Architecture
Three AI pillars
Three distinct layers that work together to make your operations intelligent.
Insight Engine
Continuously analyses your operational data to detect patterns, surface anomalies, and recommend actions — before problems escalate.
- Real-time anomaly detection across sales and inventory
- Pattern recognition for demand shifts and stock risks
- Actionable recommendations with reasoning context
- Prioritised by business impact, not algorithm output
Advisory Copilot
A conversational AI that answers operational questions using your live business context — not generic knowledge.
- Ask in plain language: "What's our best-selling category this week?"
- Grounded in your live inventory, sales, and shift data
- Understands business context, not just raw numbers
- Session history for continuity across conversations
Automated Orchestration
Policy-governed AI actions that execute workflows automatically — with a human-approval queue for anything above your risk threshold.
- Trigger automation from AI insight outputs
- Policy rules gate actions by risk level
- Every action logged in the Review Queue
- Full rollback and override controls
AI providers
Every provider is policy-governed. You control which actions they can take.
Anthropic Claude
Primary reasoning and advisory copilot
Google Gemini
Supplementary intelligence layer
Heuristic
Rule-based deterministic fallback
Every AI action requires approval. You stay in control.
No AI action in FlexotiumPOS executes without passing through your policy rules. Actions above your configured risk threshold are queued for human review — with full context, reasoning, and a simple approve/reject interface.
Create purchase order
Coffee Blend 500g — reorder qty: 120 units
Triggered by: Insight Engine · Stock fell below 18 units
Governance
AI risk levels
Policies gate every AI action by risk level. You configure what auto-executes and what needs a human in the loop.
Auto-approved. Executes immediately.
Example
Generate a daily summary report
Flagged for review. 24h approval window.
Example
Create a purchase order for low stock
Requires explicit manager approval.
Example
Modify pricing rules system-wide