Autonomy-Preserving Support Systems
Event-driven decision support for people with large support teams. Messages between team members, scheduled reminders, observations and changes to note are routed through policy-based automation that augments human judgment without replacing it.
Built because the person at the centre of a support system should be the one steering it.
The Problem
People with large support teams generate many decisions every day. Messages between team members, scheduled reminders, everyday choices, appointment coordination, observations.
The person at the centre should be the one steering those decisions, with their expressed preferences as the primary signal.
The insight: What if we could automate routine decisions while actively protecting the person's expressed preferences from being overridden by support workers?
Core Principles
Non-negotiable commitments that guide every design decision.
AI Extends People's Judgement
The system extends human judgement. Every automated decision can be overridden. Complex or novel situations escalate to a human.
Expressed Preferences First
The person's expressed preferences are the primary signal. Not what support workers think they want. Not what historical patterns suggest. What they actually indicate they want.
Graduated Autonomy
The system earns direct-send authority through consecutive human approvals. Autonomy is granted incrementally based on demonstrated accuracy, not all-or-nothing.
Power-Check Safeguard
Continuous monitoring detects when support workers' preferences override the person's actual choices. Flags capacity changes, worker anomalies, and pattern deviations.
Event Processing
Every event a support organisation handles is routed through classification, pattern matching, policy application and confidence evaluation.
| Event Type | Automation Level |
|---|---|
| Messages between team members | |
| Scheduled reminders | |
| Everyday choices | |
| Observations | |
| Changes to note | |
| Roster and appointment coordination |
The Guardian Layer
Continuous cross-stream analytics that protect autonomy and catch problems early.
Support consistency review
Flags anomalies in team communication patterns and response quality.
Change detection
Notices when a person's choices shift, so the team asks rather than assumes.
Pattern identification
Groups related events so the team can see triggers, patterns and what has worked before.
Preference Drift Detection
Notices when recorded preferences start diverging from expressed choices—a signal of external influence.
Dialectica Integration
Two-role model: offline policy generation and on-demand complex analysis.
Role A: Policy Generation
For novel decision categories, runs Dialectica to determine: "Should this type of decision be automated?" Considers multiple perspectives—accessibility, ethics, legal constraints, practical effectiveness—and outputs an automation policy.
Example: Everyday choices: automatable above an agreed confidence threshold. Hard rules apply. Anything new escalates the first time.
Role B: Complex Analysis
When routine policies can't resolve a decision, routes to Dialectica for multi-perspective analysis. Used for conflicting professional advice, ethical dilemmas, and novel situations.
Example: Two professionals give conflicting advice. Present every perspective side by side, so the person and their team decide.
Why This Matters
AxelV1 was designed for people with large support teams: many workers, several services, a full calendar.
Without technology that actively preserves a person's preferences, a system that size drifts towards whatever is easiest to roster. Choices get averaged out. Autonomy erodes through a thousand small decisions made without the person's input. AxelV1 keeps the person's expressed choices in charge.
AxelV1 exists because everyone deserves technology that amplifies their voice, not silences it.
Technical Architecture
Why aXai
aXai is a disability-led AI, software and accessibility company. We build AI systems, software, and concepts and prototypes ready to scale, and we deliver them through AI-supported work: skilled people with disability in real, properly paid roles, working to clear quality standards, with AI removing the barriers and people keeping the accountability.
Every project runs on our own infrastructure, the Decision Bus and Project Layer, so each step is planned, signed off by a person and on the record. That is our operating advantage. Read how it works.
Interested in Autonomy-Preserving Systems?
Whether you're building for disability support, aged care, or any domain where automation must preserve human agency, we'd love to discuss the architecture.