The Perfect Day of an Educator With AI

01What you will learn

• What a school day looks like when AI takes over arranging materials and paperwork and you stay with the children.
• Which task is done with what — and where the data goes: 🔒 local or 🌐 global.
• Where AI has no place: in the classroom, in marking, in an alert, in a conversation with a parent.
• How to start your own “twin” — the one-page entry and the decision card.

02Before you start

• Take the lesson “Do No Harm”, or at least read the four rules above.
• A list of the materials and documents you prepare every week — worksheets, adapted texts, progress notes, plans, letters to parents.
• A rough estimate: how many hours a week they take. Write the number down — you will need it in a month.
• Know what your school allows for AI tools and what its internal rule is for pupils’ data. If there is no rule, ask the head teacher before you begin.

03Steps

1. Recall

   Without looking at your timetable, write down the three things that eat most of your time outside lessons with the children. Almost always they are: preparing several versions of the same material, rewriting the same thing in plans and notes, and letters to parents. Keep them — the day below is built around exactly these.

2. One new idea

   AI writes the structure; you write the meaning. The machine is good at arranging and at variants: the same text with shorter sentences, a word list, a draft letter, a day plan. It is not good at — and must not — knowing the child, assessing the child or deciding what the child needs. Divide the work this way and materials and paperwork get faster, while the responsibility stays where the law puts it: with the educator and the team.

3. Worked example: the day hour by hour

   📎

   Example day — fictional

   A primary-school teacher with a class who also teaches some resource-support hours in a fictional school. The children (“K-04”, “K-11”) and the cases are made up. The day illustrates the approach; it is not a measured case — we do not promise saved hours; they are measured in your own process.
   07:30 – 08:00

   The start: the day on one page, and the materials

   Without AIThe timetable lives in three places, and the text for the lesson is one for everybody, although two pupils need an easier version. The evening before goes on rewriting.

   With AIThe local model turns your timetable into a day plan — with codes only. A global model reworks a public or your own text into shorter sentences and adds a word list — with no data about any child. You choose which version is for whom: only you know that.

   🔒 timetable with codes🌐 material without data

   08:00 – 12:30

   In the classroom and the support room: people only

   Without AITeaching, observing, supporting on the spot.

   With AIThe same. AI does not enter the classroom. No recording, no camera, no “attention dashboard”. Children’s marks and behaviour are not handed to a machine.

   ⛔ no AI

   12:30 – 13:00

   The note, straight away

   Without AIObservations stay in your head until evening — and get lost.

   With AIYou write by hand or dictate on the work device: code, fact, not interpretation — “K-04 did not start the task for 6 minutes; started after a picture prompt”, not “K-04 is lazy”. The next step then has good raw material.

   🔒 notes with codes

   13:00 – 14:00

   Paperwork: structure from the machine, meaning from the team

   Without AIThe support plan or report is written from scratch or from an old file — with a risk that someone else’s name stays in.

   With AICoded notes go to the local model, which returns the structure of the plan or report. The assessment of individual needs and the conclusions are the work of the support team appointed by the head — not of a machine and not of one person (Art. 188–189 of the Pre-school and School Education Act; Art. 70, 78 of the Ordinance on Inclusive Education). Health data is a special category (Art. 9 GDPR). A colleague from the team reads before it goes on file — four eyes.

   🔒 notes with codes

   14:00 – 14:45

   Parents and the team

   Without AIEvery invitation and every letter from scratch, hunting for the right tone.

   With AIA general model writes a template with blanks: “[date]”, “[time]”, “[pupil code]”. You fill them in on your own computer. The conversation with the parent is yours: the support team works together with the parents (Art. 188 para 5 of the Act).

   🌐 template without data🔒 filling in

   14:45 – 15:15

   The case that is not on the plan

   Without AIA pupil shows marks on their body or says something worrying about home.

   With AIThe same — and AI stays closed. You write down facts in the child’s own words, not interpretations. Anyone who learns that a child needs protection must immediately notify the Social Assistance Directorate, the State Agency for Child Protection or the Ministry of Interior — and this holds also for what you know through your profession, even if you are bound by professional confidentiality (Art. 7 of the Child Protection Act). In immediate danger — 112. The head is informed by the procedure your school has adopted ⚠️ (check it with your head — the internal procedure is not verified in this lesson).

   ⛔ no AI

   15:15 – 16:00

   Team work and learning

   Without AIThe team meeting is postponed because “there is no time”.

   With AIA general model helps you arrange your questions to colleagues — generalised, with no child who could be recognised. Team work and sharing good practice are part of general support in the Act (Art. 178 item 1 and Art. 179).

   🌐 generalised questions only

   16:00 – 16:30

   The twin: 15 minutes of arranging

   Without AIWhat was learned today stays in your head — until tomorrow or until your leave.

   With AIYou update the one-page entry and record one decision card (“how we solved it”) — no personal data. AI helps you arrange them. Tomorrow you and the colleague who covers for you start from there.

   🔒 your memory

   16:30

   The end

   Without AIExercise books, sheets and notes go home.

   With AINothing about pupils goes onto a personal device. The freed time is for the children — and for life after 16:30.

   The division of labour

   ⚙ AI takes

   • A day plan from your timetable (with codes)
   • Variants of a public or your own text — shorter sentences, a word list
   • The structure of a plan or report from coded notes
   • Templates for letters and invitations with blanks
   • Arranging your own notes and templates

   ★ The educator keeps

   • The lesson, the observation, the child
   • Every assessment of a pupil
   • The assessment of needs and the support decisions — the team, together with the parent
   • Conversations with parents
   • Alerts and entering data into official systems
   ⛔

   Where AI has no place

   Marking and ranking. AI does not write marks, does not suggest them and does not decide access to education or the level of education. Systems that do this are among the high-risk ones under the AI Act (Annex III, item 3).
   Diagnosis. Whether a child has dyslexia, ADHD or anything else is not a question for a chatbot. It is a matter for specialists and for the team’s assessment.
   Recording and tracking. No recording of lessons, no camera, no attention analysis.
   Alerts and parents. Neither the alert nor the conversation with the parent goes through the machine.

   Which task with what

   Task	With what	Mode
   Day plan	Model on the work computer, your timetable with codes	🔒
   A variant of a lesson text	General assistant — only a public or your own text, no data about children	🌐
   Structure of a support plan / report	Model on the work computer; a global one only under contract and after an impact assessment	🔒
   Invitation or letter to a parent	General assistant — template with blanks	🌐
   Assessment, diagnosis, alert, conversation with a parent	No AI	⛔

4. Guess → try → change

   Pick one text you use in class — your own, or a public one you have the right to use.

   1. Guess: how many minutes does it take you to make an easier version by hand? Write it down.
   2. Try: give a general AI assistant the text with the request below — without a single detail about a child. Read the result word by word: the machine slips in the details.
   3. Change one thing: ask for even shorter sentences or fewer words in the word list. Compare the time with your guess.
   Request for an easier version (no personal data)copy

   Retell the following text for pupils in grade 3 using shorter
   sentences (up to 12 words) without changing the facts.
   Then write 4 comprehension questions: two yes/no, one
   multiple-choice with three answers, and one open question.
   Add a word list with 5 words. Do not invent new facts.
   Text: [paste YOUR text, or a public one you may use]

5. Make it yours

   Start your twin — the arranged external memory of your work, which both you and the colleague who covers for you can read. Two templates, one page each. Not a single child in them.

   The one-page entrycopy

   THE ONE-PAGE ENTRY
   Who I am (3 sentences):
   What I do (the class, the subjects, resource support):
   My rules (up to 10, each with one word of why):
     1. Nothing personal into a global AI model — protection
     2. ...
   Where everything is (what → where):
     plans and reports → ...
     materials by topic → ...
     templates for letters to parents → ...
   How to work with me (how to hand me work, when to ask me):

   Decision cardcopy

   DECISION CARD
   Title:            Date:          Version:
   What we decided:  Notes on pupils are by code only; AI never sees names.
   Why (up to 5 lines):
   What it replaces → where the successor is:
   Who approved:

6. Explain to someone else

   Show a colleague one of your materials and the rule “AI writes the structure, I write the meaning”. Ask them to find one place where the machine could see something about a child that it should not. If they find one, fix it together. That is already four eyes.

7. Come back after a week

   • Which task got done faster this week — and what did you check in the result before using it?
   • Could a colleague start work with only your one-page entry?
   • How many hours went on documents and materials? Compare with the number from “Before you start”.
   📊

   Measuring without tracking

   We measure hours, not people and not children: how many hours a week go on documents and materials — before and one month after, self-reported. No pupil names. More: SP-02 · Measuring without tracking.

04Check

05What next

06Sources

1. Pre-school and School Education Act (Bulgaria) — Art. 177 (general and additional support are provided where the child is), Art. 178 item 1 and Art. 179 (team work between teachers and other pedagogical specialists), Art. 187 (additional support includes work on a specific case and resource support; its type and hours are set in a support plan), Art. 188 (assessment of individual needs by a support team appointed by order of the head; the team must include a psychologist or pedagogical counsellor and a speech therapist; it works together with the parents), Art. 189 (team functions: assessment, plan, monitoring). A copy of the consolidated text published online was reviewed (01.10.2026). ⚠️ The Act has been amended many times — check the numbers against the current edition.
2. Ordinance on Inclusive Education (Bulgaria) — Art. 70 (additional support is provided on the basis of an assessment by the team), Art. 78 (the assessment is functional), Art. 79 (a parent may request a repeat assessment in case of disagreement). The text in force from 27.10.2017 was reviewed (01.10.2026). ⚠️ Check against the current edition.
3. Child Protection Act (Bulgaria) — Art. 7 (duty to notify immediately; it applies also to what is learned in connection with one’s profession, even under professional confidentiality) (checked 01.10.2026).
4. Regulation (EU) 2016/679 (GDPR) — Art. 9 (special categories of data, incl. health), Art. 35 (impact assessment).
5. Regulation (EU) 2024/1689 (AI Act) — Annex III, item 3 (education: systems to determine access or assign, to evaluate learning outcomes, to assess the appropriate level of education and to monitor prohibited behaviour during tests are high-risk). Requirements for Annex III systems apply from 2 December 2027 (after the amendment by Regulation (EU) 2026/1744; checked 2026-10-01).
6. SP-01 · Do No Harm — the four rules and the alert protocol, with all norms.

This lesson is not legal advice. Norms are given with a short meaning; for a specific case turn to your head teacher, the school’s legal adviser or the regional education authority.