Novartis has worked with Market Logic Software since 2020 on Sherlock, an end-to-end platform for commercial insights and research management, and in July 2024 added Market Logic's generative AI tool DeepSights on top of it. The collaboration was a finalist for "Best Data Solution" at the 2024 Market Research Society Awards.
Before Sherlock, Novartis's insights content was spread across separate SharePoint sites, causing duplication, governance and compliance problems and extra effort to get knowledge to the right stakeholders. Sherlock consolidated those sites into a single sign-on platform, introduced governance developed with legal, procurement and other functions, added curated "Knowledge Zones" microsites for brands and therapeutic areas (more than 25 launched in one year), and let external suppliers upload research directly. By 2023 the user base had grown 640% since launch, and Sherlock has passed 7,500 unique users.
DeepSights uses retrieval-augmented generation to pull relevant insights from the Sherlock archive. Before answers are displayed it runs a self-validation loop, it provides full citations and document references, it flags "Watch Outs" when sources contain inconsistencies or contextual limitations, and it applies user controls over what is uploaded. Novartis worked with its legal and risk teams on the reliability of generated insights and on processes to prevent their misuse, and its business operations team built a business case and value proposition before approaching procurement.
"With the addition of DeepSights' generative AI, users are able to get answers faster and more reliably than ever, saving an average of 2 hours per question and millions through reduced duplication of spend," said Ian Hook, Director of Business Operations at Novartis.
Novartis reports a 56% reduction in primary market research spend, a $30 million annual reduction in research spend and two hours saved per major research query. In one example, a team wanting to know whether patients preferred blister or bottle packaging for a drug commissioned a three-month survey costing $50,000 to $100,000 with 50 respondents; another team used DeepSights to draw on existing research covering 6,700 patients in three weeks at no external cost.
Adoption was driven by a test-and-learn change programme. Comparing two behavioural-science approaches, the team found the "authority principle", emails from senior leaders endorsing the platform, drove usage spikes far more than highlighting past pain points. Novartis now requires contracted research vendors to upload their deliverables into Sherlock.