns an ambiguous request into a multi-stage brief with check or stop criteria
Expert	The participant set a prompting standard that a team or organization actually usesOwn design work only Teaches or assesses others against that prompting standard

2. Discernment & verification

Level	Criteria
Foundation	Evidence of catching obvious errors in AI output Knows which output needs checking before use, such as numbers, names or citations
Practitioner	Systematically checks facts against the original source Compares several outputs against task criteria, or logs what was corrected
Advanced	The participant built a checklist or rubric for judging AI output for a type of workOwn design work only Finds hidden errors in output that looks right, or assesses bias or provenance
Expert	The participant designed a test set to compare several systems or toolsOwn design work only Uses quality data systematically to change how the team works

3. Workflow design

Level	Criteria
Foundation	Uses AI for a single step inside an existing workflow Can split a task into the parts a person should do and the parts AI should do
Practitioner	Plans their own workflow from input to delivery, with AI in several steps Marks where a human must decide or review in the workflow
Advanced	The participant designed a workflow that chains several kinds of toolsOwn design work only Decides not to use AI, or picks a tool, with cost and risk reasoning
Expert	The participant handed over a workflow others can follow, with owners and a fallbackOwn design work only Makes policy-level calls on which work to automate, augment or keep human

4. Responsibility & data safety

Level	Criteria
Foundation	Does not put confidential or personal data into public AI tools Discloses AI use when appropriate
Practitioner	Classifies data sensitivity and masks it before giving it to AI Identifies approval points or assesses copyright or bias risk in own work
Advanced	Reviews someone else's workflow for data leaks or privacy-law issues and proposes fixes The participant designed access control or credential handling for an AI systemOwn design work only
Expert	The participant wrote team- or unit-level guidance on using AI with dataOwn design work only Sets up an audit trail or an accountability process for AI mistakes

5. Automation & governance

Level	Criteria
Foundation	Can explain which repetitive work could be automated Can separate trigger, input, output and the step a person must do
Practitioner	Writes a spec for an automation: input, conditions, output and stop points Builds or configures an automation and tests the happy path
Advanced	The participant designed multi-branch logic with validation or retryOwn design work only The participant defined test cases and human approval pointsOwn design work only
Expert	The participant runs an automation others rely on, with monitoring and a failure planOwn design work only Has a handover guide so others can maintain the system