Toward an Autonomous Business Operating System for the U.S. Private Security Industry: A Multi-Agent AI Framework for End-to-End Workflow Automation

Operational problem

U.S. private security providers coordinate scheduling, emergency coverage, timekeeping, reporting, compliance, payroll, billing, and client communication across multiple systems. Staffing continuity remains difficult: the U.S. Bureau of Labor Statistics reports approximately 1.27 million workers in the combined occupation in 2024, with about 162,300 projected openings annually during 2024–2034. A Census QWI-based analysis cited by the Center for American Progress reported 50.8% annual turnover in the security industry in 2023.

Existing workforce platforms already integrate substantial portions of these functions. ASBOS is proposed as an orchestration layer that coordinates cross-system workflows with explicit human approval, auditable events, and scoped agent permissions — not as a claim that integrated software does not exist.

Research question

Can a security-specific, event-driven multi-agent layer coordinate workforce functions autonomously within policy bounds, preserve auditability, escalate risk appropriately, and demonstrate better outcomes in a controlled field study?

Six-agent architecture

All six agents operate through a central orchestration and policy layer. Agents receive scoped permissions; deterministic rules control hard constraints; important actions can require supervisor approval; authoritative systems remain the systems of record; and events and actions are logged for traceability.

Architecture visualization

Event-driven orchestration

ASBOS is modeled as an orchestration layer above existing systems of record. Each business event receives a unique identifier, timestamp, source, confidence level, data classification, and audit history. The orchestration service receives events such as shift creation, missed clock-ins, credential expiry, overdue reports, and invoice due dates — evaluates policy, invokes specialized agents, records tool calls and decisions, and routes exceptions to a human queue.

Human oversight

ASBOS uses three action tiers. Tier 1 actions are low-risk and reversible, such as drafting a schedule or sending an approved reminder. Tier 2 actions may execute automatically but notify a supervisor. Tier 3 actions require approval — including unbudgeted premiums, payroll changes, credential exceptions, and high-severity incident communications.

Privacy and security controls

Voice deployment requires jurisdiction-specific review of call authorization, recording consent, AI disclosure, data retention, and opt-out handling. The architecture constrains voice outreach to approved pay bands, explicit disclosures, short retention where possible, and immediate human handoff for sensitive cases. Biometric voice templates should not be created unless legally approved and operationally necessary.

• Role-based access control and tenant isolation
• Typed event validation with source provenance
• Complete audit logging and policy versioning
• Reversible writes to systems of record
• Data minimization for location, voice, and biometric data

Evidence and context

Public context only. Market figures are third-party forecasts, not observed outcomes. These organizations do not endorse ASBOS.

Prospective evaluation plan

A suggested 90-day pilot includes baseline weeks, staged deployment phases, and a final observation period. The outcomes below are proposed pilot measurements, not completed results. The repository simulator provides synthetic baseline comparisons only to support protocol design.

• Administrative workload
• Coverage time (median and 90th percentile)
• Percentage of shifts filled before start
• Timekeeping exceptions
• Report delivery compliance
• Credential-related assignment blocks
• Human escalations and overrides
• System availability
• Duplicate-action rate
• Failed integrations
• Recovery time

View synthetic simulator scaffold

Limitations and future work

ASBOS is a conceptual and design contribution. It does not report a completed deployment, causal effect, measured fill time, measured administrative savings, or validated ROI. Market-size figures differ by publisher and category definition. Future work includes a registered pilot protocol, de-identified event dataset, failure-mode analysis, fairness testing of replacement ranking, and independent security and privacy review.

Author

Rabbi, K M Fazle (2026). Toward an Autonomous Business Operating System for the U.S. Private Security Industry: A Multi-Agent AI Framework for End-to-End Workflow Automation. Manuscript. St. Cloud State University.

@article{rabbi2026asbos,
  title={Toward an Autonomous Business Operating System for the U.S. Private Security Industry: A Multi-Agent AI Framework for End-to-End Workflow Automation},
  author={Rabbi, K M Fazle},
  year={2026},
  institution={St. Cloud State University},
  note={Manuscript}
}