Narrativa

Market position

Narrativa positions itself as an agentic AI documentation provider for regulated life sciences workflows, and states it was the first AI vendor to fully automate patient safety narratives for an FDA-submitted study and the first AI vendor represented on the CDISC Board.

The company cites end-to-end regulatory expertise across the clinical workflow from data collection to submission, agent-based automation that produces submission-ready outputs with traceability, a proprietary knowledge graph and text-generation language, recognition by CDISC, and compliance-oriented architecture for GxP, GDPR, HIPAA and EU AI Act requirements.

Technology

The platform combines AI agents with a knowledge graph and a pseudo-programming language for text generation developed by the founding team. Agent-based architectures were adopted in 2025 to address limitations of standalone large language models, including hallucinations, and are used to generate submission-ready R code for TLF (tables, listings and figures) production along with regulatory prose, and to work with large, complex documents while maintaining traceability. The platform includes features for security and privacy, compliance and infrastructure, with GDPR alignment, HIPAA-ready and GxP validation-ready architecture, and a risk-based AI safety framework referencing FDA/EMA guidance and EU AI Act principles. A customer-facing assistant, SideKick AI Agent, is available on the company website.

Go-to-market

Direct enterprise sales to pharma and life sciences organizations supported by published customer success stories, partnerships (VMD Sciences in veterinary regulatory documentation; UCLA Bruins for Business Health for AI training and internships), participation in industry standards work with CDISC and the ClinEco community, and startup programs run by Microsoft, Google, Tata Consultancy Services and NYU's Endless Frontier Labs.

Pharmaceutical and biotech companies, including large global pharma, plus other life sciences organizations; the company also cites opportunities in medical devices, hospitals, healthcare providers and research organizations, and previously served marketing and media customers.

Geography

The company serves pharmaceutical customers across multiple markets and has worked on projects in Spain, including election coverage for RTVE across nearly 5,000 Spanish municipalities, and in the United States, including an FDA-submitted study, a CDISC board seat and a program with UCLA.

History

The company was founded in 2015, months before OpenAI existed and roughly two years before the transformer architecture paper. According to CEO David Llorente's ten-year retrospective, the early company made mistakes by shaping itself to appeal to investors, but eventually signed its first large customers in the marketing and media industries and reached break-even at a reasonable scale. About six years before the December 2025 retrospective, the company pivoted to focus primarily on life sciences and pharmaceuticals, targeting the share of clinical trial time—stated as up to 40%—spent on analysis, reporting and regulatory approval. Progress was constrained by regulatory complexity and by limitations of large language models such as hallucinations requiring human oversight. In 2025 the company shifted to agent-based architectures, enabling AI agents that generate submission-ready R code for TLF production together with associated regulatory prose, and secured its first enterprise multi-year contracts with large pharmaceutical companies. Other milestones include selection for Microsoft for Startups (2020), recognition by Samsung's first Synthetic Media Landscape study (2020), the Google for Startups program (May 2023), NYU's Endless Frontier Labs mentorship program (October 2023), full automation of patient safety narratives for an FDA-submitted study (2023), and Jennifer Bittinger joining the CDISC Board in 2026.

Risks & controversies

The CEO has publicly described early strategic mistakes, notably shaping the company to appeal to investors rather than customers, and years of difficulty understanding the pharmaceutical industry and its regulatory processes after the life sciences pivot. He also notes that large language models alone were limited by hallucinations requiring human oversight, which constrained scalability and efficiency before the shift to agent-based systems.