Human over the loop, AI for assessment.
AI implementation
lives or dies on
the loop.
People need control over the loop, and they need to sit at the right points in it, where their judgement counts most. Get that right and the tech starts earning its keep.
That is what Human over the loop means in practice.
AI does the heavy lifting.
Experts stay in control.
How it works
One pipeline, not four products
Most quality work runs on gut feel. Ours runs on a loop that turns vibes into metrics. The same four steps work whether you are checking questions, standards, a teacher's capability, or a learner's evidence. Same engine, different lens.
-
Gates: what gets in
Cheap, mechanical checks decide what is even worth an expert's time. Is the input clearly correct, does someone own it, does it clear the minimum bar, is it machine-readable. We built this as an Auto Reject Gate for a QA team drowning in 500-plus standards. Expert attention only gets spent where judgement is actually needed.
-
Gold Loop: what gets held up as the standard
Your experts pick the near-perfect examples and say why they are good. That "why" becomes the reference the rest is measured against. Find great work, name what makes it great, pull out those elements, feed them back in.
Slow the first time, compounding after that.
-
Capability: who is allowed to judge
Before anyone judges, we check their evidence holds up: quals, teaching history, industry currency, matched against the units they deliver. Every match the AI proposes stays proposed until a human confirms it.
The machine never signs anyone off on its own.
-
RPL: the same engine pointed at people
Recognition of Prior Learning asks whether a person's existing evidence already meets a unit's requirements, and what is left to cover.
One rule keeps it honest: a transfer claim that names nothing that does not transfer is a red flag, not a strong claim.
Recognition without rubber-stamping.
The point is the pattern. Gate, then exemplar, then a qualified judge, then the criteria you extract.
It runs the same on questions, standards, teacher capability and learner evidence. One framework, many lenses.
Bring us the thing you can only judge by feel, and we will turn it into something you can measure.
Same engine, different lens.
- Questions
- Standards
- Teacher capability
- Learner evidence