AI Good Practice Stocktake

A reflective development tool to assess how deliberately, critically and responsibly you use AI in professional work.

Your use of AI

These questions establish the breadth, complexity and professional exposure of your current practice. More extensive use is not treated as more mature.

Name, optional Current pattern of professional AI use Information used with AI Consequences if an AI-assisted output is wrong How independently does AI act in your work?

What do you currently use, or expect to use, AI for?

Select all that apply. Deliberate non-use is a valid position.

Who or what could be affected?

These contextual signals contribute to professional exposure; they do not imply that the use is inappropriate.

Please complete the profile and select at least one option in each group.

Please answer every item. Choose “Not relevant” where appropriate.

Professional judgement in practice

Choose the response closest to what you would probably do under real constraints—not the response that sounds most ideal.

Please choose one response for each situation.

Your AI Good Practice Stocktake

Your result highlights current good practice, areas to strengthen and risks that may need more immediate attention.

Your practice assessment

Disclaimer: This tool does not certify competence or replace legal, professional, safeguarding, security or information-governance advice

Your good-practice dimensions

These bands describe how consistently your answers indicate good practice; they are not a measure of professional seniority.

What needs attention—and what to do

What you are doing well

Trustworthy-AI practice lens

The EU framework also includes societal and environmental wellbeing. This stocktake considers effects on people and wider society, but does not assess environmental impact in enough depth to provide an individual result.

This indicative view relates your answers to the European Commission’s requirements for trustworthy AI. It is not a system assessment or confirmation of legal compliance. View the EU Assessment List for Trustworthy AI (ALTAI).

Risk signals

This section brings together practical risk and regulatory signals from your answers. It identifies what could make your AI use more consequential and what action is proportionate now.

Good-practice risk

Why this signal appears

EU AI Act signal

Read about the EU AI Act.

The good-practice signal adapts the NIST AI Risk Management Framework: your professional context is mapped, possible harms and safeguard gaps are considered, and a proportionate response is suggested. It is not a formal NIST assessment. View the NIST AI Risk Management Framework Core.

Turn the result into a professional practice plan

Use the result to set one clearer boundary, one checking habit and one purposeful experiment. It can also support coaching, supervision, CPD or a team discussion.

The Making Sense of AI workshop provides a practical foundation in critical and responsible AI use.

AI@andynobes.co.uk

Admin and methodology

Model notes

AI Good Practice Stocktake — prototype edition. The questions, scoring rules and practice assessments are an original synthesis for reflective professional development. Practice quality and professional exposure are interpreted separately.

The four practice assessments describe current patterns of reported behaviour. They are not professional grades, certifications or measures of workplace seniority.

The trustworthy-AI lens is an indicative crosswalk to six requirements that the questionnaire can meaningfully assess. Societal and environmental wellbeing remains outside the scored result.

The risk signal adapts the NIST AI Risk Management Framework’s logic of establishing context, identifying and assessing risks, and prioritising proportionate action. It is not a formal NIST assessment.

Data, privacy, security, integrity, transparency, accountability and professional-compliance questions are broad prompts. Their relevance depends on the person’s jurisdiction, employer, profession and use case.

The cognitive-autonomy dimension combines established cognitive-offloading theory with newer GenAI research on confidence calibration, task stewardship, performance versus learning, metacognitive regulation and scaffolded rather than substitutive use. It is not a measure of cognitive decline.

The situational questions add applied judgement to self-report, but they remain illustrative and have not been validated as a performance test.

Only sources that materially influenced the dimensions, practice assessment, framework lenses, risk interpretation or wording are listed below. Wider literacy and competency frameworks were reviewed during development but are not presented as direct foundations of this version.

Sources directly informing the model

European Commission High-Level Expert Group on AI, Ethics Guidelines and ALTAIDirectly informs the trustworthy-AI practice lens covering human agency, robustness, privacy, transparency, fairness and accountability.
EU assessment list

NIST AI Risk Management FrameworkDirectly informs the contextual risk signal through its Govern, Map, Measure and Manage functions.
NIST AI RMF Core

Skills England (2026), AI foundation skills for workUsed as a coverage check for basic technical, non-technical, responsible and ethical workplace practices.
GOV.UK benchmark

OECD AI PrinciplesDirectly informs questions on human agency, fairness, transparency, robustness, security and accountability.
OECD principles

UK Information Commissioner’s Office, AI and data protection guidanceDirectly informs the privacy, data-security, transparency and human-oversight prompts.
ICO guidance

Risko & Gilbert (2016), Cognitive OffloadingProvides the general theoretical basis for deliberate externalisation, metacognitive monitoring and control.
PubMed record

Lee et al. (CHI 2025), The Impact of Generative AI on Critical ThinkingDirectly informs confidence calibration, verification, response integration and task stewardship in professional work.
Microsoft Research paper

Yan, Greiff, Lodge & Gašević (2026 manuscript), Distinguishing performance gains from learningDirectly informs the distinction between an improved immediate output and retained learning or transferable capability.
Manuscript

Li, Cui & Hagedorn (2026), The cognitive impact of ChatGPT in higher educationDirectly informs items on metacognitive regulation, active evaluation, scaffolded use and the risk of unstructured cognitive substitution.
Systematic review

AI Good Practice Stocktake — prototype