Sokros

For CMI centres — configuration-ready

Management assignments, marked at the right depth.

The hardest part of CMI marking isn't reading — it's holding the line betweendescribe and evaluate, script after script. Sokros encodes that line as explicit checks and grades to the descriptors, consistently.

One unit's journey to going live.

From your brief and a handful of marked assignments to a unit that marks itself — gated by calibration at every step.

Grade scheme configured per unit

Calibration measured in dozens of scripts

Every decision stored and replayable

01

Author the unit, not the engine

CMI units, assessment criteria, command-verb expectations and grade descriptors are authored as versioned configuration — calibrated against your assessors' own marked work before anything goes live.

02

Read management evidence properly

Reflective accounts, workplace examples and applied recommendations are read into typed fields — every claim pinned to the learner's verbatim words, first-person or not.

03

Hold the command verb line

“Describe” credited where the brief demands “evaluate” is the classic CMI referral. Depth expectations run as explicit checks per criterion, so under-pitched answers are caught with the exact gap recorded.

04

Grade to the descriptors, file the sheet

Each criterion is judged against the unit's descriptors with written reasoning, and feedback lands in the assessment template your centre already submits — referrals carrying what the resubmission must address.

Requirement check — depth of treatment
Evaluation present, not descriptionNot met

Options are listed but not weighed — evaluative treatment not found for AC 2.3

Applied to the learner's own organisation
Two management models compared

The operational layer, tuned for CMI.

Everything around the marking — intake, LMS, approvals, intelligence — runs for management qualifications exactly as it does for CIPD centres today.

Intake — 512_HughesJ.docx · received 08:47
Signature presentdeclaration page signed
Correct unit & briefUnit 512 · 2026 brief matched
Word count3,912 of 3,900 — within tolerance
Summative in summative areaplacement confirmed
Referencing attemptedno reference list found — required at Level 5
AI & similarity in rangescan complete, within centre thresholds
Reverted to draft — 09:142 min after submission

Reason attached and returned to the learner — no marking cycle spent, the queue stays markable.

Instant Revert to Draft

No manager waits two weeks to hear a signature was missing.

Your centre's eligibility rules — signature, correct unit and brief, word-count tolerance, referencing, integrity range — run the moment an assignment arrives. Ineligible work returns with recorded reasons in minutes to a daily sweep by plan, so the marking queue holds only markable work and workload forecasts stop lying.

See the intake checks
Dispatch hub — 512 · Cohort 2026-A
Flagged · 3Dispatch ready · 24Approved · 61Amends set · 2
512_PatelR.docxGates 6/6 · judged Pass · no flagsPass
512_ChenM.docxGates 6/6 · judged High Pass · no flagsHigh Pass
512_OkaforD.docxGates 6/6 · judged Pass · no flagsPass
512_JordanA.docxGate miss on AC 1.2 — held for reviewRefer
Approve & release 24Amend selectedreleases on the LMS too

Batch approval on the dispatch hub

Marked assignments queue by state with the grading basis on every row — approve a cohort in one sitting and release to the portal and your LMS together, with your platform's workflow states set for you.

Copilot, Insights and integrity — same record

Ask why any learner referred and get the stored reasoning with a drafted reply; see the most-referred criteria and command-verb failure patterns across a cohort; scan for plagiarism, within-cohort collusion and AI misuse with passages quoted.

UK-based · no LLM APIs in the marking pathNot trained on your learnersCapable from the first assignment
All capabilities, shown working →

Built for how management evidence actually arrives.

Messy documents

Templated or not — tables, appendices, reference lists and embedded charts are all read, not rejected.

Reflective writing

First-person workplace accounts are evidenced like any other claim: typed fields, verbatim quotes, provenance kept.

Assessor oversight

Sampling, review-before-release and approval modes put your IQA exactly where you want it.

Calibrate a CMI unit.

Bring the brief, the descriptors and a dozen marked assignments — we'll show you the same unit marking itself, decision by decision.