ERA is a agentic AI pipeline that takes a researcher's study description and generates a complete, REC-ready ethics document package — consent forms, participant information sheets, ethics narratives, and regulatory compliance reports — in minutes instead of months.

The pipeline works in four stages:

  1. Intake Agent — Parses a free-text study description and extracts a structured study profile: population, methods, data types, funding, institutional context.
    1. Regulatory Mapping Agent — Uses deterministic rule-based logic combined with Claude's reasoning to identify every applicable UK regulation (HRA/IRAS, GDPR, Mental Capacity Act, incidental findings protocols, Caldicott Principles, UKRI data sharing mandates). It flags requirements the researcher likely hasn't considered.
    2. Document Generation Agents — Produce a Participant Information Sheet, Consent Form, and Ethics Narrative tailored to the specific study. Consent forms are scored for readability (targeting 8th-grade level). Vulnerable population safeguards (capacity assessment protocols, consultee consent) are automatically included when detected.
    3. QA Agent — Reviews the entire document package for cross-document contradictions, missing regulatory coverage, common REC rejection patterns, and readability issues. Outputs a submission-readiness score with specific fix suggestions.

We built ethics infrastructure, not just a text generator. ERA doesn't bolt ethics considerations on as an afterthought — the entire product is an ethics tool. Every feature is a safeguard: mandatory review checkpoints, deterministic regulatory rules, cross-document consistency checking, readability scoring on participant-facing materials.

Tackling the reproducibility crisis. By forcing structured, precise study documentation from day one — explicit data retention periods, anonymization procedures, pre-registration triggers, and consistent protocol-to-consent mapping; ERA doesn't just speed up ethics approval. It produces the exact kind of thorough documentation that the research reproducibility crisis has been demanding. The most complete version is also the fastest version.

The patient access framing is real, not performative. 150,000 clinical studies registered annually. If ERA saves 2 months per study on average, that's 300,000 patient-months of delayed access to potential treatments — every year. For progressive conditions like Alzheimer's, those months represent irreversible cognitive decline.

Built With

Share this project:

Updates