◆ Assistive Specialist
Reads occupancy sensors, badge data and booking patterns to identify under-used floors and uneconomic leases, and models consolidation options across portfolio scenarios. Advisory only and aggregated (never individual surveillance); the real-estate decision agent owns every lease and consolidation call under the oversight judge.
Memory
Working The site or portfolio under analysis.
Episodic Prior utilization trends and past space decisions.
Semantic Lease terms, space standards, cost-per-seat models.
Store Occupancy time series + portfolio data warehouse
Orchestration
MCP
Harness · Managed Agents … sandboxed code execution against occupancy data.
Tools
{ } Occupancy sensors + badge data API { } Room-booking system API { } Lease + portfolio database API ›_ Optimization modelling sandbox Code exec
Evals & guardrails
- Recommendations are advisory … the real-estate decision agent owns every lease/consolidation decision under the oversight judge.
- Source-traceability: every utilization figure links to its sensor/booking data.
- Privacy guardrail … occupancy analysis is aggregated, never individual surveillance.
Frontier edge
- ▲World-model simulation: models a consolidation before the decision agent commits, projecting cost, seat-pressure and break-clause timing across portfolio scenarios.
- ▲Causal reasoning: separates real structural under-use from seasonal or hybrid-pattern noise, so it doesn't recommend giving up space the bank will need back.
- ▲Confidential compute: badge and occupancy signals are aggregated inside a secure enclave, so the analysis can never resolve to individual movement.
In numbers
100%
Portfolio under continuous analysis
tracked / quarter
Identified underutilized space
Handoffs
Hands to → Facilities Operations Agent
Across ⇢ Finance for real-estate cost-planning inputs