# 💼 Business Cases *Last Updated: 2026-01-09* --- - Data Tinkerer **[How Dropbox Made AI Evaluation Work at Scale](https://www.datatinkerer.io/p/how-dropbox-made-ai-evaluation-work-at-scale)** - Dropbox built a system where every prompt, retriever, and model must prove its value before it can be merged. - Data Tinkerer **[How Netflix Used Deep Learning to Slash Video Quality Control Time by 90%](https://www.datatinkerer.io/p/how-netflix-used-deep-learning-to-slash-video-quality-control-time-by-90)** - Netflix combined neural networks, synthetic pixel generation, and a smarter pipeline to drastically reduce video quality control time. - Data Tinkerer **[How Uber Built an AI Agent That Answers Financial Questions in Slack](https://www.datatinkerer.io/p/how-uber-built-an-ai-agent-that-answers-financial-questions-in-slack)** - Uber’s “Finch” finds the right data, runs the query, and delivers secure, real-time answers directly in Slack. - Data Tinkerer **[How Target Used GenAI to Lift Sales by 9% Across 100K+ Products](https://www.datatinkerer.io/p/how-target-used-genai-to-lift-sales-by-9-across-100k-products)** - Target used LLMs to recommend better add-ons, boosting engagement (+11%) and relevance (+12%) while lifting sales by 9%. - Data Tinkerer **[How DoorDash Used LLMs to Trigger 30% More Relevant Results](https://www.datatinkerer.io/p/how-doordash-used-llms-to-trigger-30-more-relevant-results)** - DoorDash improved relevance by combining knowledge graphs, tight vocabularies, and “just enough” AI. - Data Tinkerer **[How Uber Cut Invoice Handling Time by 70% with GenAI (Without Ditching Humans)](https://www.datatinkerer.io/p/how-uber-cut-invoice-handling-time-by-70-with-genai)** - Uber turned a messy mix of bots and spreadsheets into a scalable pipeline powered by LLMs and human-in-the-loop design. - Data Tinkerer **[How Reddit Scans 1M+ Images a Day to Flag NSFW Content Using Deep Learning](https://www.datatinkerer.io/p/how-reddit-scans-1m-images-a-day-to-flag-nsfw-content-using-deep-learning)** - Reddit uses deep learning to scan over one million images per day to detect and flag NSFW content. - Harvard Business Review **[통합된 전략을 실행하려면 리더들이 서로를 섀도잉해야 합니다.](https://hbr.org/2026/01/to-execute-a-unified-strategy-leaders-need-to-shadow-each-other)** - 체계적이고 확장 가능한 직원 섀도우잉 프로세스를 구현하기 위한 6단계입니다. - Harvard Business Review **[전략적 예측 능력이 뛰어난 기업이 다른 방식으로 수행하는 것](https://hbr.org/2026/01/what-companies-that-excel-at-strategic-foresight-do-differently)** - 500개 조직을 대상으로 한 설문조사를 바탕으로 한 통찰력. - The Sequence **[시퀀스 의견 #786: 엄청난 흡수: 시스템 코드가 모델 가중치가 될 때](https://thesequence.substack.com/p/the-sequence-opinion-786-the-great)** - 많은 시스템 및 에이전트 기능이 프론티어 모델로 제품화되고 있습니다. - Harvard Business Review **[Bristol Myers Squibb이 미래 자금을 조달하기 위해 운전 자본을 혁신한 방법 - EY-PARTHENON의 후원 콘텐츠](https://hbr.org/sponsored/2026/01/how-bristol-myers-squibb-transformed-working-capital-to-fund-its-future)** - EY-Parthenon의 콘텐츠를 후원합니다. - The Sequence **[금주의 시퀀스 AI #785: 경사 고속도로 유지 관리: DeepSeek의 최신 혁신 내부](https://thesequence.substack.com/p/the-sequence-ai-of-the-week-785-gradient)** - DeepSeek mHC는 AI 규모에 대해 확립된 가정 중 일부를 재구성합니다. - The Sequence **[시퀀스 지식 #784: 합성 데이터와 세계 모델의 융합 모델이 구현된 AI의 잠금을 해제합니다](https://thesequence.substack.com/p/the-sequence-knowledge-784-the-convergence)** - 대화형 3D 환경에 대한 합성 데이터 생성 관련성. --- *Generated by [LLM Handbook News Digest](https://github.com/springCoolers/llm-handbook)*