Education Department moves the edtech test from activation to evidence at renewal

The complement of the Tennessee trial's 17% is 83%: in more than four of every five moments when a student had gotten something wrong, the tutor stayed closed ^[5]. A vendor's dashboard reports the trial rate. A renewal decided on demonstrated learning outcomes is decided on the other number ^[3].

There are AI tutors available now, some of them free, patient and competent at eleven at night ^[10]. Moving the 83% costs something else. In Fortune's account of the post-pandemic career-exploration wave, schools had to carve out time during the school day and pay career coaches to get students through the tools, and identifying a promising career meant little unless somebody had brought employers to the table to create an internship, apprenticeship or an actual job at the other end ^[9].

Responsible design, the Department wrote, is "the floor" ^[6]. Fortune reads the document as the beginning of edtech's accountability era and says Washington last week started asking schools for proof ^[12][11].

The renewal language is where this touches revenue. A purchase can be won with a pilot and a demo; a renewal judged on outcomes has to be won with data somebody outside the company gathered, and the guidance asks vendors to publish exactly that, along with a disclosure of what their products cannot do ^[3].

MIT's 95% covers corporate generative-AI pilots ^[1]. One pilot in twenty produced a measurable return ^[7]. Fortune's case is that American schools ran the same play at national scale, buying interactive whiteboards, laptop carts and pandemic-funded platforms and activating each without the training and instructional model built around it ^[8]. That case is an argument from pattern, and the piece does not report district spending totals or student outcome data for AI in US schools ^[14].

In my view the guidance moves a vendor's cost of sale before it moves a student's score. An independent evaluation is a fixed cost paid once per product; the training and the protected class period are recurring costs paid per district per year, and Fortune calls the constraint "everything nobody builds around it" ^[15]. Two things would show me wrong. Districts renewing on usage reports straight through the next budget cycle, or evaluations turning out cheap enough that publishing one stops separating vendors at all.