Everyone's first priority, almost nobody's first requirement
Four K-12 edtech reports, read as one document. They describe a chain of custody for a single question — is this tool safe, and does it work? — and the chain loses something at every handoff.
Over the last month I’ve written about four reports: SETDA and CoSN on what states advise and districts require, then Project Unicorn and Clever on what districts can measure and what teachers actually experience. Individually each is a sector update. Stacked in order of altitude, they stop being four surveys and start being one document: a chain of custody for a single question, and a record of what gets dropped at each handoff.
The question is is this tool safe, and does it work? Four organizations asked four populations some version of it. Here is what happens to it on the way down.
The same five indicators, measured at two altitudes
There is one place where two of these reports ask a genuinely comparable question. The EdTech Quality Collaborative — 1EdTech, CAST, CoSN, Digital Promise, InnovateEDU, ISTE and SETDA — published five quality indicators for edtech: Safe, Usable, Interoperable, Inclusive, Evidence-based. SETDA then asked state leaders which indicators their AI procurement guidance covers. CoSN asked district leaders which they require vendors to provide information about.
Same five indicators. Two altitudes.
Every indicator loses ground between the statehouse and the purchase order
Share of state AI procurement guidance covering each EdTech Quality Indicator, against the share of districts requiring vendors to provide information on it. Ordered by the size of the drop.
Table view
| Indicator | State guidance | District requirement | Drop |
|---|---|---|---|
| Safe | 95% | 55% | 40 |
| Usable | 68% | 45% | 23 |
| Inclusive | 50% | 29% | 21 |
| Interoperable | 59% | 46% | 13 |
| Evidence-based | 45% | 41% | 4 |
Everything drops, which is unsurprising. The interesting part is which thing drops furthest.
Safe is the indicator states lead with — 95%, near-unanimous, the closest thing this sector has to a settled position. It is also the indicator that loses the most on the way down: a 40-point fall to 55% of districts. Evidence-based is the indicator states care least about, at 45%, and it barely moves — a 4-point fall to 41%.
So the correlation between how loudly a priority is stated at the top and how reliably it is required at the bottom is, if anything, negative. The louder the guidance, the bigger the gap between it and the paperwork.
The framework is not the problem. Knowing it exists is.
The obvious explanation would be that districts disagree with the framework. That is not what the data says. 69% of districts require vendors to provide information on at least one of the five indicators — most districts are already doing the thing the framework describes.
They are just not doing it from the framework. 54% of district technology leaders report being unfamiliar with the Five Quality Indicators. A majority of the people implementing a standard have not read it.
That is a much more tractable problem than disagreement, and it explains the shape of the drop. An improvised checklist will independently reinvent “is it safe” — that one is obvious. It will not independently reinvent “is it accessible to learners with disabilities,” which is why Inclusive sits at 29%, the lowest of the five, in a sector where accessibility is a legal obligation rather than a preference.
Why this isn’t negligence
It would be easy, and wrong, to read all this as districts not caring. The capacity numbers make the arithmetic plain.
The mandate arrived; the capacity didn't
Four figures from two of the reports, describing what districts have to work with.
Read those together and the 58% is the one that reframes the rest. Districts are not broadly understaffed — two-thirds have enough people to keep the network running. They are specifically understaffed for the work of making the technology function in classrooms, which is precisely the work that a quality-indicator review is. The staffing exists for the infrastructure and not for the instruction.
What fills the vacuum
When a district cannot run the review the framework describes, the decision does not get deferred. It gets made on whatever evidence is already lying around.
How districts decide whether a licensed tool is worth keeping
Share of school decision-makers citing each factor. Respondents could select more than one, so the bars do not sum to 100%.
Table view
| Factor | Share citing it |
|---|---|
| Budget impact and cost per user | 60% |
| Student performance outcomes | 45% |
| Usage analytics and login data | 44% |
| Direct feedback from users | 35% |
| Teacher adoption and training completion | 20% |
| Input from district leadership | 18% |
| Not sure | 16% |
Cost per user is the only input on that list that requires no infrastructure to obtain. You divide the invoice by the headcount. Every other row needs something the reports say most districts do not have — and note that “not sure” at 16% outranks nothing except itself, which is its own kind of finding.
The chain, in four handoffs
Put end to end, the four reports describe one process with a loss at every step.
- States publish. 95% of state AI procurement guidance leads with Safe — but only around 40% of states issue procurement guidance at all.
- Districts don’t receive the framework. 54% of district technology leaders have not encountered the Five Quality Indicators; they improvise a review instead, and 55% land on requiring safety information.
- Procurement can’t document it. Project Unicorn finds 61% of districts say they “always” or “usually” weigh privacy and interoperability, but only 33% put privacy language and 28% put interoperability language in an RFP. An intention that isn’t in the contract isn’t a requirement.
- Teachers aren’t asked. 21% of educators report having any say in tool selection or privacy safeguards, and 20% of keep-or-kill decisions weigh teacher adoption — despite Clever’s own segmentation showing engaged teachers reaching 60–70% of disengaged students against 15% for reluctant ones.
Not one of those four steps is unreasonable on its own. A state that publishes guidance without a mandate is respecting local control. A district that improvises a review with no staff for it is doing the best available thing. A procurement office that considers privacy without writing it into the RFP is still considering privacy. A district that doesn’t poll teachers on every license is managing its time.
The chain still doesn’t hold. Nobody in it is failing, and the question still arrives at the classroom unanswered.
What we’re doing about it
The reason I keep coming back to these reports is that they describe our actual go-to-market conditions, and they are not flattering to the standard vendor playbook. If 54% of your buyers have never seen the framework, you can pass a lot of reviews without being good.
So the two commitments for Lerad are the ones this chain argues for. Answer the five indicators before anyone asks — in writing, in one document, whether or not the district’s process has a field for it. Our NY Education Law §2-d pack exists for the same reason and is with education-privacy counsel now; it was drafted before we had a single user.
And make the evidence cheap enough to beat cost per user. Not a case study we wrote about ourselves — usage and outcome data a district can pull without first standing up a data team, because the funding numbers above say most of them will not be standing one up.
We start piloting in one Brooklyn classroom next month. The bar I actually care about is whether a teacher who didn’t choose Lerad still opens it in October.
A note on the numbers
Writing this meant reading all four reports against each other rather than trusting my own earlier summaries, and three things turned up that are worth stating plainly.
CoSN’s report disagrees with itself by a point. Its executive summary says 56% of districts require vendors to provide safety information; the procurement section says 55%. I’ve used 55% throughout, because that’s the figure in the section that carries the full five-indicator breakdown and it’s internally consistent with the other four.
Two of these surveys measure the same practice 15 points apart. CoSN reports that 43% of districts include interoperability requirements in their RFPs. Project Unicorn reports 28%. Different populations, different question wording, both credible. I’ve cited each to its own report and I’d treat the gap between them as the honest error bar on this whole genre.
Two figures from my July post didn’t survive. I cited a fall in ESSER-sustained edtech from 27% to 6% to CoSN’s 2026 report. It isn’t in that report — there is no mention of ESSER at all — and it isn’t in the 2025 edition either, so I can’t source it and have removed it. In the same paragraph I reported 65% of districts as understaffed for cybersecurity; 65% is the share naming a lack of dedicated budget as their top cybersecurity barrier, which is a different claim. Both are now corrected in that post, with a note saying so. Its argument didn’t depend on either number, which is lucky rather than to my credit.
Which is, more or less, the argument of the last three posts pointed at myself. A claim without a document behind it is an intention, and this is what it costs to check.
Sources: SETDA, 2025 State EdTech Trends Report; CoSN, U.S. State of EdTech 2026; Project Unicorn / InnovateEDU, State of the Sector 2025; Clever, Classroom of the Future Report 2025.