Skip to content

Trust belongs in
every interaction.

Trust in healthcare AI depends on controlling what information the AI can use, how patient data is handled, what actions the system can take, and when a person needs to step in.

Security & compliance

Healthcare-grade security.
Across the experience.

  • HIPAA Compliant
  • SOC 2 Type II
  • ISO 27001
  • FDA 21 CFR Part 11
  • GDPR Ready

Inside the E1™ architecture

Separate identifiers.
Preserve accountability.

The platform architecture separates patient identifiers from AI processing and records access for review.

  • Approved knowledge sources
  • Tokenization / de-identification
  • Separate PHI Vault
  • Role-based access
  • Audit logging
  • Consent
  • Human escalation
Before processing

Identify and tokenize

Detect patient identifiers and replace them with tokens. Keep the identifying information in the encrypted PHI Vault, separate from the approved-content repository used for responses.

During the interaction

Use de-identified context

Route de-identified information into the response workflow, with approved knowledge sources and program rules.

Across the workflow

Control and record access

Apply role-based access, authentication, and audit logging so teams can review who accessed information.

Keep every response
within bounds.

The response workflow uses approved content, limits what the agent can do, and checks a reply before it reaches the patient.

Information boundaries

Approved-content repository

E1’s Private Vault holds your organization’s approved content for responses. It is separate from the PHI Vault, which protects patient identifiers. Questions outside the approved content receive a configured fallback.

Permitted actions

Intent analyzer

The intent analyzer checks which workflow actions are permitted. Out-of-scope requests receive an approved fallback or move to human support through the configured escalation. A request for a person follows that same route.

Before delivery

Response validation

A large language model drafts a reply. A small language model checks it against approved information and response rules. Failed drafts are withheld; the configured workflow supplies an approved fallback or routes the request to human support.

Controls, in context.

See each product’s controls. Capabilities vary by product and implementation.

E1™Information, consent, and escalation

E1 applies approved-information boundaries, consent, and human escalation to keep automated responses within the organization’s rules.

  • Responses use approved knowledge sources and program rules.
  • Consent steps form part of the workflow.
  • Failed validation and out-of-scope requests follow the configured fallback or human escalation.
Explore E1™
HXP™ Marketing CloudPatient choice and engagement

HXP brings patient preferences and consent into communication decisions, with compliance controls applied throughout the engagement workflow.

  • Opt-in and opt-out management governs communication choices.
  • Patient preferences inform content, channel, and delivery timing.
  • Compliance controls apply within the engagement workflow.
Explore HXP™ Marketing Cloud
EIDM 360™Context for interpretation

EIDM 360 keeps behavioral inferences distinct from facts and preserves the context people need to review an interpretation.

  • Patient and interaction context remains available alongside analysis.
  • Behavioral inferences are kept distinct from established facts.
  • Dashboards and visualizations support a person’s review.
Explore EIDM 360™
VOCE™Conversations and continuity

VOCE combines secure audio and video with interaction records that support traceability and review after a conversation.

  • Audio and video engagement uses a secure conversation surface.
  • Transcripts and summaries let teams review what was discussed.
  • Audit trails retain a record of the interaction.
Explore VOCE™
PIXI™Safety review and reporting support

PIXI retains source context and a human review gate as potential safety information moves toward applicable reporting workflows.

  • Potential adverse event information retains its source context.
  • A person reviews the information before the next reporting step.
  • Reporting support follows the applicable workflow, including FDA MedWatch for relevant US workflows.
Explore PIXI™

Good questions are the start of a trusted relationship.

Talk with HXT