How it works

One cycle. Every stage connected.

AT.OS is opinionated by default and configurable where law differs. It ships with working playbooks for intake, referral, investigation and reporting, aligned with international labor standards. Each country then adapts them to its own laws, institutions and languages.

Detect

Tips from every channel arrive in one queue. Satellite imagery flags potential exploitation sites for field verification. Supply-chain signals connect discoveries to the economic activity behind them.

Protect

Survivors and at-risk workers reach help by voice and text on WhatsApp. Caseworkers manage intake, referrals, services and follow-up in one case file. That file connects to national social protection networks: healthcare, housing, income support, legal aid and training.

Investigate

Tips and case records become investigation priorities and case clusters. Each institution trains its own models, under human oversight.

Prosecute

Prosecutors see patterns across cases and the actors and supply chains linked to them.

Account

National dashboards bring together administrative records and official statistics. Action-plan tracking shows which commitments are on track and where implementation lags.

Prevent

Outreach reaches people where recruitment happens, guided by where risk is concentrated.

Learn

A shared research framework produces evidence that is comparable across places and reproducible over time.

Across every stage: Partner.

A shared data layout lets agencies, civil society and other stakeholders hand off work without re-entry, through authorized connections only.

Supply chains, in a shared language.

AT.OS includes a supply-chain tool with a standardized layout used by three groups:

Civil society organizations

Document risk in sectors and regions.

Governments

Connect discoveries to the actors and products involved.

Companies

Map their own chains and share supplier data in the same format.

Clear boundaries keep investigations and survivor data out of reach of commercial users.

AI-native, by design.

Messy data welcome

Records are taken in as they are, and AI handles extraction and normalization.

Sensitive data stays home

Survivor and case data is processed only inside the adopter's own environment.

Conversational interfaces

Voice for the public, plain-language questions for staff. Less time learning where to click, more time on judgment.

Humans decide

AI suggests; accountable people approve, and every step is recorded.

Adopt all of it, or start with one part.

Every component works on its own. Connections between components are switched on deliberately, never by default. An agency can begin with a single capability and grow from there.