For CIPD centres · live today
Same-day marking across Levels 3, 5 and 7. Every grade replayable for moderation.
Live in 34 centres · 10,000+ assessments a year · Refer to High Pass
Reverted to Draft · 09:41
Signature page missing. Returned with reasons, nothing joined the queue.
01 · Pain points, fixed
The queue, the ineligible script, the moderation scramble. Each one closed by design, not by asking markers to work faster.
Pain 01 · the marking queue
Cohorts that queued a fortnight with self-employed markers come back the same day: grades resolved, feedback written, ready to dispatch. The SLA is measured in hours at every point in the term, not just the quiet weeks.
Pain 02 · the ineligible script
Signature, unit and brief match, word-count tolerance, referencing attempt at Level 5, AI and similarity range. Every eligibility rule runs the moment work arrives, and ineligible work is Reverted to Draft in minutes with its reasons attached. The queue holds nothing but markable scripts.
Reason attached and returned to the learner — no marking cycle spent, the queue stays markable.
Pain 03 · the moderation scramble
Deterministic grading with zero third-party AI in the marking path. Nothing probabilistic or statistical decides a grade, so nothing drifts between cohorts or discriminates by phrasing. Every grade traces to a verbatim quote, and IQA sampling and the EQA visit read straight from the record.
02 · Marking capabilities
Formative or summative, first attempt or resubmission, essay or recorded presentation: one deterministic pipeline marks it all, against your brief year.
Every evidence mode
Recorded presentations are transcribed and checked criterion by criterion, the same pipeline as an essay. Charts and tables are read as evidence, not skipped.
Integrity first
External and cohort originality checks built in. Flagged passages are quoted, and a flag is always a signal for your assessors, never an automatic penalty.
Second attempts
Attempt-keyed, additions checked against the gap, CIPD caps applied automatically. The first attempt's record stays untouched.
Brief years
Nothing resets, nothing retrains. Both years stay runnable and auditable, side by side.
03 · Proof
Moderation week used to be a weekend of reconstruction. This year the pack came off the record before the meeting started.
Learners hear about a missing signature the same morning, not after two weeks in a queue. The waiting complaints have simply stopped.
“Why was this referred?” used to be a week of finding the marker. Now it's a lookup while the learner is still on the phone.
Live today across CIPD units at Levels 3, 5 and 7.
04 · Reach
Sokros marks CIPD assessments from Ottawa to Jakarta. Different time zones, one deterministic record, the same grade wherever the qualification runs.
London · Dublin · Paris · Berlin · Lisbon · Amsterdam · Bern · Valletta · Ankara · Cairo · Riyadh · Manama · Abu Dhabi · Muscat · Abuja · New Delhi · Kuala Lumpur · Singapore · Jakarta · Ottawa
05 · Workflow
Nothing generative decides a grade. Four moves take a 5CO01 script from submission to outcome.
The determinism, replay proofs and calibration underneath:how the engine works →
01 · Extract · AC 1.2 · two theories required
Every AC is read into typed fields, each claim pinned to a verbatim quote from the learner's own words. No extraction, no credit: a criterion cannot be passed on fluency alone.
theory_1 · Maslow's hierarchy
“Maslow's hierarchy suggests employees are motivated by…
theory_2 · no extraction found
No credit without evidence. Fluent prose cannot fill this field.
A plainly worded answer that meets the criteria extracts in full. An international learner's English is never the thing being graded.
02 · Gate · AC 1.2
The brief's hard rules (“two theories”, “applied to the case organisation”, “stages included”) run as present-or-not checks before any quality judgement is allowed to start.
Binary, present or not, decided by the extraction and recorded with its evidence.
03 · Judge · AC 2.1 · rubric band
Once the core requirements are met, each criterion is banded against the unit's own descriptors, from Refer through Low Pass and Pass to High Pass, with written reasoning. The order a professional marker works in, held every time.
Quality is judged only after the requirements stand, never instead of them.
04 · Resolve · task 1 outcome
Band boundaries, gate caps, resubmission caps and dispatch rules resolve every judgement deterministically. Confidence scores and review flags are mathematical, not predictive, and every decision is stored.
Same script in, same grade out, replay-proven byte for byte on every release.
See these four moves on one of your own units.
Book a demo06 · Integrations
LMS live connections
Sokros connects to the popular LMS platforms and knows where each assessment lives: unit, level, cohort year. It pulls on your schedule, marks, then completes your platform procedure end to end. Document attached, grade buttons selected, page feedback written, workflow state set, saved and released. Fully automated, or held in a dispatch queue and released in batches from the Sokros Portal.
Why no files change handsLive connections to the platforms CIPD centres already run
07 · Platform
Ask the record, not the room. Every grade, referral reason and RTD cause your centre has ever produced, retrievable in seconds.
Sokros Copilot
Less a chatbot than instant retrieval over every stored decision. “Why this grade?” stops being a hunt through SharePoint folders, email threads and a self-employed marker's calendar — admin and IQA answer from the saved reasoning the same day, and moderation windows and EQA audits pull their evidence packs from the same record.
From the marking record: 34% of referrals trace to AC 3.2 — double the 2025 rate. The 2026 brief reworded task 2, and the learner-confusion flag is up on that question.
Referrals by criterion — Cohort 2026-A
vs 2025 cohort · brief-year difficulty shift flagged
Every figure is retrieved from stored decisions — ask anything, drill anywhere.
The current AC 2.4 gate requires two methods (2026 brief · v3). I've raised an amendment request and attached your six assessor overrides as evidence.
Marking logic never changes silently — approved, numbered, rationale written, replay-tested.
38 in cohort: 31 released, 4 dispatch-ready, 2 flagged for review, 1 reverted to draft. Oldest unreleased is 2 days in — comfortably inside SLA.
Status is read live from the queue — and the action sits right next to the answer.
From the stored record: AC 1.2 requires two theories explained — extraction found one.
“only Maslow's hierarchy is explained in task two…
To: A. Jordan — learner email on file
Subject: Your 5CO01 result — what AC 1.2 needs
Hi Aaliyah — here's exactly what happened, and what to do next.
AC 1.2 asks for two motivation theories. Your Maslow explanation is strong; a second theory isn't there yet.
For resubmission: add one more theory in your own words, applied to Meridian BioTech.
— Student Success, Meridian Training Centre
Feeds from student-success inboxes, automatically or on forward.
Most-referred criteria
AC 3.2 referrals doubled under the 2026 brief — question wording flagged
RTD causes this month
word count 41% · missing signature 28% · wrong brief year 17%
Unit health
weakest unit: 5CO02 · pass-first-time down 6 pts vs 2025 cohort
Most-referred ACs, RTD causes, brief-year difficulty shifts and the questions that confuse learners, read straight from the marking record.
Flagged, up-for-review, dispatch-ready and approved queues, with the grading basis on every row. Batch approve and release, to portal and LMS alike.
Sampling reports and the EQA evidence pack (quotes, gates, reasoning, amendments) exported from the record on demand.
Task 2 closely matches 5CO01_SmithK.docx — matched passages attached side by side
flagged passages quoted with the signal that raised each one
Signals for the assessor — evidence attached, never an automatic penalty.
Plagiarism and AI-misuse analysis with within-cohort similarity, flagged passages quoted. Signals for your assessors, never auto-penalties.
08 · What makes us different
No external APIs and no LLM in the marking path. Learner work stays inside the boundary you choose, in-region or fully on-premise, and assessments never train a model.
Compare the fine print →Nothing probabilistic or statistical decides a grade: no distribution to drift toward, no phrasing bias to discriminate. A criterion fails on missing evidence, and the record shows the quote.
Full capability from day one. No thousand-assessment warm-up, no term spent watching your markers. The five-hundredth outcome is decided exactly like the first.
Downloads submissions, uploads the marked document, writes page feedback, fills the scoring table and sets the workflow status. Approvals and amendments made in the platform go straight to your LMS. No file shuffling, no data handover.
09 · Going live
The brief, the rubric, your feedback-sheet template and dozens of human-marked scripts: the whole calibration pack.
Your units become versioned configuration: ACs, case studies, word-count rules, minimum requirements, bands, the official CIPD feedback sheet.
Outputs are checked against your markers' outcomes, the unit runs in shadow mode with everything held for approval, and it goes live only when calibration says so.
Units & criteria · case studies · word-count rules · minimum requirements · rubrics & bands · the official sheet
10 · Extras
Recorded presentations get transcripts, per-criterion evidence detection and a draft observation form.
Attempt-keyed, marked against exactly what was referred, with CIPD caps applied automatically.
A new cohort year is a new configuration version. Both stay runnable and auditable.
Sampling, review-before-release and batch approval at the points your IQA policy names.
Extraction credits what is present, not how fluently it is phrased. Multilingual stack included.
Feedback rendered row for row into the CIPD feedback-sheet template your centre files.
Under two minutes on instant plans; a daily sweep on standard. The eligibility checks run the moment work arrives, and an ineligible script goes back with its reasons attached. Either way, no learner waits a marking SLA to hear their signature page was missing.
Yes. Sokros connects to the popular LMS platforms and knows where an assessment lives: unit, level, cohort year. It pulls on your schedule, marks, then completes your platform procedure: attaching the document, selecting the criteria grade buttons, writing page feedback, setting the workflow state and saving or releasing. Fully automated, or held in the dispatch queue for one-click batch release.
Because nothing probabilistic decides a grade. Core requirements are graded as binary present-or-not against extracted evidence. Quality is judged only after they stand. Written policy resolves the outcome, and every decision carries its verbatim quotes. Same script in, same grade out, replay-proven on every release, with the audit record EQA expects.
Wherever you want them. Sampling, review-before-release and batch approval modes put human judgement exactly where your IQA policy asks for it, and integrity signals are always for a person to decide on, never an automatic penalty.
Referred scripts return keyed by attempt. The resubmission is marked against exactly what was referred: additions checked against the gap, CIPD resubmission caps applied automatically, and the first attempt's record untouched.
The brief, the rubric, your feedback-sheet template and a calibration set of human-marked scripts: dozens, not thousands. Sokros is fully capable from the first assessment. It does not need to watch your markers for a term before it works.
More across the whole platform on the full FAQ →
Dozens of your marked scripts calibrate it. The gates decide go-live. The record defends every grade after that.
On the call: 30 minutes, one unit scoped, a calibration plan. You watch your own scripts marked before anything integrates.