Over the past year, we have worked on several AI systems where the instinct was clear: “make it agentic.” Add autonomy. Add tool use. Add memory. Let the model decide.
In multiple cases, that instinct made the system worse.
This talk examines real-world scenarios where agent-based designs introduced instability, evaluation collapse, unpredictable costs, boundary violations, and long-term maintenance complexity. We’ll explore when deterministic pipelines outperformed autonomous agents, how evaluation degrades as system freedom increases, and why smaller, composable components often prove more reliable than large, generalized agents.
Rather than arguing against agents, this session focuses on architectural trade-offs: autonomy vs. control, flexibility vs. reproducibility, and experimentation vs. operational reliability. We’ll also share open questions we’re still grappling with around observability, safety boundaries, and maintainability in agent-driven systems.
Our goal is to go beyond just presenting a finished framework, but to invite technical dialogue on when autonomy truly adds value - and when it doesn’t!
Jigyasa Grover is a 12-time award-winning AI lead and 'Sculpting Data For ML' author Jigyasa Grover drives rider personalization innovation at Uber after transforming Twitter/X, Facebook/Meta, Faire, and Bordo AI with large-scale ML systems. Handpicked by Google for their I/O 2024 keynote, she serves on Google's Developer Advisory Board while advising social search engine Diem and other Silicon Valley startups.
As a LinkedIn Learning instructor, Jigyasa educates thousands of professionals worldwide on cutting-edge AI-powered applications and agentic AI systems, solidifying her status as a thought leader in artificial intelligence education. As a Google Developer Expert, Women Techmaker Ambassador, and World Economic Forum Global Shaper, Jigyasa has also been featured in Forbes, Business Insider, VentureBeat, and International Business Times, and has elevated panels with Harvard University, Preston-Werner Ventures, Norwegian Business School, Humanitarian Frontier in AI, Women in Data, and more to her name.
The UC San Diego alumna has secured funding from the Canadian and Norwegian governments, the Linux Foundation, and multiple tech giants, enabling work that transcends geographical boundaries. With 200+ media features and contributions to open source recognized by Apache and Python Software Foundations, she mentors next-generation talent while shaping AI's future through advisory roles at Bezoku AI, Las Positas College, and various AI forums.
Rishabh Misra is a Principal ML Engineer and Researcher with over a decade of experience in the AI and machine learning space. He currently drives LLM training and generative personalization efforts at Atlassian, and has previously led deep learning and GenAI-powered user personalization initiatives at a late-stage conversational commerce startup (Attentive), as well as at Twitter and Amazon. He specializes in designing low-latency, large-scale deep learning systems and successfully deploying them to production.
He has an extensive publication record in NLP (Large Language Models / GenAI), deep learning, and applied machine learning, with over 1,000 citations. He actively contributes to the research community as a committee member for leading AI conferences such as ICML, KDD, and TheWebConf. In recognition of his contributions to machine learning research, he has been acknowledged by the U.S. government as an outstanding researcher.
He is a strong advocate for a data-centric approach to AI and has co-authored the book Sculpting Data for ML, which serves as a practical guide to curating high-quality datasets as the foundation of robust ML pipelines. Building on this philosophy, he regularly shares insights and best practices through technical talks, panels, podcasts, and blog posts. His machine learning content has reached over 150,000 people globally.
His work has been widely featured in media outlets including TechCrunch, Times of India, The Sun, Hindustan Times, Gizmodo, NBC, and Slash Film. Additionally, his research has been incorporated into DeepLearning.AI’s Natural Language Processing in TensorFlow course, training materials for Google’s TensorFlow Professional Certification program, and an NSF-funded project, Narrative Modeling with StoryQ, aimed at teaching text classification concepts to high school students. His research artifacts have also been extensively used by the data science community on platforms like Kaggle to learn and apply machine learning concepts. Find out more about him at https://rishabhmisra.github.io.