The teachers who get real value out of an AI assistant are the ones who already know what a good lesson looks like, because every judgment the tool cannot make falls back to them. AI skill is an extension of teaching skill. What is worth learning is less how to phrase a request than how to decide what to ask for, what to check, and what not to hand over at all. Those are decisions about your own practice, which is a different job from teaching students to use AI and turns on different things.

The skill is knowledge, verification, and refusal

Teachers who treat prompt-writing as the whole skill end up judging output by how confident it sounds, and confidence is the property AI produces most reliably. Prompting is a small procedural piece. The rest is subject knowledge, verification, ethical judgment, attention to privacy and bias, honesty about when AI was used, and the willingness to not use it at all.

Use AI like someone in the discipline would. A historian reads a claim and asks where it came from, who benefits from it, and what the record supports. A teacher of history working with AI runs the same reflexes over what the model hands back. Without them you have a fast writer with no accountability and no memory of your students.

Fluent use turns out to mean more reading. You read the draft against what you know, check the quotations, and notice the paragraph that would confuse the class that struggled with the same idea last week. The first version arrives sooner and you spend the saved minutes looking at it harder.

A 2026 systematic review of 28 empirical teacher–AI co-design studies found that teachers used AI mostly for ideation, lesson planning, prompt generation, and content production. Only a small portion of those studies showed AI working as a genuine partner in design. The recurring problems were generic output, misalignment with the actual curriculum, factual unreliability, bias, and opacity about how the system reached its answer, with efficiency crowding out the rest. The review’s recommendation is to keep questioning, filtering, and adapting, and to keep governance with the teacher.

UNESCO’s competency framework for teachers names five things a teacher needs: a human-centered mindset, ethics, AI foundations and applications, AI pedagogy, and professional learning. The last two are competencies in their own right there, not byproducts of the first three. The same framework warns that overreliance may weaken teacher agency and professional competency, which is the risk a fluent workflow makes hardest to notice. Punya Mishra and colleagues arrive near the same place from another direction. TPACK is the account of what a teacher holds at once about content, pedagogy, and technology. Their revision of it for generative AI widens the contextual half, to reach the social and institutional consequences of using the tool at all. The professional-development SEE framework reduces the same ground to a test you can run on the spot: is this use safe, ethical, and effective? It is one of several sorted in the guide to which framework answers which question. Each of the three asks what the teacher knows, which is the part no prompt supplies.

Hand over what you can check in a minute

Before anything else you sort the work, and the sort follows one rule: AI suits tasks whose output is provisional, quickly checked, and not a high-stakes judgment. Where any of the three fails, the task moves to the next tier or stays with you entirely.

teacher tasks, sorted by how much of your judgment they need

DRAFT Hand it over You can check it fast, and nothing high-stakes rides on the first version.
- Alternative examples and nonexamples
- Quiz and question variants
- First activity ideas
- The misconceptions a topic tends to produce
- Draft discussion questions
- First-pass differentiation
- Formatting
- Conversion into tables and organizers
- Draft feedback statements
- Several versions of the same directions
- Role descriptions for a simulation
- Preliminary rubric language
- Family-communication drafts
- Patterns in de-identified information you supply
- Counterarguments to a lesson you designed
- Adaptations of an activity you built
REVIEW Generate, then read it hard Worth generating and never worth accepting as it arrives.
- Lesson sequencing
- Unit design
- Assessment items
- Exemplars
- Reading-level adaptations
- Translations
- Culturally responsive examples
- Accommodations
- Source packets
- Scoring assistance
- Recommendations drawn from student work
KEEP Keep it yourself Final responsibility the tool cannot hold.
- Curriculum goals
- The intellectual sequence of a unit
- Source selection and validation
- High-stakes grades
- Diagnosis of student needs
- Special-education decisions
- Discipline
- Safeguarding
- Counseling
- Culturally or politically sensitive judgment
- Parent communication carrying conflict or consequence
- Determinations of cheating
- Any decision that materially affects a child’s educational opportunity

The review bin is where the damage hides. Its failure mode is not a visible error but a lesson sequence that is coherent, plausible, and pointed at a slightly different objective than the one you wrote down. Substantial review is the price of generating one at all.

The keep bin is the one worth being stubborn about. Nobody minds a machine drafting a first pass at rubric language. The call on whether a student gets a second attempt is the job, and delegating final responsibility there hands over an accountability the tool has no way to carry.

That sort covers your own work. Whether students may use AI is a separate question, and five questions settle whether AI belongs on a given assignment at all.

Write the brief before you write the prompt

The AI generates alternatives, questions, and critiques; you hold the goals, the sequence, the factual integrity, and everything you know about the students in front of you. That stance is the critical collaborator, and it is what the emerging teacher–AI co-design literature argues for and mostly does not yet find. One polished answer has nothing to be judged against, so you end up judging it by how it reads. Three give you something to choose between.

critical-collaborator-workflow

1 Write the design brief The standard, the objective, what you will accept as evidence of mastery, the prior knowledge, the misconceptions you expect, the required sources, the reading level, the cultural and ethical constraints, and what the tool must not do. sets everything after it
2 Ask for alternatives, then make it argue with itself Three lesson openings, two contrasting explanations, competing assessment approaches. Then ask what the draft assumes about prior knowledge, where it drifts from the objective, and what would work better with no technology at all.
3 Verify Facts, quotations, citations, standards alignment, reading demands, developmental fit, accessibility, cultural framing, safety, and whether the assessment measures what it claims to.
4 Localize What the system cannot know: your class history, the relationships in the room, the misconception a few students hit last Tuesday, the community you teach in, the materials you have, where you are in the unit.
5 Pilot one component Test a discussion protocol or a question set before redesigning a course around it, then ask whether students learned more on their own, whether their time with you and each other went up or down, and where the saved time went.

An EEF/NFER randomized trial with 259 teachers across 68 English secondary schools studied ChatGPT-supported planning in Key Stage 3 science. Teachers given access and guidance spent 56.2 minutes a week on the target planning tasks against 81.5 for the control group, a drop of about 31%. A blinded expert panel detected no apparent fall in resource quality, and those teachers generally used AI for one or two components — quiz questions, activity ideas, examples — rather than whole lessons. That is the argument for keeping a pilot narrow.

Writing down what you will accept as evidence of mastery keeps the rest of the sequence pointed at your own unit. It is also the easiest step to skip, because the tool will answer without it.

Every output is a draft until you check it

Prompt-writing as the whole of AI literacy is one framing that misleads, and five more show up alongside it. Each costs something you can name.

The framings that mislead

Treating AI as an oracle hides the teacher’s responsibility for curriculum and invites acceptance of plausible unverified output.
Calling it a neutral tool ignores that its answers depend on training data, system design, provider rules, prompting, and processes nobody outside the provider can see.
Counting it mainly as a time-saver measures the wrong thing, since time saved is worth having only where instructional quality held and the hour went somewhere useful.
“Students will use it anyway” mistakes inevitability for a pedagogical reason, and students meet calculators and search engines without every task being built around those.
Putting the options at prohibition or unlimited access skips the middle ground that the Four Modes of AI Assignments works out, one assignment and one phase at a time.

The framings that hold up

Every output is provisional: a draft, a possibility, or a hypothesis.
Deciding the cognitive role before you ask gets you a critic, a coach, a translator, or a practice partner instead of a general-purpose answer machine.
The participant is fallible, and you stay responsible for the truth and the consequences of the work either way.
Support fades as competence rises, for you as much as for students.
Human relationships are the instructional infrastructure, and the reason to hand off a rubric draft is that discussion, observation, mentoring, and feedback are better uses of the hour.

Kindred K-12 is built around this workflow

You bring the objective and the expertise; the assistant in Kindred K-12 returns options, and you verify and adapt what comes back. It builds the materials for a lesson and reworks the sources you already teach with. It earns you time back for the parts of teaching that need a person in the room.

Kindred K-12 sits on the teacher’s side of the work: no student accounts, no student personal information, and the assistant never chats with students. What you make with it does reach students, and what it says is your call.

If this is the kind of thing you want to keep thinking about, the Kindred K-12 newsletter carries more of it.