The Signal Boost

Most AI conversations are still stuck on prediction — can we forecast X better than before? But, what if the decisions that actually matter are causal?In this episode, co-host Carolyn Seminario sits down with Susmitha Mutapally. She is an AI and data science leader with experience at Amazon, where she led Alexa voice AI analytics, and Medtronic, where she applies causal inference to customer and business strategy. She's also a NYU Stern MSBAi alumna and Chief Data Officer at Young Aspiring Thinkers, a South African nonprofit whose AI-powered career guidance platform has already reached 1,000+ students without access to professional networks.In this conversation, she breaks down the gap between predictive and causal AI, explains why recommendation algorithms are working exactly as designed and makes the case for why 2026 is a real inflection point for AI governance: not another year of principles, but the moment accountability becomes enforceable.

What happens when psychology, healthcare, and AI collide? In this episode, co-host Luciana Rigotto sits down with Joshua Singer, Senior Director of Clinical Quality Reporting & Data Science at Mount Sinai Health System, to explore what it really takes to build analytics tools that people trust and actually use. Joshua shares how a background in behavioral health shaped his approach to enterprise data, the lessons behind Mount Sinai's first automated quality dashboard, and how he thinks about responsibility when predictive models produce unexpected outcomes. The conversation also takes a more personal turn, diving into the growing use of AI chatbots as a substitute for therapy, the tension between engagement-driven design and genuine mental health outcomes, and what a healthy relationship with AI might actually look like. If you work at the intersection of data, healthcare, and human behavior, this one's for you.

In this episode of The Signal Boost, host Sergio sits down with Gordon Burtch, professor of Modern AI at NYU Stern MSBAi Program, Allen and Kelli Questrom Professor of Information Systems at Boston University’s Questrom School of Business and a Research Scientist with Meta’s Central Applied Science group. They get into what agentic AI actually means in practice, the real obstacles companies face when deploying these systems in regulated industries like banking and healthcare, and why the unpredictability of large language models is something businesses may need to accept rather than solve.Professor Burtch also weighs in on the sustainability of AI training data, the economics pushing companies toward leaner and more efficient models, and why the biggest bottleneck in an AI-saturated world is no longer producing content but verifying it.The conversation covers a look at where the next wave of improvements is coming from and the use cases that have barely been touched yet.