Live on your units, gated at every step.

One call establishes the unit, level and brief year, your grade scheme, your RTD rules and where you want humans in the loop. Fixed-price pilot, scoped upfront.

Criteria, case study, word-count rules, minimum requirements, rubrics, extraction schemas and your official feedback sheet are authored as versioned configuration: no engine code, no engineering ticket.

The unit runs against your human-marked calibration set and the outputs are compared side by side. Go-live is gated: zero tolerance for a script flipping between Refer and a clear Pass.

Start with every script held for your approval in the dispatch hub. Loosen toward sampling or full automation at your own pace. The modes are settings, not promises.

The plugin is pointed at your platform, the pull cadence set to your plan, and the release procedure mapped to your workflow states. From here, submissions arrive and grades return without file handling.

New brief years arrive as new configuration versions, resubmissions follow awarding-body caps, every behavioural change is a numbered amendment, and the audit record grows from day one.