AI Settings
Configure how AI features work across your practice, including which capabilities are available, who can access them, and how usage is managed.
Overview
AI settings are configured at two levels:
| Level | Who configures | What it controls |
|---|---|---|
| Practice-wide | Practice owner / manager | Available features, AI roles, global toggles |
| Per-practitioner | Individual practitioners | Personal preferences, suggestion level, specialty focus |
Practice-wide settings act as the ceiling -- individual practitioners can only enable features that are available at the practice level.
Subscription Tier Capabilities
Your subscription tier determines which AI capabilities are available to your practice:
| Feature | Student | Essentials | Pro | Enterprise |
|---|---|---|---|---|
| AI chat assistant (help, scheduling, billing) | Yes | Yes | Yes | Yes |
| Clinical AI (notes and documents) | -- | -- | Yes | Yes |
| Research assistant (evidence, MBS lookup) | -- | -- | Yes | Yes |
| Analytics AI (revenue, practice insights) | -- | -- | Yes | Yes |
| Proactive suggestions | -- | -- | Yes | Yes |
| Custom AI instructions | -- | -- | -- | Yes |
To upgrade your plan, go to Settings > Subscription or contact our sales team.
queries (scheduling, billing) use efficient models for fast responses, while clinical tasks (note writing, report drafting) automatically use more capable models for accuracy. Your subscription tier determines the maximum model capability available. :::
Practice-Wide Configuration
Navigate to Settings > AI Preferences to configure practice-wide AI settings.
Enabling AI Features
- Go to Settings > AI Preferences
- Toggle the master Enable AI Features switch
- Configure data access controls:
| Setting | Description | Default |
|---|---|---|
| Allow client data | AI can access client records for context | On |
| Allow clinical notes | AI can assist with note writing and review | Off |
| Allow voice transcription | Speech-to-text for sessions | On |
| Require client consent | Client must have AI consent before data is processed | On |
- Click Save
reasons. Practice owners must explicitly enable it after reviewing the privacy implications. :::
AI Agents
Praxi includes six specialist AI agents, each focused on a specific area of practice management:
| Agent | Specialty | Category |
|---|---|---|
| Praxi | Clinical notes, report drafting, document generation, compliance checks | Clinical |
| Ada | Research, evidence grading, MBS item lookup, knowledge management | Clinical |
| Jarvis | Scheduling, billing, Medicare claims, client communications, help | Administrative |
| Warren | Practice analytics, revenue tracking, growth insights, marketing | Administrative |
| Brene | Staff wellness monitoring, HR support, team dynamics, supervision | Support |
| Linus | Workflow automation, app development, integrations, technical support | Support |
Each agent can be individually configured at the practice level:
- Go to Settings > AI Preferences
- Select an agent to configure
- Adjust settings:
- Enabled/Disabled -- turn agents on or off for the practice
- Tone -- Professional, Friendly, Clinical, or Empathetic
- Verbosity -- Concise, Balanced, or Detailed
- Custom instructions -- practice-specific guidance for this agent
- Daily request limit -- cap daily usage per agent (optional)
Cost Controls
Manage AI spending with daily and monthly budget limits:
- Go to Settings > AI Preferences
- Configure:
| Setting | Description | Default |
|---|---|---|
| Daily limit | Maximum AI spend per day (minimum $1) | $10 |
| Monthly limit | Maximum AI spend per month (minimum $10) | $100 |
| On daily exceeded | Action when daily limit is reached | Notify |
| On monthly exceeded | Action when monthly limit is reached | Pause |
When a limit is reached, Praxi can take one of four actions:
| Action | Effect |
|---|---|
| Notify | Alert the practice owner but continue allowing AI requests |
| Downgrade model | Switch to a more efficient model tier to reduce costs |
| Pause | Temporarily pause AI features until the next period resets |
| Block | Fully block all AI requests until the limit is adjusted |
A budget status indicator appears in the AI panel showing current spend against limits. When usage reaches 80% of any limit, the indicator changes to a warning state.
Approval Thresholds
Control which AI-suggested actions require human approval before execution:
| Threshold | Behaviour |
|---|---|
| All | Every AI-suggested action requires approval |
| High risk | Only actions flagged as high-risk require approval |
| None | AI suggestions are auto-approved |
data always require practitioner approval. This is a regulatory requirement under the PsyBA Code of Conduct (2025). :::
Model Preferences
| Setting | Description | Default |
|---|---|---|
| Default model | Which model tier to use for general queries | Sonnet |
| Allow model fallback | Permit automatic fallback to a different tier | On |
Praxi selects the appropriate model tier based on the task. Routine queries (scheduling, billing) use efficient models, while clinical tasks (note writing, report drafting) automatically use more capable models.
Conversation Retention
Set how long AI conversation history is retained:
- Adjustable from 0 to 365 days
- Default: 90 days
- Setting to 0 disables conversation history entirely
Custom Instructions (Enterprise)
Enterprise plans can provide custom instructions that guide AI behaviour across the practice:
- Go to Settings > AI Preferences
- Enter instructions such as:
- Preferred clinical terminology
- Practice-specific documentation standards
- Tone and communication preferences
- Specialty-specific guidance
- Click Save
Custom instructions apply to all AI interactions for your practice, in addition to any per-agent and per-practitioner preferences.
Monitoring AI Usage
Audit Trail
All AI interactions are logged in the audit trail:
- Go to Settings > Compliance > Audit Logs
- Filter by Category: AI Interactions
- View details including:
- Who made the request
- What was asked
- What the AI suggested
- Whether the suggestion was accepted or rejected
- Timestamp and duration
Audit logs are retained for 7 years in compliance with the Privacy Act 1988.
Usage Dashboard
Monitor AI usage across your practice:
- Go to Settings > AI Preferences
- View metrics including:
- Daily and monthly spend (in dollars) against your configured limits
- Total requests today and this month
- Breakdown by agent (Praxi, Jarvis, Ada, Warren, Brene, Linus)
- Per-agent request counts (daily and monthly)
- Whether AI is currently paused due to budget limits
Privacy and Compliance
| Requirement | How Praxi meets it |
|---|---|
| Privacy Act 1988 | All AI processing complies with Australian Privacy Principles |
| No model training | Practice data is never used to train AI models |
| Australian hosting | AI processing occurs on Australian-hosted infrastructure |
| Data isolation | Strict tenant separation -- your data cannot be accessed by other practices |
| Consent | AI processing of clinical data requires client AI consent |
| Right to erasure | Client data can be permanently deleted upon request |
| AHPRA compliance | AI-generated clinical content meets documentation standards |
Client AI Consent
Before AI features can process clinical data for a specific client, the client must have AI processing consent recorded:
- Navigate to the client's profile
- Go to Consent tab
- Record AI Processing consent
- This enables features like AI note assistance for that client
Without AI consent, the assistant can still answer general help questions but cannot access or process that client's clinical data.
Troubleshooting
| Issue | Solution |
|---|---|
| AI features greyed out | Check your subscription tier supports the feature |
| "AI consent required" message | Record AI processing consent on the client's Consent tab |
| AI paused or blocked | Check your budget limits in Settings > AI Preferences |
| Slow AI responses | Clinical queries use more capable models -- allow a few extra seconds |
| Unexpected AI suggestions | Review custom instructions on the agent and per-practitioner settings |
| Agent not responding | Check the agent is enabled in Settings > AI Preferences and has not hit its daily request limit |
Related
- Chat Assistant -- Using the interactive AI chat
- AI Agents -- Overview of specialist agents
- Note Assistance -- AI-powered note writing
- Audit Logs -- Compliance monitoring
- Permissions -- Role-based access control