Local-first intelligence
We investigate useful AI that can work close to the user and minimise unnecessary movement of educational information.
SIRA Research
Intelligence that strengthens responsibility.
We explore how AI can reduce administrative friction while keeping people, permissions, and educational judgment in control.
Active researchControlled testingRoadmap, not a release promise
Human-controlled action
We design approval boundaries that keep sensitive decisions with authorised people and preserve a clear activity trail.
Role-aware assistance
We study how an assistant can provide relevant help without exposing information outside a user’s responsibilities.
Current direction
Research in progress.
SIRA is presently focused on reliable local assistance, transparent automation, educational coordination, and privacy-preserving personalisation. Formal publications and technical reports will appear here as they become available.
Exploration roadmap
What we are investigating next.
These areas guide experiments and tester feedback. They may change, remain on hold or never become production features.
Reliable administrative assistance
Improve role-aware summaries, task planning, document organisation and safe draft generation across everyday educational work.
- Better context within permitted roles
- Clearer sources and assumptions
- Faster feedback from controlled testers
Connected educational workflows
Study how email, calendars and shared workspaces can trigger useful actions without removing human control.
- Google Workspace connections
- Reusable workflow templates
- Approval and exception handling
Additional platforms and versions
Explore broader device support and alternative product experiences after the core desktop testing experience is stable.
- Mobile companion concepts
- Web workspace research
- Additional operating-system support
Deeper educational coordination
Investigate carefully scoped tools for cross-team reporting, service coordination and responsible organisational insight.
- Privacy-preserving trend summaries
- Department and faculty coordination
- Configurable governance controls
How we evaluate
Useful is not enough. It must also be understandable and controllable.
Start with a real problem
We study repetitive or fragmented administrative work rather than adding AI where it is unnecessary.
Test within boundaries
Experiments use defined roles, permitted information and clear expectations for human review.
Measure the trade-offs
We look at usefulness, accuracy, time saved, clarity, privacy, accessibility and failure behaviour.
Decide with evidence
A concept can advance, change direction or remain on hold depending on feedback and risk.
Items marked “next”, “later” or “on hold” are research directions, not commitments or announced release dates. SIRA will only describe a capability as available after it has been implemented and verified for the relevant testing or production stage.