markus
  • Product
  • About
  • Pricing
  • FAQ
Sign in
Markus
AI Transparency & Trust  ·  April 2026
How Markus Works
Built by educators. Designed for real classrooms.

We built Markus to deal with the parts of teaching that take the most time and give the least back. The AI handles marking and feedback. Teachers handle everything that matters. That distinction is intentional.

What Markus Does

We work with what students actually produce. Written responses, including handwritten work, assessed the same way a teacher would assess them.

Marks student work
Automatically processes and marks written responses against mark schemes.
Generates personalised feedback
Produces tailored feedback grounded in each student's actual response.
Tracks progress over time
Gives educators a clear view of individual and cohort progress.
Highlights strengths and gaps
Surfaces what students know and where they need support.

How the AI Works

Most AI tools read a response once and give a mark. We don't think one pass is reliable enough.

1
Read more than once
Each response goes through multiple independent passes. No single interpretation is trusted on its own.
2
Approached from different angles
Each pass looks at the response differently, which reduces the chance of the same blind spot appearing every time.
3
Passes compared before a mark is given
If they agree, good. If they don't, that matters too. A final outcome is only produced once the passes have been weighed against each other.

Evidence Based Marking

A mark without supporting evidence is a guess. Before any score is given, the system has to point to what in the student's writing justifies it.

  • The system identifies where in the response the marks are coming from
  • It checks the evidence is genuinely present, not assumed
  • It won't award marks it can't back up
The feedback references what the student actually wrote. Not what we think they meant. What they put on the page.

Confidence and Reliability

Not every piece of work gets marked with the same level of certainty. When the passes disagree, or the evidence is thin, that gets flagged. A high confidence outcome means the passes agreed closely and the evidence was strong. A lower one means a teacher should take a closer look.


Teacher Control and Oversight

Every mark Markus produces can be reviewed, questioned, and changed. Teachers are never locked into what the AI has decided.

Teachers can always
Review all marks Override where needed Use insights to guide teaching
Markus doesn't make final decisions. It gives teachers better information to make them.

Safeguarding and Monitoring

We take our obligations under the DfE's Generative AI: Product Safety Standards seriously, particularly where the platform is used with young people. Safeguarding isn't a feature we added on; it's part of how the platform works.

Designated Safeguarding Lead (DSL) Support

Schools and colleges can nominate their DSL directly in the platform's admin settings. When student work is flagged as potentially harmful, the DSL is alerted so they can review it themselves, rather than waiting to hear about it secondhand.

  • Institutions nominate their DSL through the platform's administration settings
  • The DSL is alerted to flagged submissions for human review. Markus does not take autonomous action
  • DSL nominations can be updated by the institution's platform administrator at any time
  • Flagging covers indicators such as self harm language, safeguarding concerns, and other harmful content identified within submitted work
Keeping Children Safe in Education (KCSiE)

The safeguarding features are built to help schools and colleges meet their obligations under Keeping Children Safe in Education Part 1. The platform surfaces concerns that come through student work. The DSL reviews and acts on them. Policy responsibility stays with the institution.

  • The platform flags concerns. It doesn't investigate or act on them independently
  • DSL nominations can be updated by the institution's administrator at any time
  • Flagging covers indicators including self harm language and other harmful content in submitted work
DSL nomination is built in. Harmful content flagging happens automatically. The platform won't act on a flag itself, but it won't sit on it either.

What Markus Does Not Do

Worth being direct about:

  • Replace teacher judgement
  • Make high stakes decisions independently
  • Use student work to train AI models
  • Track or profile students for non educational purposes

Data and Privacy by Design

We keep the data footprint small by design. In most cases, we only process:

Student name Student email Assessment responses Feedback and marks
  • We do not use advertising, tracking, or behavioural analytics tools
  • All data is processed securely and in line with UK GDPR

Why This Matters

Assessment decisions affect students. If a teacher uses Markus to inform a grade, they should be able to explain to a student, a parent, or an inspector why that grade was given. That requires the tool to be honest about how it works, and it requires us to be honest about what it can and can't do.

Transparent
You can see how marks were reached and what evidence was used
Reliable
Multiple passes, compared outputs, flagged uncertainty
Honest
We tell you what the AI can do and where you still need a teacher

Our Approach

If you're using AI to assess students, you should be able to explain what it's doing. So should we. That's what this document is. If something here raises a question you'd like to discuss, get in touch.


Questions or want to know more?

We're always happy to explain how Markus works in more detail.
info@markus.tech
markus

Markus Tech Ltd · Company No. 16443285
Registered in England & Wales
ICO Registered · ZC106423
1 Trowbridge Road, London E9 5LD

© 2026 Markus Tech Ltd. All rights reserved.

Product
Initial Assessment Mock Exams Self Assessment Living Colour Data & Insights
Company
About Pricing FAQ Contact Wordhunt
Legal
Privacy Policy Terms & Conditions SLA DPA Sub-Processors AI Transparency
Security
markus
Privacy Policy Terms & Conditions SLA DPA Sub-Processors AI Transparency Contact
Markus Tech Ltd · Company No. 16443285
Registered in England & Wales · ICO Registered · ZC106423
1 Trowbridge Road, London E9 5LD

© 2026 Markus Tech Ltd. All rights reserved.
Security