# AI Fluency for educators: teach judgement, not just prompts

> A guide to designing educator AI training around pedagogical judgement, verification, delegation boundaries and responsible practice.

- Author: Viktor Berthelius (BRTHLS)
- Published: 2026-09-03
- Category: ai operating models
- Tags: education, ai-fluency
- Language: en
- Canonical: https://www.brthls.com/magazine/ai-fluency-educators-judgement-en
- Source: BRTHLS Magazine — https://www.brthls.com

---

## Problem

Much AI training for educators begins and ends with a collection of prompts. That may help produce a first activity, but it does not prepare someone to decide when to use AI, how to review its output or which responsibility must never be delegated.

## Thesis

AI Fluency is decision capability, not writing speed. A fluent educator can state a pedagogical intention, choose an appropriate role for AI, verify the output and explain its limits to learners.

## Framework

Develop four layers of competence:

- **Intention:** what learning should occur and why AI adds value.
- **Interaction:** how to provide context, constraints and useful examples.
- **Verification:** how to check facts, bias, suitability and traceability.
- **Responsibility:** which decisions and relationships remain human.

| Situation | AI may | The educator retains |
| --- | --- | --- |
| Preparing activity variants | Propose drafts | Objective and suitability |
| Giving formative feedback | Detect patterns | Interpretation and conversation |
| Assessing learning | Organise evidence | Final judgement and right to reply |

## Why it matters now

UNESCO frames teacher competence across five dimensions, including a human-centred mindset, ethics and pedagogy. DigCompEdu also goes beyond technical skill by connecting resources, teaching, assessment and learner capability. Article 4 of the EU AI Act adds a contextual AI-literacy responsibility for providers and deployers.

## Anti-example

A workshop teaches twenty “perfect” prompts without examining data, errors or assessment. Educators leave with recipes that expire when the tool changes, but no method for making decisions.

## Protocol (3 steps)

1. **Start with a pedagogical decision:** not a product feature.
2. **Practise through contrast:** compare acceptable, doubtful and unacceptable outputs.
3. **Close with traceability:** record what AI did, what the person verified and what the learner learned.

## Related

- [What a startup should not automate yet](/magazine/what-not-to-automate-startup-en)
- [The 14-day AI experiment for a startup](/magazine/14-day-ai-experiment-startup-en)

## Sources consulted

- [UNESCO AI Competency Framework for Teachers](https://www.unesco.org/en/articles/ai-competency-framework-teachers)
- [European Commission DigCompEdu](https://joint-research-centre.ec.europa.eu/projects-and-activities/key-competences-lifelong-learning/digcompedu_en)
- [EU AI Act, Article 4](https://eur-lex.europa.eu/eli/reg/2024/1689/oj?locale=en)

## Next step

Redesign one real activity: define the learning first, then AI's role, and finally the evidence that will show whether the intervention helped.

---

_Cite as: Berthelius, V. (2026). "AI Fluency for educators: teach judgement, not just prompts". BRTHLS Magazine. https://www.brthls.com/magazine/ai-fluency-educators-judgement-en_