Piarrel
ENESPLBook a demo (opens in a new tab)

AI marking, with a teacher approving every mark

Piarrel proposes; the teacher decides. Each mark is proposed point by point against your own mark scheme, and nothing reaches a pupil until the teacher releases it.

What Piarrel marks, and what it leaves to you

Piarrel proposes marks for typed answers, calculations, numeric answers, multiple choice, equations and tables. Graphs, drawings and labelled diagrams are always left for the teacher to mark.

What the teacher sees

  • Each answer beside the mark-scheme point and Piarrel's reasoning.
  • A confidence level on every proposed mark, so the answers worth a second look stand out.
  • Your conventions: accept, reject and ignore lists on every marking point.
  • One press: full-marks answers proposed with high confidence approve together.
  • Every decision recorded: change any mark, and the change is kept in an audit trail.
The marking queue: a pupil's answer beside the mark-scheme point, with Piarrel's proposed mark, its reasoning and its confidence.
The marking queue: each answer beside its mark-scheme point, with the proposed mark, the reasoning and the confidence.
The button that approves, in one press, the full-mark answers proposed with high confidence.
One press approves the full-mark answers proposed with high confidence. Partial and zero marks stay in the queue.

How it held up with real classes

In a September–October 2026 trial at a British international school in Spain, 3,650 of Piarrel's marks were reviewed. Teachers kept Piarrel's mark on 94.4% of reviewed answers, measured against each teacher's own final mark. Of the 203 marks teachers changed, 195 were raised and 8 lowered. Teacher time per class set was 26 minutes in Piarrel, against about 1 hour 45 minutes on paper, both timed at the school.

Feedback pupils can use

When the teacher releases the results, each pupil gets feedback written from their own answers: what went well, and what would make it even better. The class gets a summary too.

A pupil's released feedback: overall feedback, an area to focus on, then what went well and what would make it even better, question by question.
A pupil's feedback, once the teacher releases it: overall, then question by question.

Questions

Does the AI decide pupils' marks?

No. Piarrel proposes a mark with its reasoning; a teacher approves or changes it, and only then does a pupil see it.

Who marks graphs and drawings?

The teacher, always.

Is pupils' work used to train AI?

No. Answers are marked through OpenAI's API without pupils' names, and are not used to train AI models.

Can we see why a mark was given?

Yes: every proposed mark comes with its reasoning, against the mark-scheme point it refers to.

Book a 20-minute demo (opens in a new tab)Watch the film (2 min, with sound)