Articles · AI in education · Teaching

What the AILE framework changes in your Monday lesson

AILE keeps the lesson structure good teachers already use and sharpens what they can see. Here is what changes in a Monday lesson, phase by phase.

By Dr. William Hickman · 6 min read

Most conversations about AI in schools start with a product. I would rather start with a Monday: twenty-eight students, one teacher, a unit that has to land by Friday, and several students already behind before the bell rings. If a framework cannot tell you what changes in that room, it is a slide deck.

The AILE Framework, the Artificial Intelligence-Instructional Learning Ecosystem, is my answer to that test. It upgrades the ITI Instructional Model I built, and the core idea is simple. The architecture of a good lesson stays the same. What multiplies is what the teacher can know, see, and reach.

Here is where the change actually lives.

What does not change

Start with what you keep, because many of you are already running it.

The four phases stay: I Do, We Do, They Do, You Do. That sequence rests on Vygotsky’s zone of proximal development and on the gradual release of responsibility described by Pearson and Gallagher and refined by Fisher and Frey. The direction of travel is always toward independence.

The four High Impact Centers stay too: Co-Pilot, Chauffeur, Carpool, and Caboose. The names tell you who holds the wheel. At Co-Pilot, the teacher works directly with three to six students. At Chauffeur, students practice on a route the teacher designed. At Carpool, peers share the ride, with defined roles and a product they owe the group. At Caboose, the student drives and the teacher follows behind.

Underneath everything sit the 5 C’s and the 4 R’s, the competencies and character the whole design exists to grow.

AI replaces none of this. It gives each piece a sharper set of eyes.

Before the bell: Entry Points that keep updating

In the original model, Entry Points were a quick check before a unit that sorted students into three groups: below the entry threshold, at it, or above it. Useful, and frozen the moment you gave it.

In an AILE classroom, the evidence keeps arriving through the work students are already doing: a paragraph written at Chauffeur on Thursday, a contribution logged at Carpool, a mastery snapshot from Caboose. By Monday morning, your groups reflect Friday, not three weeks ago.

You still decide. The tool proposes. You look at the student work behind the proposal and make the call.

I Do: you still model, but you stop guessing

I Do looks the same from the front of the room. You think out loud, name your own confusion, and show the moves an expert makes. What is different is what you know while you do it.

Short checks embedded in the lesson send a signal to your screen: who answered correctly, who missed and in what pattern, who answered so fast it may have been a guess, who has been idle for a while. I call the habit the Ninety-Second Check, a loop short enough to close inside the lesson it belongs to. You glance, see the misconception cluster, and reteach before you move on.

This matters because feedback is one of the strongest levers we have. In Hattie’s 2017 synthesis of 252 influences on achievement, feedback carries an effect size of 0.70, well above the 0.40 hinge point he uses to mark an influence worth a teacher’s time.

We Do: feedback reaches the back row

Here is the honest math of guided practice. In a thirty-minute We Do, one teacher can give specific feedback to perhaps six or eight students. Everyone else gets whatever general comment reaches the whole group.

This is where AI pays its first real dividend. A well-built tool responds to each student’s work with the question that pushes thinking forward, or flags an error with a prompt instead of a correction. You still lead. The tool reaches the back row.

The centers: one of you, many of them

During rotations, AI plays a different part at each center. It informs at Co-Pilot, so you know who belongs at your table today and why. It scaffolds and gives feedback at Chauffeur. It facilitates and flags at Carpool, surfacing a group that has stalled before you have to walk over and discover it. It personalizes and assesses at Caboose.

Expect some flags to be wrong. Picture a quiet student flagged for low contribution after four minutes. You know she listens a long time before she speaks, and that the group usually follows when she does. You dismiss the flag. Now picture another student at Caboose who gets every problem right while asking for a hint on every third one. You might never have caught that pattern from across the room. The screen did.

That is the partnership. The tool sees patterns. You know children.

In inclusive classrooms, a special education co-teacher can run Chauffeur while you run Co-Pilot, doubling the number of students who get intensive instruction in one period.

You Do: the rule you cannot skip

You Do is where AI can do the most good and the most damage, so it carries one rule: the tool may make the task fit the student, but it may never do the student’s thinking.

The research behind that rule is sharp. In a study of about 1,000 high school math students in Turkey, Bastani and colleagues (2025) found that unrestricted chatbot access raised practice performance by 48 percent, and then those students scored about 17 percent lower on the exam than peers who never had AI. A tutor version with guardrails raised practice performance by 127 percent and did no harm. Same technology. The design decided the direction.

So redesign at least one independent task so a chatbot cannot complete it on a student’s behalf. Ask for a three-minute live defense of the student’s own claim. Ask for the draft trail, with a note on what changed and why. Ask for a short teach-back recording, or an error autopsy on a mistake from her own earlier work. Each one requires presence, specificity, or a visible process.

After the bell: the intelligence layer works for you

The fourth layer of AILE is the intelligence layer. Learning Agents work with students in the centers. Teacher Agents work for you in the background, pulling evidence from all four centers into next-step recommendations and drafting differentiated materials. Assessment Agents turn student work into mastery evidence. Intervention Agents watch for early signs of drift.

A good dashboard answers three questions. What is happening right now? What does it mean? What should I do next? Everything else is noise.

The line is clear. Teacher Agents do not make instructional decisions. They give you better information to make them. The expertise and the authority stay with you.

The fifth layer, Extended Learning Pathways, carries practice home and feeds what happens there back into tomorrow’s groups, so the work a student does at the kitchen table actually counts in your room.

What to do this week

None of this requires a purchase order.

  1. Run a trust test on one tool before it touches student work. Write down where student work goes, what happens when the tool is wrong, who might be missing from its training, whether it does the thinking or prompts it, and whether you can explain it to a parent in one sentence.
  2. Redesign one You Do task with two pieces of process evidence, such as a live defense and a draft trail.
  3. Use AI in one We Do or Chauffeur activity with guardrails you wrote, and override it at least once with a one-line written reason.
  4. Co-write a classroom AI agreement with your students, and sign it first.

The full framework, with every tool and figure, is in The Intelligent Classroom. The first step fits on a sticky note: keep the lesson, sharpen the eyes, and never hand over the thinking.

Dr. William Hickman (Dr. Philip William Van Hickman) is an inventor, former superintendent, and the founder of MindAble Solutions. He is the author of eleven books, including The Intelligent Classroom. About him · Official site