v1.2.0May 2025
AI Operations: Review Queue, policy governance, risk-level routing
v1.2.0 ships the full AI Operations layer — a unified approval inbox, configurable governance policies, and intelligent risk-level routing that keeps humans in control without slowing operations down.
What's new
- Review Queue — a unified inbox that surfaces every pending AI action, invoice approval, and inventory adjustment that needs human sign-off. Filter by type, assignee, or risk level. Approve or reject in one click.
- Policy governance — define approval thresholds per role. Set monetary limits, inventory variance tolerances, and supplier-level rules. Any action that crosses a threshold is automatically routed for review.
- Risk-level routing — the AI now tags each generated action with a risk score (Low / Medium / High / Critical). High and Critical actions are blocked from auto-execution and always require a manager or owner approval.
- Bulk actions in Review Queue — approve or reject up to 50 items at once with an audit comment. Full audit trail maintained for every bulk operation.
- Review Queue notifications — get in-app and email alerts when new items land in your queue. Configurable digest frequency (instant, hourly, daily).
- Policy templates — five pre-built policy packs (Retail Standard, High-Volume, Multi-Location, Franchise, Minimal Friction) to get you set up in under 5 minutes.
Improvements
- AI Copilot now cites the specific data source and timestamp for every recommendation, making it easier to audit suggestions before acting.
- Forecasting — stock risk alerts now appear inline in the Review Queue as low-risk informational items, not just on the forecasting page.
- Audit log — all Review Queue decisions (approve/reject/bulk) are now indexed and searchable by date, actor, item type, and outcome.
- Role permissions — Managers can now be scoped to review only their assigned departments, reducing noise in high-SKU environments.
- Performance — Review Queue loads up to 3× faster on accounts with more than 500 pending items.
Bug fixes
- Fixed: Review Queue count badge in the sidebar was not clearing after bulk approval.
- Fixed: Policy threshold editor was accepting negative values for monetary limits.
- Fixed: AI-generated purchase orders occasionally skipped the approval step when the supplier had a pre-approved status set.
- Fixed: Risk score labels were not localising correctly in non-English browser locales.