In short: artificial intelligence (AI) can help with some special educational needs coordinator (SENCO) administration, but a school should not paste identifiable pupil information into an unapproved tool and hope for the best. The safe starting point is a defined task, an approved system, a clear data-protection position, human checking and a record of how the output was used.
AI should support drafting and organisation. It should not decide a pupil’s needs, determine provision without review, or replace the professional conversation with the pupil, family and staff.
This is practical guidance, not a substitute for advice from the school’s data protection officer, information technology (IT) lead or legal advisers. The school remains responsible for checking that a particular use is lawful and approved.
Start with the task, not the AI label
“SENCO admin” covers very different activities. Some are low-risk drafting tasks. Others involve special category data, decisions about children or information that could materially affect a pupil.
Before choosing a tool, define:
- what the person is trying to produce;
- what information the task actually needs;
- who will see the output;
- whether the output affects a decision about a pupil;
- what a human will check before it is used;
- how the source and final record will be stored.
A school may find that the safest first use is structuring an already approved set of information into a draft checklist. It may decide that another use, such as generating a recommendation about a pupil, should not be automated.
Eight checks before using AI with special educational needs and disabilities (SEND) information
1. Is the tool approved for this purpose?
Check the school’s policy, procurement route and approved-tool list. A free public chatbot and a contracted education service do not automatically have the same controls, terms or data flows.
Ask the supplier what the tool is designed to do, what data it processes, where processing occurs, how long it is retained and whether it is used to train a model. Keep the answer with the school’s procurement and data-protection record.
2. Can the task be completed with less data?
Use data minimisation. Remove names, dates of birth, addresses, unique identifiers and unnecessary details when the task does not need them. Do not include a full report when a short, de-identified summary is enough.
Pseudonymisation is not the same as anonymisation. A school that can reconnect a record to a pupil should continue to treat it as personal data.
3. Has the DPO or IT lead checked the arrangement?
The Department for Education advises schools to seek advice from their data protection officer or IT lead and to understand how an AI tool uses personal data. This should happen before staff start using the system with pupil information, not after a problem occurs.
The school should understand its role, the supplier’s role, the instructions for processing, security measures, breach process, sub-processors, retention and deletion route. The contract label alone does not answer every question about what happens in practice.
4. Is the lawful and transparent use understood?
A school needs a lawful basis for processing personal data and should be open about how it uses generative AI. Its privacy information, policies and staff training should not leave families guessing about whether personal information is processed by a new tool.
The exact lawful basis and any additional condition for special category data need to be confirmed locally. Do not treat consent as a shortcut for every school use.
5. Is there a human review before the output is used?
AI can produce fluent text that is wrong, incomplete or based on a misunderstanding of the source. A named member of staff should check the output against the source information and the pupil’s actual circumstances.
The reviewer should check names, dates, pronouns, needs, provision, professional advice, pupil voice and any statement that could be read as a decision. Keep the final professional responsibility with the school.
6. Could the output introduce bias or erase the pupil’s voice?
A summary can make a complex pupil sound simpler than they are. It can also overemphasise deficits, copy assumptions from historic records or omit the pupil’s priorities.
Compare the draft with the original source. Check that strengths, communication preferences, family views and reasonable adjustments have not disappeared because they were harder for the system to summarise.
7. Can the school explain what was done?
Record the purpose, tool, type of information used, reviewer and final decision where the use is material. Staff should know when they are looking at AI-assisted draft text rather than a source record or professional assessment.
This does not mean adding a warning to every sentence. It means keeping enough governance information to answer a sensible question about how the document was created.
8. Can the school retrieve and delete the information?
Ask how the school can export its records, correct them, respond to rights requests, manage retention and delete information when it is no longer needed or the contract ends. A tool that makes drafting easy but makes exit unclear is not a complete school solution.
A safer workflow for AI-assisted SEND documents
A practical workflow looks like this:
- The SENCO defines the document and the decision it supports.
- The school confirms that the tool and use are approved.
- Staff provide the minimum necessary source information.
- The system produces a draft or structured suggestion.
- A qualified member of staff checks it against the sources and pupil context.
- The pupil and family views are added or confirmed where relevant.
- The final record is approved, dated and stored in the school’s controlled system.
- The school records any follow-up action and review date.
The important control is not the word “AI”. It is the chain from source to draft to human decision to current record.
What to ask an AI SEND supplier
Before procurement, ask:
- Does pupil data train a model?
- Is AI processing zero retention, and what does that mean in the contract and technical design?
- Where is data processed and stored?
- What information is logged and for how long?
- Can the school control access and see an audit trail?
- Can the school export and delete all customer data on exit?
- How are generated documents checked and presented to staff?
- Can the supplier explain uncertainty, source grounding and known limitations?
- What support is available for the school’s DPO and IT team?
These are procurement questions, not badges that automatically make a product safe. The school still needs to assess the proposed use.
How MeritDocs fits
MeritDocs uses AI-assisted drafting for SEND documents with a workflow built around source information and professional review. Its approved product controls include UK and European Union (EU) processing, zero-retention AI drafting, no use of pupil data to train models, encryption, access controls, audit logs, export and full customer-data deletion on exit.
The platform is intended to help staff draft and manage documents, not make decisions about pupils. Schools still control permissions, check the output, decide what belongs in the record and apply their own policies.
Implementation support for schools
MeritDocs can arrange an on-site visit to demonstrate the platform live to the staff who need to use it, including relevant SEND staff and the existing IT team. The team can help set up the workspace and support migration of existing SEND records alongside those staff.
That support should be agreed around the school’s requirements and migration plan. Existing records vary in format and quality, so the school should review imported information, warnings and unmapped content before adopting it as the current record.