Official data show a hiring gap AI tools are not built to close

The classroom note

A principal who cannot fill a vacant maths post by September is increasingly offered an AI tool as partial relief, sold on the promise of a workload reduced enough that fewer staff can cover the same students. Official US data describe the scale of the underlying gap that pitch responds to. A National Center for Education Statistics School Pulse Panel release, surveying 1,392 public K-12 schools nationwide between 13 and 27 August 2024, found 74% of public schools reported difficulty filling one or more vacant teaching positions, down from 79% the year before, with 'an overall lack of qualified candidates (64 percent)' and 'too few candidates applying (62 percent)' the two most commonly cited reasons.

What the evidence says

A second, later wave of the same NCES survey, collected from 1,491 schools between 8 and 22 October 2024, quantifies the gap more precisely: 3% of all public school teaching positions were vacant, 35% of schools reported one or more teaching vacancies, and 50% of school leaders felt their school was understaffed, an increase from earlier measurements according to the same release. That release also notes the net change in teaching positions rounded to zero, since 31% of schools reported losses and 27% reported additions. These are large, nationally representative samples with a stated survey period, though the figures describe vacancies and hiring difficulty, not attrition causes or teacher workload specifically.

The implementation question

The mechanism to name plainly is that an AI tool addressing marking or lesson-planning time does not put a qualified candidate in an empty maths classroom; the NCES data describe a recruitment and applicant-pool problem, with 'lack of qualified candidates' as the leading stated cause, not primarily a problem of existing staff overloaded with marking. A workload tool might make a teacher's day more sustainable, which could plausibly affect retention over time, but the cited data do not measure that link, and no source here reports a study connecting AI adoption to reduced vacancy rates.

What holds and what fails

What holds is that official, sampled, dated workforce statistics are a far sturdier basis for discussing staffing than a vendor's framing of 'the teacher shortage' as a single problem an AI product solves. What fails is the implied logic, common in sales material rather than in these official releases themselves, that a workload tool is a staffing policy: the NCES data locate the bottleneck substantially in candidate supply, which a classroom AI tool does not increase. This is an editorial distinction the cited releases do not make explicitly but that their own reported causes support.

• Is our vacancy problem primarily a candidate-supply issue, a retention issue, or both, and which does an AI tool actually address?
• Has our district measured whether existing staff workload, not candidate supply, is the constraint an AI tool should target?
• Would freeing marking time change how many qualified candidates apply for open posts?

Vacancy statistics and workload tools describe two different mechanisms, and a staffing plan that conflates them risks solving the wrong one.