The Framework

How much an educator uses AI, and how wisely. Two questions, twelve positions.

Reading the grid

Engagement runs up the grid: how much an educator uses AI, from Peripheral at the bottom to Generative at the top.

Discernment runs across: how wisely, from Receptive on the left to Prudent on the right.

Each cell pairs a stage with a level. 3B, for example, is Integrated and Critical.

Engagement: the rows

  1. Generative

    AI is something the person builds with: custom tools, configured context, and new workflows that extend what they can produce.

  2. Integrated

    AI is part of recurring workflows. It is a routine partner in core work like communication, grading, and content preparation.

  3. Functional

    AI handles discrete tasks on demand, such as drafting a message or looking something up, but it is not built into how the work gets done.

  4. Peripheral

    AI sits at the edge of the work. Use is occasional or exploratory, and no regular task depends on it.

Discernment: the columns

  1. Receptive

    Output is trusted by default. Little scrutiny is applied, and understanding of AI's limitations is thin.

Critical vs. Prudent

Critical

Output is scrutinized after it is produced: compared against a source and checked before use. Discernment as hygiene.

Prudent

Judgment comes first: whether AI belongs in the task, how it should be used, and what must stay human. Discernment as architecture.

When judgment happens

The columns differ in when judgment happens, not in how much effort it takes. Two people can check output the same way and still sit in different columns.

AReceptive
  1. Before the task
  2. Using AI
  3. Output
  4. Before it's used

Output is trusted by default. No checkpoint.

BCritical
  1. Before the task
  2. Using AI
  3. Output
  4. Before it's used Is it accurate?

One checkpoint, after the output. Discernment as hygiene.

CPrudent
  1. Before the task Does AI belong here? How? What stays human?
  2. Using AI
  3. Output
  4. Before it's used Is it accurate, and is it fitting?

Judgment comes first, and checking follows. Discernment as architecture.

Same task, three levels

Hold Engagement constant and the columns separate on judgment alone. Each example uses AI about as often; what differs is whether, when, and how the person decides.

Teacher

Feedback on a stack of student essays

  1. A Receptive

    Pastes each essay with the rubric and returns the comments as written.

  2. B Critical

    Generates comments, then reads each one against the essay and rewrites what misses.

  3. C Prudent

    Decided beforehand that the first read and the grade stay hers. Writes her own comments, then uses AI only to tighten the wording.

Counselor

A message to a family about a student's falling grades

  1. A Receptive

    Asks AI for a draft and sends it.

  2. B Critical

    Drafts with AI, checks every fact against the gradebook, and softens the tone before sending.

  3. C Prudent

    Decided this message should come from her directly. Uses AI only to pull the grade history together beforehand.

Administrator

Summarizing a 60-page accreditation guidance document for the leadership team

  1. A Receptive

    Circulates the AI summary as given.

  2. B Critical

    Checks the summary section by section against the document and corrects two misreadings.

  3. C Prudent

    Reads the standards sections himself, because the team will be held to them. Uses AI to summarize the appendices and draft a reading guide.

Growth

  • Up · Engagement Growing engagement mostly requires time and permission: try things, set up context layers, experiment with prompting.
  • Right · A → B: Hygiene Learn scrutiny. Compare output against a source.
  • Right · B → C: Architecture Learn to ask the harder questions in advance. Study AI capabilities and limitations.

Grow right before up. Advisory, not a rule. Where you can, build discernment before you add engagement. Checking and prior judgment are easier to learn at lower volume, and they tend to slip when volume grows first. Engagement grows mostly with time and permission; discernment has to be learned on purpose.

Common trajectories

People move on the grid over time. Four patterns recur.

Starting cell Sound growth Risk path

Two toolkits

Discernment grows in two moves, and each has its own practices. Your profile adds practices specific to your position.

A → B

Hygiene

Learn scrutiny. Compare output against a source.

“Is it right?”

  • Source-first check. Before anything AI-drafted goes out, open the original (handbook, calendar, gradebook, email) and compare the key facts.
  • The three-fact rule. Find the three facts that would matter most if wrong, and verify those every time.
  • Know the failure modes. Invented sources, outdated information, misread context, and a confident tone on uncertain claims.
  • Read as the recipient. Reread the output as the person receiving it. Would anything confuse or mislead them?
  • Audit after the fact. Now and then, check five outputs you have already used. Count the errors.

B → C

Architecture

Learn to ask the harder questions in advance. Study AI capabilities and limitations.

“Should AI be here at all?”

  • The pre-task pause. Before opening the tool: What is this task for? What must stay mine? What would I lose by handing off the first draft?
  • A stays-human list. Name the kinds of work AI won't draft for you, with a reason for each. Revisit it each term.
  • Task triage. Sort your work into what AI handles, what it assists with, and what stays human, before the tasks arrive.
  • Think first, then prompt. Form your own position in a few lines before asking. Compare it with what comes back.
  • Study capabilities and limits. Keep current on what the tools can and can't do, so boundaries rest on knowledge rather than assumption.

Task triage, illustrated

  1. AI handles With a designed check
    • Reformatting a schedule
    • First-pass summary of a public document
    • Practice items you will review
  2. AI assists You lead, AI supports
    • A newsletter you will revise
    • Organizing meeting notes
    • Brainstorming discussion questions
  3. Stays human Decided in advance
    • The first read of student work
    • Discipline and pastoral conversations
    • Recommendation letters
    • Staff evaluations
Illustrative only. Each person and school will draw these lines differently. The point is drawing them in advance.

Engagement needs no toolkit of its own. It grows mostly with time and permission: schools provide both, and individuals use them.

Offload and debt

Handing work to AI is not the problem. What gets handed off is.

Output Capacity, judgment kept Capacity, judgment handed off Time using AI →
Conceptual, not measured.
Offload
Handing AI a task while keeping the judgment the task requires. Capacity is preserved because the person still decides and still checks.
Debt
Handing off the judgment itself. Capacity erodes quietly, and the cost comes due later: when the tool is wrong, unavailable, or the skill is suddenly needed.
  • Offload is a design choice. Debt is a default.
  • Debt accrues fastest where use is high and judgment is thin: 3A and 4A.
  • The Prudent column is how offload stays offload.

Risk

Integrated · Receptive

3A: Capable but Uncritical

Integrated practice without meaningful checks, and thin understanding of AI's limitations. Some are silent verifiers who check but don't describe it.

The primary institutional risk by prevalence. If an employee is customer-facing, AI discernment is non-negotiable.

Peripheral · Prudent

1C: Passive Caution

Low engagement by deliberate choice, often with a stated position. This posture can serve as ballast when the conviction is rooted in truth and not merely avoidance.

Risk: stagnant growth as the broader culture develops. Tradeoff: opportunity cost as tools become more capable.

Generative · Receptive

4A: Active Risk

Enthusiastic adopters who fail to set appropriate boundaries. Heavy AI use without understanding or forethought, with output trusted by default.

If adoption is over-rewarded culturally, the pressure leads here. Cognitive offload becomes cognitive debt.

Common misreadings

Is 4C the goal for everyone?

No. Generative use isn't required, and 2C or 3C can be the right place for a given role. The goal is the C column wherever you sit on Engagement: always discerning.

Is 1C a bad place to be?

Not bad, but flagged. Low engagement by conviction can serve as ballast. The risk is stagnant growth, and the tradeoff is opportunity cost as tools become more capable.

Isn't Critical good enough?

Checking catches errors of fact. It doesn't ask whether AI belonged in the task in the first place. That prior question is the difference between B and C.

Does more use mean more risk?

Not by itself. Risk comes from use outrunning discernment. 3A is the primary institutional risk because many people sit there and their errors reach students and families.

Can someone be Prudent without using AI much?

Yes: 1C and 2C. Judgment can come first, and engagement can follow with time and permission.

Is a placement a performance rating?

No. It's a starting position for reflection, not an evaluation. Inside a school, formal placement should pair the interview with human review.

Formation

Behind the grid is an old idea. The classical tradition named practical wisdom (phronesis, or prudence) as the virtue that judges rightly what to do in a particular case, before acting. The Prudent column is that virtue applied to a new tool: deciding in advance whether AI belongs in a task, how it should be used, and what must stay human.

Tools shape the people who use them. Schools exist to form people, so the question is never only whether AI makes the work faster. It is also what the work, done this way, makes of us.

Your position, in depth

Placement opens a full profile for your position: what it gets right, its blind spots, a growth path, practices, and an experiment to try.

Find your position

Always Discerning.