The teacher is the decider, by design.

"The AI reasons. The teacher decides." That is the SLE motto. It is on the home page. It is in the dark architecture band on every edition page. It is the sentence I want every educator who looks at the platform to leave with. And it is an engineering commitment, not a marketing promise.

On the lesson surface: the teacher approves the curriculum before students see it. AI-generated course schedules are surfaced to the teacher first. The teacher reviews. Edits. Signs off. Then the students get the lesson.

On the assessment surface: the teacher attests every grade. AI suggests; the teacher attests. The platform records the suggestion and the attestation as distinct events. The academic record never carries a grade that a teacher did not attest. Auto-grade-to-record is not behind a toggle; it is not a setting we shipped with off-by-default. It does not exist.

On the disclosure surface: students see, in-line on every AI-touched surface, that AI was involved. Districts subject to California AB-1791 (and the growing list of state-level AI transparency statutes) get a configurable disclosure posture that satisfies the statute without requiring a one-time consent that students forget.

Three reasons. First, because the moment of grading is the moment of attribution. A student looks at a grade and asks who decided. The honest answer has to be: a person did, with AI as decision support. Anything else fails the attribution test. Second, because the moment of grading is the moment of relationship. A teacher who has to attest the grade has to look at the work, and looking at the work is a relational act. Third, because the moment of grading is the moment of liability. The educator is the licensed professional in the room. The platform is not.

We are not the only company that says some version of this. We are one of very few that engineers it. The auto-grade feature is the cheapest thing on the shelf in this market. It exists in most EdTech platforms because it sells. We chose not to ship it because we believe what we say about the educator being the decider, and shipping auto-grade would make us liars.

ArthurAI does the work that supports the decision rather than making the decision. The platform reviews student responses, computes accuracy, identifies patterns in the misses, drafts feedback the teacher can edit, surfaces the students whose performance trajectory warrants intervention, and prepares the session-by-session evidence the teacher would otherwise have to assemble by hand.

That is genuine work. It compresses the cognitive load on the teacher from the assemble-evidence-then-decide pattern that consumes evenings down to the decide-with-evidence-already-assembled pattern that respects the teacher's professional time. The teacher still decides. The teacher decides faster, with more confidence, on more students, with better evidence.

I expect over time to be asked to soften this posture — by customers wanting throughput, by competitors who auto-grade and look faster, by analysts who think the productivity narrative requires it. The answer will be no. The educator-attested workflow is not a feature we can compromise without compromising what the platform is. The motto holds because the workflow holds.

How the assessment surface actually works →