Does AI Require Oversight? Dr. Michael Housman Speaks to General Assembly
In a talk with General Assembly, Dr. Michael Housman, Chief Data Scientist of RapportBoost.AI, explains why deploying artificial intelligence is not a “set it and forget it” endeavor. He challenges the misconception that AI systems, once built, will continue to deliver value without ongoing human involvement.
Housman emphasizes that AI models are built on historical data, which reflects a specific moment in time. As customer behavior, markets, language, and business strategies evolve, the assumptions embedded in those models begin to degrade. Without continuous monitoring and adjustment, even high-performing AI systems will slowly drift away from reality.
A central issue is that data distributions change. Customers adopt new behaviors, new words and symbols enter everyday language, and external conditions shift. When this happens, models trained on past patterns can become less accurate or even counterproductive. AI must be retrained, recalibrated, and validated to remain effective.
Housman also highlights the importance of feedback loops. The most successful AI deployments include mechanisms for observing outcomes, testing alternative strategies, and learning from real-world results. This allows organizations to distinguish between correlations and true causal effects, rather than blindly trusting model outputs.
Another key theme is the role of humans in the system. AI works best in a human-in-the-loop framework, where people evaluate recommendations, override them when necessary, and provide contextual judgment. This human input not only improves decisions but also helps the AI learn what actually works in practice.
Housman concludes by stressing that deploying AI is closer to managing a living system than installing traditional software. It requires ongoing investment, experimentation, and governance. Organizations that treat AI as a dynamic capability—rather than a one-time implementation—are far more likely to see sustained value and competitive advantage.
