Skip to main content

Project Submission and Feedback: How AI Marking Works

This guide explains how the AI marking tool assesses project submissions, expected feedback timing, automatic retries, grade meanings, resubmission, image requirements, and when to request an Expert Review.

Summary

Instant Feedback gives the Apprentice the insights needed to master skills and progress through a programme. This guide explains the full process, from initial project submission on The Platform to Expert Review.

How the AI Marking Process Works

When the Apprentice submits a project, a dedicated AI marking tool assesses the work against the required Knowledge, Skills, and Behaviours (KSBs) for that submission. The AI marking tool is a specialised model designed for grading. The AI marking tool operates separately from the Atlas chatbot.

Atlas can advise on submissions and help the Apprentice generate ideas. Atlas cannot predict whether a submission will receive a passing grade. Only the AI marking tool assesses work against the specific rubrics and evidence of KSBs.

Feedback Timing and Automatic Retries

Feedback typically arrives within two minutes of submission. In some cases, feedback can take up to 24 hours.

If feedback does not appear immediately, the system automatically re-processes the submission at the following intervals:

  • Five minutes after the initial submission.

  • 20 minutes after the first retry.

  • Every six hours after that.

Important: Wait 24 hours before raising a support ticket. This allows the automated retry cycles to complete.

What Each Project Grade Means

The following table explains each grade the AI marking tool can award.

Grade

Meaning

Exceeds Expectations

The submission successfully demonstrates the required criteria to a high standard.

Meets Expectations

The submission successfully demonstrates the required criteria.

Needs More Work

One or more criteria for a KSB are missing. The feedback identifies which areas require further development.


How to Resubmit a Project After Feedback

Multiverse encourages an iterative approach to learning. After feedback arrives, follow these steps:

  1. Read the feedback and select a level of agreement.

  2. Provide a brief reflection if prompted.

  3. Edit the project to address the feedback.

  4. Resubmit the project for further feedback.

The Apprentice can edit and resubmit a project as many times as needed to achieve the required KSBs. If the requirements for a passing grade are unclear, the Apprentice can discuss the AI feedback with the Coach.

Non-portfolio Assessments

Some programmes contain assessments that are not included in the portfolio. These relate to Assessment Method 2 (AM2), the interview or presentation element of the End Point Assessment (EPA). Non-portfolio assessments practise presentation skills, review progress, or prepare the Apprentice for Assessment Method 1 (AM1), the work-based project. Every assessment that receives written feedback follows the process described in this guide.

Image and Technical Requirements for Submissions

Follow these standards so the AI marking tool can accurately analyse the work:

  • Image limit: Include no more than 50 images per project submission. A full portfolio can contain more than 50 images in total.

  • Image formats: Supported formats are PNG, JPEG, WEBP, and GIF. Only the first frame of a GIF is processed.

  • Charts and graphs: The AI marking tool can struggle with dense data or overlapping labels. If a human cannot interpret the chart at a glance, the AI marking tool is likely to misinterpret it.

  • Complex visuals: If a necessary but complex chart causes an incorrect "Needs More Work" grade, use the Disagree function to escalate for human review.

  • Resolution: Blurry or low-resolution text can cause the AI marking tool to stall. Check that everything is legible after uploading.

How to Request an Expert Review

If the AI marking tool has made a genuine error, the Apprentice can trigger a human review. Select Disagree on the feedback form and provide a detailed explanation.

Important: The project is locked during the human review. The Apprentice cannot edit or resubmit the project until the Instructor completes the assessment, which takes up to 10 working days.

Valid Reasons for Escalation

  • Missed evidence: The submission received a "Needs More Work" grade, but the feedback states a process, detail, or KSB is missing that the Apprentice is certain is present.

  • Technical graph detection: The AI marking tool failed to recognise a complicated technical graph or chart that is essential for a specific KSB.

  • Inaccurate passing grade: The AI marking tool awarded a passing grade for a KSB despite incomplete evidence.

Invalid Reasons for Escalation

  • General preference: The Apprentice would prefer a human review over an AI assessment.

  • Grade improvement: The submission received "Meets Expectations" but the Apprentice wants "Exceeds Expectations". Instead, discuss the work with the Coach and resubmit for further feedback.

  • Technical failures: The submission failed or stalled because it exceeded the 50-image limit or used an unsupported file format.

How Feedback Quality Is Monitored

The AI marking tool performs at a level comparable to human markers. Escalations to human reviewers are a critical part of the quality assurance loop. Any discrepancies identified during human reviews are tracked and relayed to the Product and Learning teams to improve the tool.

Multiverse also conducts proactive moderation by comparing disputed projects with non-disputed ones. This checks consistency and refines the models. Every piece of feedback the Apprentice provides helps improve the tool for all learners.


Frequently Asked Questions

Questions

Answers

Question: How long does project feedback take?

Answer: Feedback typically arrives within two minutes of submission but can take up to 24 hours. The system automatically retries stalled submissions, so wait 24 hours before raising a support ticket.

Question: Can Atlas tell the Apprentice whether a project will pass?

Answer: No. Atlas can advise on submissions and help with ideas, but only the dedicated AI marking tool assesses work against the rubrics and KSBs.

Question: What is the difference between AM1 and AM2?

Answer: Assessment Method 1 (AM1) is the work-based project element of the End Point Assessment (EPA). Assessment Method 2 (AM2) is the interview or presentation element.

Question: Why is a project locked after selecting Disagree?

Answer: Selecting Disagree triggers an Expert Review by an Instructor. The project stays locked until the review completes, which takes up to 10 working days.

Question: How many times can the Apprentice resubmit a project?

Answer: There is no limit. The Apprentice can edit and resubmit as many times as needed to meet the required KSBs.

Did this answer your question?