What we've shipped, fixed, and improved. Updated with every release.
Launched Smart Escalation Paths — define multi-tier fallback rules when primary contact lists are exhausted. Configure secondary pools, cross-site borrowing, and automatic premium rate offers after configurable thresholds. Available via dashboard and POST /v2/shifts/{id}/escalation-rules.
Fixed a race condition where two workers accepting the same shift within a 200ms window could both receive confirmation messages. Added distributed locking via Redis with a 5-second TTL on shift assignment mutations. Zero double-bookings since deployment.
Reduced voice agent cold-start latency from 2.1s to 680ms by pre-warming TTS model instances during off-peak hours. P99 first-word latency for outbound calls now under 900ms across all regions.
Added native integration with UKG Pro Workforce Management (formerly UltiPro). Bidirectional sync for shift schedules, worker profiles, and time-off requests. Supports multi-entity configurations. Setup takes under 15 minutes via OAuth flow.
Shipped Conversation Analytics Dashboard — real-time visibility into AI agent conversations. Track acceptance rates by script variant, identify common objections, see sentiment analysis per worker segment. Exportable reports for compliance auditing.
Resolved an issue where webhook delivery for shift.filled events was being delayed by up to 45 seconds when the destination endpoint returned 3xx redirects. Webhook client now follows redirects with a 3-second ceiling.
Rotated all internal service-to-service mTLS certificates as part of quarterly rotation schedule. Upgraded to ECDSA P-384 keys. No customer action required. Completed SOC 2 Type II annual audit — report available upon request.
Optimized the worker eligibility query for large pools (5,000+ workers). Moved from sequential certification checks to a materialized view with 30-second refresh. Eligibility resolution for a 50-worker contact list dropped from 1.2s to 85ms.
Fixed timezone handling for agencies operating across DST boundaries. Shifts scheduled during the 2 AM "spring forward" hour were being assigned incorrect start times in the worker-facing SMS. Now correctly handles ambiguous and skipped local times using IANA tz database.
Released Bulk Shift Import API — POST /v2/shifts/bulk accepts up to 500 shifts per request in CSV or JSON format. Supports dry-run mode for validation. Async processing with webhook callback on completion. Designed for agencies migrating from legacy scheduling systems.
Migrated primary database from RDS PostgreSQL 14 to Aurora PostgreSQL 16.1. Zero-downtime migration using logical replication. Write latency improved 18% due to Aurora's storage layer. Failover time reduced from ~60s to under 10s.
Launched Slack and Microsoft Teams notifications. Coordinators can receive real-time shift fill alerts, daily summary digests, and urgent escalation pings directly in their team channels. Configure per-channel routing rules in Settings → Integrations.
Shipped Worker Preference Learning — the system now tracks and adapts to individual worker preferences over time (preferred shifts, locations, overtime appetite). Contact prioritization uses learned preferences to improve acceptance rates. Workers can review and edit their inferred preferences in the worker portal.
Fixed a bug where the voice agent would occasionally misinterpret "I can't" as "I can" due to audio compression artifacts on low-bandwidth connections. Added a confidence threshold check — if speech-to-text confidence drops below 0.82, the agent now asks for clarification rather than assuming intent.
Implemented API key scoping — customers can now create API keys with granular permissions (read-only, shift management only, worker management only, full access). Existing keys remain full-access. Deprecation of unscoped keys scheduled for Q2 2026.
Voice agent end-to-end latency reduced from 340ms to 180ms by moving to streaming TTS (text-to-speech). The agent now begins speaking while still generating the remainder of the response, resulting in more natural conversation flow. Measured on P50 across 48,000 daily calls.
Released MyHR SDK v1.0 for Python and Node.js. Typed client libraries with full API coverage, automatic retry with exponential backoff, webhook signature verification helpers, and comprehensive documentation. Install via pip install myhr or npm install @myhr/sdk.
Resolved intermittent 504 timeouts on the GET /v2/reports/fill-rate endpoint when querying date ranges exceeding 90 days. Root cause: unindexed join on the shift_assignments table for historical data. Added composite index on (agency_id, filled_at). Query time for 6-month reports dropped from 12s to 340ms.
Added Deputy integration — bidirectional sync for shift schedules and worker availability. Supports Deputy's area and location hierarchy. Auto-detects Deputy webhook format for real-time updates. Third scheduling platform integration after Bullhorn and ADP.
Expanded to US-West-2 (Oregon) region for west coast customers. All new accounts in Pacific/Mountain time zones are automatically routed to the western region. Existing customers can request migration via support. Voice call latency for west coast workers improved ~40ms.
Fixed an edge case where cancelling a shift that was actively being filled could leave orphaned outreach threads — workers would receive a shift offer followed by a cancellation message 30-60 seconds later. Now gracefully aborts in-progress outreach within 2 seconds of cancellation.
Launched Multi-language Voice Support — voice agents now available in English and Spanish, with automatic language detection based on worker preference settings. Bilingual workers receive messages in their preferred language. Portuguese and Haitian Creole planned for Q1 2026.
Completed TCPA compliance audit with external counsel. All outbound communication flows validated against current FCC regulations. Added automated consent verification check — system blocks outreach to any worker without a valid consent record on file. Audit report available to enterprise customers.
Dashboard load time reduced from 3.4s to 1.1s. Migrated from server-side rendering to incremental static regeneration for the analytics panels. Added Redis caching layer for frequently accessed metrics (fill rate, active shifts, worker availability). Cache hit rate: 94%.
Introduced Shift Templates — save and reuse common shift configurations (time, location, required certifications, contact rules). Create templates from the dashboard or via POST /v2/shift-templates. Reduces shift creation time for recurring schedules by ~80%.
Launched ADP Workforce Now connector. Real-time sync for employee records, pay rates, and schedule data. Supports ADP's Enhanced V2 API with automatic token refresh. Certification and compliance fields map directly to MyHR's eligibility engine.
Migrated telephony infrastructure from Twilio to a hybrid Twilio + Telnyx architecture. Outbound call costs reduced 34% with automatic carrier selection based on destination and time of day. Inbound call quality unchanged (MOS score 4.2+). Failover between carriers is automatic with <1s switchover.
Patched a memory leak in the voice agent connection pool that caused degraded call quality after ~72 hours of continuous operation. Root cause: WebSocket connections were not being properly released when workers hung up during the TTS phase. Added connection lifecycle monitoring with automatic recycling at the 4-hour mark.
Shipped the Webhook Events API — subscribe to real-time events for shift lifecycle (created, filled, cancelled, unfilled), worker responses (accepted, declined, no-response), and system events (outreach started, escalation triggered). HMAC-SHA256 signed payloads. Retry with exponential backoff up to 24 hours.