Plain-English summary. We are SLT AI, the people behind School Assistant. Schools are the controllers of the data in their assistant; we are the processor. We log the text of parent questions for 30 days (no names, no IP addresses). When a question is routed to a person, the verbatim message is emailed to the school in real time - that email, in the school’s own inbox, is its safeguarding record. Sub-processors are listed in section 11 below.
1. Who we are
School Assistant is operated by SLT AI, a UK partnership. We are registered with the Information Commissioner’s Office under registration ZB378294. You can reach us at info@schoolassistant.co.uk. Postal correspondence for data-protection queries: email first and we will reply with a postal address.
2. What data we collect
Schools (our customers)
- Account information: contact name, email address, role, school name, billing address.
- School metadata pre-populated from the DfE’s GIAS register (see section 7).
- Documents uploaded or emailed in (see section 8).
- Pages scraped from the school website with the school’s consent.
- Designated Safeguarding Lead (DSL) email address for safeguarding routing (see section 5).
- Usage analytics (query counts, topic summaries and the text of visitor questions - see below).
- A log of school-staff actions in the portal, kept while the subscription is active (see section 6).
Website visitors (parents, carers and other visitors)
- We store the text of the question a visitor asks the chat widget (truncated to 200 characters), together with a timestamp, whether it was answered and a topic classification. This powers the school’s analytics dashboard and the weekly knowledge-gap digest.
- We do not store visitor names, email addresses, phone numbers, IP addresses, cookies or any other identifying information against these questions by default.
- If a visitor uses the “email the office” button, the name and email address they enter are passed to the school’s office email and are not retained beyond email delivery logs.
- If a visitor requests a callback, the name and phone number they enter are stored so the school can respond; the school controls retention and deletion of that record.
- If a visitor’s question is routed to a person under the sensitive-question policy (see section 5), the verbatim message is also emailed to the address the school has nominated for that routing.
- The widget detects the language of each question for multilingual replies (see section 4).
3. How we use data
- School documents are processed and stored to provide AI-powered answers to visitor questions.
- Each school’s data is isolated from all other schools at the vector-namespace and database-query level.
- Visitor question text is used to provide usage reports to the school, power the weekly knowledge-gap digest and identify common topics.
- School-staff actions in the portal are logged to support account security and troubleshooting.
- Answers are generated only from content the school has approved for indexing, with a source citation on every reply; the assistant states when it cannot find an answer rather than guessing. See the FAQ for how outdated or conflicting documents are handled and the residual risk of AI error.
4. Multilingual processing
The chat widget detects the language of each visitor question and replies in the same language across eleven supported languages (English, Polish, Romanian, Urdu, Punjabi, Arabic, Bengali, Somali, Portuguese, Spanish and French). To do this, the question text is sent to OpenAI, which performs language detection, translation and answer generation under our data-processing terms. Non-English question counts are aggregated to drive the school’s “% non-English” EAL dashboard; individual non-English text is treated identically to English text and retained for the same 30-day window.
5. Safeguarding data
The widget recognises parent questions that should be handled by a person rather than AI - for example questions about reporting bullying, raising a concern about a member of staff, or a child’s wellbeing, as well as language consistent with self-harm, abuse or mental-health crisis. In these cases the assistant does not attempt an answer: a classifier short-circuits the LLM and the visitor sees a canned response signposting the school’s designated safeguarding contact and, where appropriate, external services such as NSPCC and Childline. In parallel, a transactional email containing the verbatim message and timestamp is sent to the school-designated contact (Designated Safeguarding Lead address or office address, as the school configures). The lawful basis for this processing is vital interests (Article 6(1)(d) UK GDPR) and the substantial public interest in safeguarding under Schedule 1 of the Data Protection Act 2018. Schools are required to maintain a current routing email address and to handle these notifications under their KCSIE policy.
No automated classifier can guarantee it will recognise every safeguarding-relevant message. As a safeguard against a missed detection, schools should keep their normal safeguarding-reporting channels (posters, staff briefings, the DSL’s published contact details) visible on their site independently of the assistant, and can use the portal’s chat log at any time to review conversations, including ones the classifier did not flag.
6. Portal activity logging
Actions taken by school staff inside the portal (document uploads and deletions, settings changes, template approvals, access grants, exports and safeguarding triage decisions) are logged so the school can review account activity. These logs are kept while the subscription is active and are deleted, along with all other school data, within 30 days of cancellation (see section 13). The school’s own safeguarding record is the real-time routing email sent to its DSL or office address under section 5, held in the school’s own systems under its KCSIE policy.
7. GIAS data
We ingest data from the Department for Education’s Get Information About Schools (GIAS) register, which is published under the Open Government Licence v3.0. During onboarding, we use the school’s URN/DfE number to pre-populate school metadata (name, address, phase, headteacher, pupil count and similar fields). We refresh this data periodically so the school’s record stays current. Schools can edit any pre-populated field, and inaccurate GIAS data should be corrected through DfE channels.
8. Email-to-knowledge
Each school is allocated an inbound email address (e.g. docs-TOKEN@inbound.schoolassistant.co.uk). Documents emailed to this address are ingested into the school’s knowledge index. We log the sender email address, subject line and arrival timestamp; spam and attachments above size limits are rejected without ingestion. The Postmark inbound service is the sub-processor for this pathway (see section 11).
9. Knowledge-gap analysis
To produce the weekly knowledge-gap digest and the “top questions” report, we group anonymised visitor question text into topic clusters and present aggregates to school staff. The text shown in these aggregates may include the original wording of parent questions, with no identifiers attached. Schools can opt their portal display out of question-text aggregation in the settings panel; the aggregated counts continue regardless.
10. Security
- All data transmission is encrypted via TLS 1.2 or higher.
- Per-school namespace isolation in Pinecone; per-school role isolation for portal access; per-school API tokens with origin lock for widget calls.
- Bot/scraper protection via origin lock, per-token rate limits, per-IP rate limits and a hard monthly cost cap.
- Vector store rebuilds are blue/green with atomic namespace swap, so there is no widget gap during the rebuild.
- We notify schools of personal-data breaches without undue delay and within 72 hours of becoming aware where required by Article 33.
11. Sub-processors
Last reviewed: 14 September 2026. We use the following sub-processors to deliver the service. Each is bound by data-processing terms equivalent to those we offer schools.
Our AI sub-processors (OpenAI, Anthropic) are contracted under API/business-tier terms, not consumer product terms: none of them use documents, website content or visitor questions submitted through School Assistant to train or improve their underlying models.
| Sub-processor | Purpose | Region |
|---|---|---|
| SiteGround | Web hosting for the marketing site, portal and widget API | UK / EU |
| Pinecone | Vector database for school knowledge embeddings | US (SCCs in place) |
| OpenAI | Embeddings, language detection, answer generation | US (SCCs in place) |
| Anthropic | Template synthesis and page classification during onboarding | US (SCCs in place) |
| Firecrawl | School website scraping with the school’s consent | US (SCCs in place) |
| Postmark | Transactional email and inbound email-to-knowledge | US (SCCs in place) |
| Google Cloud Vision | OCR for image-based PDFs, on demand | US (SCCs in place) |
We notify schools by email before adding or removing a sub-processor. Schools may object within 14 days; if a substitute cannot be agreed, the subscription can be terminated with a pro-rata refund.
12. Cookies
The marketing site (schoolassistant.co.uk) uses a small number of cookies, grouped as follows:
- Strictly necessary - WordPress session cookies for logged-in users, cart cookies on the shop and checkout pages. These are always on.
- Analytics - first-party analytics if you accept the cookie banner. We do not run third-party advertising trackers.
- Widget on school sites - the embedded chat widget does not set cookies. A per-tab session identifier is stored in the browser’s
sessionStorageand the visitor’s chosen language preference inlocalStorage.
You can manage your cookie choices via the banner that appears on first visit. Withdrawing consent at any time will not affect the lawfulness of prior processing.
13. Data retention
- School documents are retained while the subscription is active. Documents with expiry dates are auto-removed when they expire.
- On subscription cancellation, all school data (documents, vectors, settings, templates) is retained for 30 days then permanently deleted.
- Visitor question text is retained for 30 days then automatically deleted. Aggregate counts (lifetime totals) are retained beyond this window to drive the school dashboard.
- Portal staff-action logs are kept while the subscription is active and deleted within 30 days of cancellation (see section 6).
14. Accessibility
We aim to make schoolassistant.co.uk and the embedded widget conform to Web Content Accessibility Guidelines (WCAG) 2.2 at Level AA. The widget supports keyboard navigation, screen-reader landmarks, language switching to match the detected reply language, and high-contrast colour pairs. Issues can be reported to info@schoolassistant.co.uk; we aim to respond within five working days. This statement is published in line with the Public Sector Bodies (Websites and Mobile Applications) Accessibility Regulations 2018 for the benefit of public-sector buyers.
15. Your rights (UK GDPR)
As a school, you are the data controller for the documents you upload and the visitor questions logged against your namespace. School Assistant acts as a processor. You have the right to access, rectify, erase, restrict, port and object to processing. To exercise any of these rights, contact us at info@schoolassistant.co.uk. You also have the right to complain to the Information Commissioner’s Office at ico.org.uk.
16. Data Processing Agreement
A Data Processing Agreement is available at /dpa/ and is auto-applied at checkout. Schools may also countersign their own DPA template, subject to review.
17. Changes to this notice
We update this notice when our processing changes. The last-updated date at the top reflects the most recent material change. Schools are notified by email of material changes that affect processing of their data.
18. Contact
For privacy-related queries, email info@schoolassistant.co.uk.