The Two Tools That Cleared The Bar

On July 20, Dr. Sanée Bell walked the Katy ISD school board through a new artificial intelligence framework. The Houston-area district, nearly one hundred thousand students, will start the school year on August 12 with generative AI chatbots blocked on the devices of every child in kindergarten through sixth grade.¹ Microsoft Copilot is out. Google Gemini is out. Any open chatbot the child might type a paragraph into is out.²

Two tools cleared the bar and stayed in the room. The reason they cleared it is the design conversation the country has not been having.

The Distinction Under The Ban

The two tools that stayed are Amira Learning and Writable.² Amira is a reading tutor. It listens, in real time, while a first grader reads a passage out loud, catches the syllable she stumbled on, and offers a small, targeted intervention on that specific error.³ Writable is a drafting platform. When a fifth grader turns in a paragraph, Writable can guide the student to clarify an argument or strengthen the evidence, but it will not write the paragraph for her.⁴

Neither of these tools produces the finished work. Neither answers the question. Amira does not read the book to the child. Writable does not write the sentence for the student. What they do is coach a task the child is already inside of. The child is still the one doing the reading. The child is still the one writing the paragraph. What the AI does, in both cases, is watch and guide.

The chatbot the district blocked has a different design. Type a prompt, receive a paragraph. The transaction the tool is built for is production, not coaching. That is not a moral fact about the software. It is a design fact. And Katy ISD, in language most policy debates never quite reach, drew the line between the two.

The Third Position

Most districts making a call this summer are picking one of two positions. Position one, allow the chatbots and hope the students learn to use them responsibly. Position two, block the chatbots and hope the students do not find the workaround on their phones. Katy did neither. It blocked the open, general-purpose tools, and it named, by product, the narrower tools it would keep. The names matter. When a policy document says "district-approved AI," most parents cannot tell whether that means Amira or ChatGPT. When it says Amira Learning and Writable, they can.

The framework moves in a ladder from there. Seventh graders get access to district-approved tools under teacher supervision. Eighth through twelfth graders may use generative AI for specific purposes with teacher authorization, and they must cite the tool when they use it.¹ The distinction being scaffolded, grade by grade, is not just how much AI. It is which kind of AI, doing what kind of job, in the presence of which kind of adult.

What A Tool Is Asked To Do

The reason a first grader with Amira is in a different pedagogical situation than a first grader with a chatbot is not the model size, or the vendor, or the terms of service. It is what the tool is asked to do. Amira is asked to listen. The chatbot is asked to produce. A child who is being listened to while she reads is doing the reading. A child who is asked what she wants the paragraph to say, and then hands back the paragraph the machine wrote, is not, in the deepest sense, doing the writing.

Neither tool is neutral. Neither is the enemy. What Katy's framework does, without quite saying it, is treat the design of the tool as the object of the policy, and not just its brand name or its underlying model. That is a more careful move than the yes-or-no fight the national debate keeps having.

The Layer That Survives

At Koan we think about the layer of a student's work that survives, or does not, contact with the tool. When the tool watches and coaches, the paragraph the student writes is still legibly hers. The revisions, the pauses, the moment she reached for a different word, still belong to her. When the tool produces on request, the paragraph belongs, at best, to the two of them, and often to the tool alone. There is no version history worth reading in a paragraph the student did not write.

Katy's decision is not, in the end, about AI. It is about which layer of a child's thinking a school is trying to protect. The two tools that cleared the bar cleared it because they leave that layer intact. The one that did not was asked to leave.

If your district let two AI tools into its elementary classrooms this year, could you say, in one sentence each, what those tools do not do?