Instructions for Use
How Ryght's AI-powered site selection and feasibility tools work, their intended use, built-in human oversight, known limitations, and security controls.
Network Navigator & Feasibility Accelerator
Version: 1.0 | Effective: February 1, 2026 | Review Cycle: Annual
What Are These Products?
Network Navigator is an AI-powered site selection tool that analyzes clinical trial protocols and recommends the best-fit research sites from a global database of 100,000+ AI Site Twins — digital profiles of clinical research sites worldwide. Users upload a protocol, receive a ranked list of recommended sites, and can explore individual sites through conversational AI.
Feasibility Accelerator streamlines the feasibility questionnaire process. It converts PDF questionnaires into digital web forms, pre-populates site-specific answers using AI Site Twin data, and distributes them to sites for review and confirmation. A real-time dashboard tracks response progress.
Both products are built on AI Site Twins — structured digital representations of clinical research sites that capture investigator expertise, therapeutic experience, facility capabilities, patient populations, historical performance, competing trials, and regulatory history.
Who Is This For?
Clinical operations professionals, feasibility managers, site selection specialists, and project managers at CROs, pharmaceutical sponsors, and biotech companies. Users should have professional experience in clinical trial operations. All users must be 18 or older.
Intended Use
These products are AI-assisted decision-support tools for clinical trial site selection and feasibility assessment. All AI-generated outputs — site rankings, pre-populated questionnaire data, and conversational insights — are advisory only and require human review before any operational action is taken.
These products support the operational planning phase of clinical trials (site identification, qualification, and feasibility). They do not interact with patients, process individual patient data, or make clinical decisions.
How the AI Works
The platform combines large language models fine-tuned for life sciences, knowledge graph technology, and agentic AI workflows. Key AI functions include:
- Protocol analysis: Extracts therapeutic area, indication, phase, eligibility criteria, and site requirements from uploaded protocol documents.
- Site matching and ranking: Scores sites against protocol requirements using multiple factors (therapeutic experience, investigator expertise, patient access, facilities, competing trials, regulatory history).
- Conversational site exploration: Enables natural-language queries against individual AI Site Twins to investigate site-specific strengths and risks.
- Questionnaire conversion: Analyzes PDF questionnaires and generates interactive web forms.
- Data pre-population: Fills feasibility questionnaires with site-specific data drawn from AI Site Twins.
- Feedback loop: Site corrections and confirmations update the corresponding AI Site Twins, improving accuracy over time.
Human Oversight Built In
Human-in-the-loop review is built into the core product workflows:
| Stage | Human Review |
|---|---|
| Protocol analysis | User confirms extracted parameters before site matching |
| Site recommendations | User evaluates ranked list and applies professional judgment |
| Conversational AI answers | User critically assesses responses and cross-references key findings |
| Questionnaire conversion | User reviews generated web form before distribution |
| Pre-populated data | User reviews auto-filled responses before issuing to sites |
| Site confirmation | Site personnel independently verify, correct, and confirm all data |
| Final site selection | Human decision-makers approve all site activation decisions |
Users may override, modify, or discard any AI-generated output at any point. The AI does not make or execute decisions autonomously.
Known Limitations
- AI outputs may be inaccurate. Like all LLM-based systems, outputs may occasionally be incomplete, inconsistent, or incorrect. Human review is required at every decision point.
- Data dependent. Output quality depends on the completeness of underlying data. Newer or smaller sites may have less complete AI Site Twins.
- Temporal lag. Site Twins may not reflect very recent changes (e.g., investigator departures, new competing trials) until the next data refresh or site confirmation.
- Protocol interpretation. Complex or ambiguous protocol language may be interpreted differently by the AI than by human readers. Always verify extracted parameters.
- Language and geography. Models are primarily trained on English-language data with strongest coverage for North American and European sites. Performance may vary in other regions.
- Requires internet connectivity. The platform is cloud-hosted and does not support offline use.
What This Product Is NOT
- Not medical advice. Does not diagnose, treat, or prevent disease. Does not create a physician-patient relationship.
- Not a replacement for professional judgment. AI recommendations augment — never replace — the expertise of clinical research professionals.
- Not autonomous. Must not be used in workflows that bypass human review of AI outputs.
- Not a substitute for regulatory due diligence. Does not replace qualification visits, regulatory inspections, or compliance assessments.
- Not for patient-level decisions. Must not be used for individual patient enrollment, treatment, or care decisions.
Security and Compliance
The Ryght platform is SOC 2 Type 2 certified (audited November 2024 – November 2025; unqualified opinion, no exceptions). Key controls include:
- Multi-factor authentication required for all users
- Data encrypted at rest and in transit (TLS/SSL)
- Role-based access controls with quarterly reviews
- Network segmentation, firewalls, and intrusion detection
- Quarterly vulnerability scanning and annual penetration testing
- Formal incident response plan with annual tabletop exercises
- GDPR and UK data protection compliance
- Hosted on Microsoft Azure with documented business continuity and disaster recovery plans
Platform status: status.ryght.ai
Support and Reporting
- Technical support: ryght.ai/support
- AI output concerns: Report inaccurate, biased, or misleading outputs through the support portal with details of the input and output in question.
- Security incidents: Report immediately through the support portal.
- General inquiries: ryght.ai/en/contact-ryght
This document is based on publicly available product information and the Ryght SOC 2 Type 2 report. For full terms, see Terms of Service and Privacy Policy.