Student complaints, handled properly.
ReCourse runs your published procedure from the student's form to the disposal review. The AI reads each case against your own policy and advises. Your staff decide. Hosted by us or by you.
Without a system, complaints live in inboxes
Inboxes and spreadsheets
A complaint arrives by email, gets a row in a spreadsheet and lives in three mailboxes. Nobody can say who has read it, what stage it is at or when it is due.
One complaint, three answers
The outcome depends on who picks it up and which copy of the procedure they have to hand. The OIA notices.
Staff already use AI. You cannot see it.
Handlers paste complaints into consumer chatbots to summarise and draft. Special-category data leaves the university under terms nobody has read, with no record it happened. The problem is not the AI. It is the invisibility.
Deadlines missed quietly
Working-day targets are counted by hand, or not at all. You find out when the student asks, or when the OIA does.
One system for the whole procedure
From the student's form to the disposal review, in four steps.
- 1 The student submits. A validated form, a reference, and a due date counted in working days from your stage targets.
- 2 The AI reads it against your policy. A pseudonymised copy goes to the AI provider with your published procedures. Back comes a summary, a next step, a rating and quotations from the policy.
- 3 Every quotation is checked. Each one is matched to the real policy text before staff see it. No match, no quote.
- 4 Your staff decide. The handler records a decision with a rationale and writes the letter the student receives. The student never sees the assessment.
-
Submit
-
Pseudonymise
-
Assess
-
Verify
-
Decide
Why universities choose ReCourse
Less time per case
- Statuses, stages, deadlines and a queue sorted by what is due.
- Cases open with a summary, the relevant clauses and a next step already there.
- One message thread, attachments and assignment replace the email chain.
- Re-assess in one click when a case moves stage.
A process off the shelf
- Three stages. Your names, windows and targets, set on a settings page.
- Every stage change carries a note the student reads.
- Categories you control. The AI works out which procedure applies.
- Deadlines are advisory. The system never escalates or closes a case on its own.
Less risk
- One outbound connection: the assessment call. No analytics, no CDN, no external fonts.
- Pseudonymised on your server before anything is sent. If that fails, nothing is sent.
- An audit log of who viewed each complaint, not only who changed it.
- Erasure, anonymisation and a disposal list for a person to review. Nothing is destroyed by the clock.
- A public Report and Support form, anonymous or with consent. Never assessed, never sent anywhere.
- Emails say something happened. Never what.
Built for the sector
- One institution per installation.
- Handler, investigator and read-only HR roles.
- An accessibility statement, full keyboard use, and no task that needs JavaScript.
- A strict content-security policy and a privacy notice students can read.
How the AI is used
What it does, and what it does not.
- It advises. It does not decide.
- Every decision is recorded by a named person with their own rationale. The letter is written by a person, in a separate field from the AI's reasoning.
- It cites, and we check the citations.
- Each quotation is matched to the policy on file. Unmatched quotations are counted and never shown as policy.
- Its output is structured.
- The provider's interface constrains the response to a fixed shape.
- You can see what was sent.
- Every assessment keeps the exact text that left the server. Staff read it on the case.
- Pseudonymised is not anonymous.
- Names, contact details and student numbers are replaced first. Free text can still identify someone, so it stays personal data and is treated as such.
- Provider terms, as at 15 August 2026.
- Complaint text is not used to train AI models and is held for 30 days. Zero data retention is available by application, and we recommend it before live use. It is a contract, not a setting.
- Students never see the assessment.
- Separate response models and a permission check on every route, pinned by tests.
Hosted by us or by you
On your servers
A container image with PostgreSQL, behind your proxy on your domain. Your AI provider account, your backups, your retention schedule. One outbound connection: the assessment call.
On ours
The same software, run for you under a written processing agreement. One institution per installation, one outbound connection, no analytics. We are your processor and the contract says so.
The data-protection story is the same either way. It belongs to the software, not to where it runs.
What is next
The roadmap as it stands. No dates promised.
- Single sign-on through Shibboleth and the UK Access Management Federation.
- OIA Completion of Procedures letters from the case record.
- Reporting: counts by category, time to resolution, outcomes.
- Attachments on incident reports.
- An independent accessibility audit and testing with disabled users.
- Zero data retention with the AI provider as the default.
- Password reset and emailed staff invitations.
See it with your own procedure
We upload your published policy to a sandbox. The assessment cites your clauses, not ours.