Sokros

For ACCA learning providers — configuration-ready

Numbers people expect the marking to add up.

Applied accountancy evidence — workbooks, objective statements, portfolio submissions — marked the way finance expects everything: reconciled, evidenced, and auditable to the line.

Objective evidence — PO4 · statement review
Objective addressed in the learner's own role
Sufficiency — worked example presentNot met

The statement asserts the outcome but shows no worked instance — returned with the gap recorded

Sources cited are real and dated

The same ledger logic, applied to marking.

Extract the evidence, check the rules, judge with reasoning, keep the record — a pattern any auditor would recognise as double-entry for assessment.

Every claim, evidenced

Statements and workbooks read into typed fields with the learner's verbatim words attached.

Sufficiency as a check

“Show a worked instance” is a rule, not a vibe — misses are flagged with the exact gap.

Integrity signals

Reference credibility, AI-usage analysis and collusion indicators, always with evidence, always for a human.

An audit trail by default

Every verdict and reason stored per learner and attempt — reviewable years later, unchanged.

Integrity a finance body would sign off.

Accountancy programmes attract the highest integrity scrutiny — so the scanning here goes further than an originality percentage.

Integrity scan — PO4_JordanA.docx · attempt 1
External similarity12% — in range
Within-cohort similarity41% — flagged

Task 2 closely matches PO4_SmithK.docx — matched passages attached side by side

AI-usage analysis2 sections flagged

flagged passages quoted with the signal that raised each one

Reference credibility7 of 9 verified · 2 unverifiable

Signals for the assessor — evidence attached, never an automatic penalty.

Super-charged integrity

Collusion shows up even when the internet can't see it.

Similarity against external sources is the easy half. Sokros also compares learners within a cohort — catching paraphrased sharing no external database contains — analyses AI usage with the flagged passages quoted, and classifies every cited source as real, dated and plausible. Each signal carries its evidence and lands with an assessor, never as an automatic penalty.

See a scan
LMS connection — Moodle · Cohort 2026-A
Pulled 12 new submissions06:00 sweep · unit, level and cohort recognised in place
Marked 11 · reverted 1 to draftgrades, feedback and referral reasons written
LMS procedure completeddocument attached · criteria grade buttons selected · general feedback written on the page
Workflow states setpass → review · refer → release-to-learner, per centre rules
Awaiting batch approval11 in the dispatch queue — approve or amend, then releaseQueued
Release 11 to LMSReview firstno files moved by hand

Intake to release

The operational loop runs here too.

Instant Revert to Draft keeps ineligible portfolio submissions out of the marking queue; live LMS connections pull, mark and complete your platform procedure — grade selections, page feedback, workflow states, release; and the dispatch hub batches review and approval so a cohort clears in one sitting.

All capabilities

Deterministic

same submission in, same outcome and reasoning out — replay-proven on every release

In-region

UK, EU and Gulf hosting or fully on-premise; learner data never reaches consumer AI services

Human-final

sampling, review and approval modes keep qualified assessors signing off

Reconcile a cohort with us.

Bring your evidence formats and marking guidance — we'll map them to the pipeline and calibrate on your own assessed examples.