AI agents that automate the coordination work in clinical trials — so nothing falls through the cracks between systems.
Medidata, Veeva, and IQVIA have improved their platforms. But the work that lives in the seams between systems — chasing documents from sites, reconciling data, tracking follow-ups — remains a manual job.
CDA executed to site going live takes weeks of document collection, regulatory checks, and back-and-forth with coordinators. Delays here cascade into enrollment delays.
$500K+/day in enrollment delay costs (Smith, DiMasi & Getz, Tufts CSDD, 2024)eTMF and EDC are often populated after the fact for regulatory purposes — not maintained in real time. The actual source of truth during a trial lives in email, Slack, spreadsheets, and people's heads.
50+ vendors and a dozen platforms per Phase III trialWhen sites don't respond, the escalation ladder runs from coordinator to field monitor to country monitor to study lead. Tracking where each site stands is a full-time job.
70% of clinical trials experience delaysMedian tenure in clinical operations is under 3 years. Every departure takes knowledge — which sites are slow, which sponsors have unusual requirements — with it.
~30% annual CRA turnover (BDO CRO Compensation & Turnover Survey)Each agent handles a specific operational workflow — fitting into existing processes, not replacing them. Humans supervise; agents execute.
Ingests documents via email or upload, classifies against the TMF reference model, runs ALCOA++ quality checks, and flags issues for review. Handles CVs, licenses, training certificates, regulatory forms.
Deterministic rules, not generative models, for compliance logicTracks the end-to-end path from CDA execution to site go-live. Monitors milestone completion, identifies blockers, and coordinates across sponsor, CRO, and site teams.
Measurable KPI: CDA-to-activated timeManages escalation cadence with intelligent batching — outstanding items grouped per contact, reminders paused when awaiting response, site responsiveness profiles adjusting timing.
Configurable escalation ladder per sponsor SOPTracks site visit schedules, enrollment progress, and data review status. Surfaces anomalies and deviations before they become inspection findings.
Institutional knowledge persists as data, not tribal memoryNytel sits above your existing clinical systems as an orchestration layer — it reads state from CTMS, eTMF, and EDC and writes actions back, but does not replace them.
Quality checks use code and rules, not generative models. Every decision is auditable and reproducible. LLMs handle document understanding, not compliance logic.
OCR confidence below 0.80 triggers human review. Safety exceptions skip the escalation ladder. Study teams can pause, override, and flag sites as quiet.
Email, Slack, spreadsheets — Nytel meets trial teams where they actually work. No API credentials, no IT procurement, no changes to partner systems required to start.
Universal rules handle the majority of checks. Sponsor overrides — stricter thresholds, custom templates, additional fields — configure on top without touching the rule engine.
Associate Professor at George Mason University in Systems Engineering and Operations Research. Previously a Principal Computational Scientist at Argonne National Laboratory and Visiting Professor at the University of Chicago Booth School of Business.
Research in Bayesian statistics, machine learning, and computational modeling of complex systems. Built POLARIS, a large-scale simulation framework used by the Department of Energy and the US Department of Transportation. Recent work applies Bayesian methods to clinical trial design — including Bayesian perspectives on clinical trials and anytime-valid monitoring for adaptive trials. Author of Clinical Trials in Practice — a guide to clinical trial methodology from foundations through Phase III operations.
Read the full technical whitepaper on how AI agents can automate clinical trial operations — from system architecture to follow-up cadence to quality frameworks.