THB Group

Their value prop is 'leveraging deidentified data repository to generate scientific evidence' and 'AI-based medical recommendations' — which sounds cutting-edge until you realize GPT-4 already outperforms most clinical decision support systems on medical licensing exams, and it learned from PubMed, not from THB's '1bn+ clinical parameters.'

Medical writing, biostatistics, protocol design, and clinical operations — literally every service they list under 'end-to-end evidence generation' — are exactly what Claude and specialized medical AI are automating first. Their PhD biostatisticians are competing with models that process clinical trials faster than humans read abstracts.

The 60-million-patient longitudinal dataset and 5,000+ clinical partnerships are real assets — that's a decade of relationship-building and regulatory navigation you can't replicate with a API call. The moat is in the data access and clinical trust, not the analytics on top of it.

They're positioning as 'AI-based' everything while their website reads like it was written in 2019 — lots of 'big data engines running continuously' but zero mentions of LLMs, foundation models, or any AI architecture that wasn't already standard practice three years ago.

Regulatory-compliant clinical evidence generation survives — pharma companies still need human-signed studies for drug approvals. But 'hyper-personalized patient care recommendations' and 'targeted HCP engagement' are exactly the workflows that medical AI eliminates, not enhances.