Useful to teachers, and restrained about students.

Education is the sector where we most often recommend doing less. The data is about children and young adults, the inferences are consequential, and the temptation to over-model is considerable.

01 / What is hard here

The problem underneath the problem.

A dropout-risk model is technically straightforward and ethically loaded. Label a student at risk and you may trigger support that helps — or an expectation that follows them and becomes self-fulfilling. The model does not distinguish between these outcomes; the system design around it has to.

Historical education data also encodes historical inequity. A model trained on who succeeded in the past will faithfully reproduce the advantages that produced that success, and will present the result as objective.

02 / What we build

Work we take on in education.

Administrative automation: Timetabling, admissions processing, fee reconciliation and document handling — the least controversial and often the highest-return work in the sector.
Assessment support: Assistance with marking structured and short-form responses, with human review of every consequential grade rather than automated finalisation.
Early-support identification: Flagging students who may benefit from support, designed with educators, surfaced to staff who can act, and deliberately not exposed as a persistent label attached to the student.
Content and resource discovery: Search and retrieval across institutional teaching material so staff can find what already exists instead of rebuilding it.
Operational forecasting: Enrolment, capacity and resource planning at department and campus level.

03 / Constraints

What shapes the build.

These are design constraints agreed at the start, not compliance work bolted on before launch. Retrofitting them is expensive and usually incomplete.

Minors' data: Stricter handling, minimal collection, short retention and clear consent — with parents and institutions, not just users.
Bias review: Any model touching student outcomes is examined for differential performance across background before deployment, and monitored after.
Educator authority: Systems inform teachers; they do not grade, stream or exclude autonomously. This is a design constraint we hold to.
Transparency: Students and parents should be able to find out that a system is in use and what it does. We build for that being answerable.

04 / What we advise against

Where we would say no.

We would advise against automated final grading, against persistent risk labels visible to students, and against any behavioural monitoring product in a learning environment. We do not build those.

We would rather lose the work than build something that creates a liability for you or a harm for the people on the other side of it. If that rules out what you had in mind, it is better established now than at delivery.

05 / Other sectors

Adjacent work.

Different domains, the same delivery discipline — a measured baseline, a real evaluation set, and monitoring that catches decay early.

Next step

Working in education?

Tell us which process is slow, expensive or error-prone. The diagnostic will tell you whether this is a problem worth solving with machine learning — including if it is not.

Request a diagnostic All industries