AI that supports people, governed like it matters
In human services, AI touches people at their most vulnerable moments. Here is exactly how we use it, how we govern it, and how we protect the data behind it.
The AI handles the paperwork. Your team handles the people.
We use AI to reduce documentation burden, surface risk earlier, and automate compliance reporting. We do not use it to replace professional judgment. Every AI output is reviewed by a person before it becomes part of a client record or a decision.
What our AI does
- Drafts case notes from caseworker input for review and approval
- Flags possible risk factors at intake and over time
- Surfaces high-need cases for supervisor attention
- Assembles funder and compliance reports from case data
- Supports on-demand training and technical assistance
What our AI never does
- Decide eligibility, placement, or service levels
- Deny, reduce, or end services
- Contact clients without staff direction
- Finalize a record without human approval
- Train public or third-party models on client data
Six commitments that govern every deployment
Human in the loop
AI drafts, flags, and summarizes. A caseworker or supervisor reviews, edits, and approves every output before it is used. AI never makes eligibility, placement, or service decisions.
Privacy & data minimization
We collect only what the program needs, apply role-based access, and follow HMIS, HIPAA-aligned, VAWA, and VOCA confidentiality requirements, including survivor safety protections.
Fairness & bias monitoring
Risk flags and screening tools are reviewed for disparate impact across populations, and staff can override any flag with a documented reason.
Transparency
Agencies can see what the AI suggested, what staff decided, and why. Clients are told when AI supports their case documentation.
Trauma-informed design
Screening and intake flows are designed to avoid re-traumatizing questions and to respect survivor and veteran safety.
Accountability
Every AI output and human decision is logged for audit, quality improvement, and funder reporting.
How we protect client data
Human services records include some of the most sensitive information a person can share. Our security practices are built around that reality.
Specific controls are documented per engagement to meet each agency’s contract, funder, and state requirements.
Encryption
Data encrypted in transit and at rest.
Access control
Role-based, least-privilege access with multi-factor authentication.
Audit logging
AI outputs, edits, and approvals are recorded with user and time stamps.
No model training on client data
Client data is not used to train public or third-party AI models.
U.S. data residency
Client data is hosted in U.S. data centers.
Incident response
Documented incident response and breach notification aligned to contract and legal requirements.
Vendor due diligence
We prioritize cloud and AI providers with independent security attestations, such as SOC 2 or FedRAMP authorization, where the program requires them.
Retention & disposal
Records are kept and disposed of according to funder, contract, and state requirements.
Built to align with the standards you answer to
Our governance practices are designed to align with the frameworks federal, state, and local programs already use, so AI adoption fits inside your existing compliance obligations.
- NIST AI Risk Management Framework (AI RMF 1.0): Govern, Map, Measure, Manage
- NIST SP 800-53 security and privacy controls, where required by contract
- HIPAA Security Rule safeguards (HIPAA-aligned handling of health information)
- HUD HMIS Data Standards and Coordinated Entry data requirements
- 42 CFR Part 2 confidentiality, where substance use records are involved
- VAWA and VOCA confidentiality and survivor safety requirements
- Section 508 and WCAG 2.1 AA accessibility
Documentation available on request
AI use inventory
Each AI capability, its purpose, inputs, outputs, and the human review step that governs it.
Data flow & security
Data flow diagrams, hosting details, and responses to your security questionnaire.
Accessibility
Our accessibility approach and testing results. See our accessibility statement.
Questions from your compliance or security team?
We’ll walk your privacy, security, and program reviewers through how the AI is governed, what it can and cannot do, and how every output is reviewed.
Request a governance briefing