Three things: (1) links to your public programme and module pages, (2) a grades CSV per cohort and year, (3) an indirect-measures CSV (alumni/exit surveys, employer feedback). The wizard provides downloadable templates for both CSVs, and the upload screens validate the files with plain-language error messages. No system integration is required for a pilot.
From data handover: about two weeks for a pilot, including faculty training and the approval rounds. The pipeline itself runs in minutes — the calendar time is for the human steps: your faculty reviewing and approving competency goals and findings.
The AI drafts; your faculty decides. Competency goals — the intellectual anchor of the report — are blocked in an approval queue until a named person signs them off. Every paragraph in the final PDF is backed by an evidence row in the bundle, and the SHA-256 manifest proves nothing was altered after signing. You can also reject or amend anything at the approval gates.
No. Language processing runs on Anthropic's commercial API, which under its commercial terms does not use customer data for model training. Your data stays in your isolated tenant, and survey rows with fewer than 5 respondents are dropped by design to protect individuals.
Today: AACSB Standard 5 (Assurance of Learning), the most labour-intensive part of the cycle. The architecture is standard-agnostic — the agents read the standard text they are given — so EFMD/EQUIS and QAA variants are on the roadmap. Ask us about your specific framework.
The pilot is a fixed £2,500 — AI usage included. For context, the AI cost of one full pipeline run is around £0.10, and the platform enforces per-run and per-day budget caps, so costs cannot run away. Annual plans start from £400/month after the pilot.
Yes — the platform deploys to an EU region (Frankfurt) by default. For pilots we can also discuss running on infrastructure you control. TLS in transit, secrets in the host's secret store, container runs as a non-root user.
A submission bundle of three files: REPORT.PDF (the typeset narrative report), EVIDENCE.JSONL (every data point behind every claim, machine-readable), and MANIFEST.JSON (SHA-256 hashes chaining the two, so any reviewer can verify integrity). Plus access to the platform to re-run or explore.
No — if you can export a CSV, you can run Atomic AoL. The grades template is deliberately minimal (module, competency goal, year, cohort, pass-rate or mean score). Direct LMS integrations are on the roadmap, but no pilot has ever been blocked by an export.
You keep the bundle and everything you approved. If the quality convinces you, the annual plan (from £400/month) covers unlimited cycles on up to three programmes. If not, you walk away with a complete AoL cycle delivered and no further obligation.
Write to us — a real person answers, usually within one working day.
Ask away → Book a call →