How Product & Data Science collaborated to halve classification costs
summary
A product–data team uses ChatGPT via Azure OpenAI, a local encoder and a Gen X evaluation tool to build a prompt‑engineered production pipeline that ranks labels from a 7,800‑item taxonomy and compares results to historic human tags. The work runs 17 rounds over six months, tunes precision then recall, and reaches >85% acceptance, replacing contractor classifiers and cutting the manual classification budget by half — classification engine saving $0.5M/year — now classifying 90% of papers automatically.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 1:07Elsevier and SSRN overview
- 4:57Gen AI integration
- 6:45SSRN classification challenge
- 10:35Current classification issues
- 13:23Precision and recall
- 14:49Data science collaboration
- 16:56AI proof of concept
- 20:54AI classification challenges
- 23:24Lessons from AI project
- 25:52Cost and efficiency impact
- 27:38AI in paper approval
- 29:25Other Elsevier AI initiatives
- 31:29Q&A
related talks
29mTransformation in a New Era of AI
01h 12mChatGPT and OpenAI API solutions: 👑 My Favourite Implementations
01h 09mChatGPT and OpenAI API solutions: 👑 My Favourite Implementations
36mUnleashing Generative AI: Improving Developer Productivity in SRE
36mAI for Government Efficiency with ChatGPT: From Hype to Reality
32mSmarter AI for Smaller Teams: Scaling Automation Without Breaking the Bank
07mHow About Now? Can ChatGPT Do My Job Yet?
01h 16mChatGPT - and How we Can Benefit from it