Assesses academic work before submission. Running in four countries since May 2026. Writes nothing for the user and gives no grade.
- Unattended
- From request to delivered report without intervention.
- Calibrated
- Per discipline, per level of study and per country.
- In house
- Engine, calibration, payment flow, infrastructure.
How it is built - 01
Judgement built in stages
More
The assessment is built up in several separate steps, each with its own instruction and its own failure path. If one step drops out, the assessment does not.
- 02
Verification against invention
More
Every remark must be literally traceable in the submitted document. Whatever fails that test is removed, even when it reads well.
- 03
Cost control by design
More
Cost per run is a design choice, not an outcome. It is measured, bounded and monitored, so the system stays predictable at peak volume.
- 04
Calibration
More
Discipline, level of study, type of institution and country are settings on one engine, not separate versions. The separations between them are under tests, so one domain cannot leak into another.
Esmond does not in our assessment qualify as a high-risk AI system under Annex III of Regulation (EU) 2024/1689 (AI Act).