The worry, and what it gets right
Teachers who fear that AI will replace them are not being naive. They have watched software take over tasks in other professions, and they have been told, sometimes by vendors, that a tool can mark, plan and tutor. The worry is worth taking seriously, because the way a school introduces AI decides whether it is felt as help or as pressure.
It is also possible to say something reassuring that is true. The parts of teaching that matter most, knowing a particular class, noticing the child who has gone quiet, deciding what to do next, are the parts AI does not do. The parts it can do are the parts most teachers would happily hand over. The design question is how to make sure only the second kind moves.
What AI takes off a teacher's plate, and what it cannot
A clear division helps a staffroom more than a general reassurance does.
| Teacher work | AI's part, and the teacher's part |
|---|---|
| Planning a lesson | AI: a first draft of structure, activities and questions. Teacher: fits it to this class and this week. |
| Making resources | AI: worksheets and quizzes with an answer key. Teacher: checks accuracy and level before anyone sees it. |
| Giving feedback | AI: a draft from the teacher's notes. Teacher: what the student hears, and how. |
| Marking | AI: at most a suggestion on some question types. Teacher: the mark. |
| Writing to families | AI: a draft. Teacher: reads, corrects and chooses to send. |
| Knowing the class | AI: nothing. This is the job. |
Five rules that keep the teacher in charge
Schools that introduce AI well tend to write these down before the tool spreads.
- A teacher owns every use: someone can say who asked for it and who approved it
- Draft, do not decide: AI output is an input to judgement, never the judgement
- Nothing leaves unread: no student or family sees output a person has not checked
- Every use is visible: requests are recorded and the school knows what the tool did
- Measure time returned: judge success by hours given back to teaching, not tasks automated
Marking and feedback: where to draw the line
Feedback drafted from a teacher's own notes is a low-risk use, because the teacher supplied the substance. Marking is different, because a mark is a judgement that affects a child and carries the school's authority.
The safe arrangement is a suggestion that becomes a mark only through a deliberate act by a named teacher, with the tool declining to score work it cannot read rather than guessing, and with moderation continuing as before. Anything that lets a score flow into a report without that act has removed the teacher from the loop, however it is described. The wider set of guardrails for staff-facing tools is in teacher AI assistants for schools.
Student-facing AI: owned by a teacher or not at all
An AI that talks to children with no adult in the room needs stricter rules than one that drafts for staff. The questions to settle before a school allows one are who created it, who can switch it off, what it is told to do when a student asks it to do their work, who can read the conversations, and what happens when a student says something that suggests they are in distress.
The safe answers are that a teacher creates and activates each one, that its instructions are to teach rather than answer, that staff can read every exchange, and that anything worrying goes to a person rather than being handled by the model. Those are design choices, so ask to see them working before agreeing.
How skodle helps
This section names our own product, so weigh it accordingly. The rules above are the test to apply to us as much as to anyone.
In skodle, each use has a teacher behind it. Lesson plans, worksheets, quizzes and feedback are drafts that a teacher saves; a suggested score on a short-answer, long-answer or rubric question becomes a mark only when a named teacher accepts it, and a blank or unreadable answer gets no score rather than a zero; a family email goes out only when someone presses send; and a student tutor is created and switched on by staff, cannot be created by students, is instructed to teach rather than hand over finished work, is readable by staff, and passes a narrow set of distress phrases to a person. On the leadership side, copilot drafts on higher-stakes topics need a school leader to approve them, each request is recorded, and staff can mark an answer helpful, unhelpful or wrong. AI is included in every plan, and every new school starts with a free 30-day trial, so teachers can judge it on their own planning: start the free trial. See skodle AI and the platform overview.
The limits, plainly. A tutor's instruction to teach rather than answer is guidance to the model, not a guarantee, which is why staff can read the conversations. Report comments are drafted from the notes a teacher supplies, not from a per-student reading of the gradebook, and the lesson planner does not read the school's curriculum map. Requests, and any records a tool retrieves, are processed by an external model provider, and there is a daily usage limit that depends on the plan. skodle is not an AI detector, does not predict grades, and does not do a teacher's work automatically: it drafts, and a person decides.
What to promise staff, in writing
Reassurance lands better as a short written commitment than as a speech. A school can decide, for example, that AI will not be used to make or draft decisions about the performance of staff, that records of how the tools are used exist for safety and support rather than for ranking teachers, that a teacher can decline a drafting feature without it counting against them, and that students will be told when AI has been used in something they receive.
These are choices for a school to make, not features of any product, and they cost little. What they buy is the trust without which a tool goes unused in the staffroom whatever it can do.
Training that helps rather than alarms
The training that works is hands-on and specific to the subject. Give a department an hour to use the tool on a real lesson they are about to teach, ask them to find where the draft is wrong, and share the failures openly. A teacher who has spotted a confident error in a draft trusts their own judgement more, and the tool less blindly, which is the outcome a school wants. Planning is the natural first task; see AI lesson planning tools for teachers for what to check in a draft.
Measuring success in returned time
Ask a small group to time a routine planning or feedback task before and after, including the time spent checking, and to note how often a draft was usable with light edits. Ask what they stopped doing, and whether they used the time for students. If the answer is that they simply produced more paperwork, the tool has not helped, however many tasks it automated.
Signs that AI is replacing judgement
Watch for these, and treat any of them as a reason to stop and review.
- Marks or comments reaching students or families that no teacher has read
- Staff unable to say who approved a piece of AI output
- Teachers adjusting their judgement to match what the tool suggested
- Usage records being used to compare teachers
- Students receiving AI answers to work they were meant to think through
