ShiftPretty uses AI to support approved administrative and care-support tasks. The AI may help organize information, draft summaries, identify items for attention, support scheduling or assist users in navigating workflows. It does not provide diagnosis, treatment, emergency response or autonomous consequential decisions.

Clear disclosure

We identify AI experiences at or before the first interaction. A footer notice alone is not sufficient. The interface will state the feature’s purpose, allowed data, important limitations and how to reach a person.

Standard interaction notice: “You are interacting with ShiftPretty AI. It may generate incomplete or incorrect information. Qualified staff must review information before it is used for care, safety, staffing or another consequential decision. This feature does not provide medical advice or emergency services. If this is an emergency, contact local emergency services.”

Human oversight

AI may assist; authorized people decide. ShiftPretty does not permit an AI output to be the sole basis for care escalation, shift eligibility, employment-like opportunity, access to benefits, discipline or another consequential decision. Customers must assign qualified reviewers and maintain escalation procedures.

Data boundaries

Public and Phase 1 AI experiences must not receive PHI. Approved Phase 2 AI may process PHI only through documented data flows under the customer BAA and applicable subprocessor BAAs. PHI is not used to train general-purpose models. Texas Medicaid/EVV data remains prohibited unless separately authorized.

Before each feature launches, ShiftPretty will document the model/provider, input and output types, retention, data location, training use, safeguards, known limitations, human reviewer, monitoring and incident owner.

Safety and dignity

ShiftPretty prohibits AI designed to deceive users about its identity, manipulate people into harm, unlawfully discriminate or impair protected rights. We pay particular attention to foreseeable risks affecting older adults, people with disabilities, limited-English users and careworkers.

Senior/patient-facing AI must provide a human handoff and defined escalation for emergency, self-harm, abuse/neglect, medication and urgent safety content. It must not imitate a licensed professional or discourage contact with caregivers, clinicians or emergency services.

Careworkers and fairness

If AI recommends careworkers or shifts, authorized staff must review the recommendation. Users must be able to correct relevant information and request human review. ShiftPretty will test for inappropriate proxies and disparate outcomes and will not claim that an AI system is “unbiased” or error-free.

Questions and feedback

Use /contact?topic=responsible-ai or email privacy@shiftpretty.com. Report a security issue to security@shiftpretty.com. Do not include PHI in public email or forms.

—