AI recognized by pharmaceutical giants! Novo Nordisk: AI is finally reliable enough to generate sensitive documents

The annual cost is less than one employee's salary! Novo Nordisk has introduced AI into the drafting of drug regulatory documents and has achieved significant results. According to media reports, last year, Danish pharmaceutical giant Novo Nordisk began testing Anthropic's Claude 3.5 Sonnet model. Waheed Jowiya, the strategic director responsible for overseeing AI usage, found that the number of errors had significantly decreased. Currently, Novo Nordisk has started using Claude to draft clinical research reports, each of which can be hundreds of pages long. For years, Novo Nordisk has been testing AI chatbots, including OpenAI's ChatGPT and Meta's Llama, hoping to leverage these technologies to assist in writing drug approval documents submitted to regulatory agencies. However, industries with strict regulations, such as pharmaceuticals and insurance, have been cautious about using generative AI to handle legally sensitive documents. The company's head of technology strategy, Louise Lind Skov, stated that early technologies are prone to errors, and sometimes correcting AI mistakes is more time-consuming than writing from scratch. It is reported that Novo Nordisk has adopted a common method to reduce AI error rates: Retrieval-Augmented Generation (RAG). For example, when Claude generates a clinical definition of obesity that a human expert considers good, the human will instruct Claude to reuse that description in any future documents regarding obesity trials. Skov revealed that this practice has greatly shortened the document drafting time, from about 15 weeks to less than 10 minutes. Previously, such documents required the involvement of over 50 writers, but now only 3 human writers can complete them with the help of Claude. Moreover, Novo Nordisk's annual expenditure on Claude is less than the salary of one writer. The company stated that it has not laid off any writing staff but plans to reduce new hiring and allocate the saved resources to recruitment in other departments. Skov commented on this: “This increase in efficiency means we can accomplish the same or even more work with fewer people.” “We are rapidly moving AI from the experimental phase to the development phase, which is a rare quick advancement in the pharmaceutical industry.”