How Do Brands Get Started with AI? Dr. Michael Housman Speaks to General Assembly

In a talk with General Assembly, Michael Housman outlines the three foundational things brands need to get started with AI. Rather than focusing on hype or advanced algorithms, he emphasizes practical building blocks that determine whether AI initiatives succeed or fail.

First, brands need a clearly defined problem. AI is most effective when applied to specific, well-scoped questions with measurable outcomes. Organizations that begin with vague goals like “use AI” or “be more data-driven” often struggle. Success comes from identifying a concrete decision or behavior to improve—such as increasing conversions, reducing churn, or improving response quality—and working backward from that objective.

Second, brands need the right data. Algorithms do not create value on their own; data does. Housman stresses that data must be relevant, reliable, and tied directly to the problem being solved. In many cases, companies already have useful data but haven’t structured or connected it in a way that allows meaningful analysis. Importantly, better data often matters more than more data.

Third, brands need people and processes that support AI. This includes individuals who understand the business context, people who can work with data, and leaders who are willing to trust insights generated by algorithms. AI works best in a “human-in-the-loop” model, where machines surface recommendations and humans apply judgment, experimentation, and accountability.

Housman emphasizes that these three elements—clear problems, quality data, and the right organizational mindset—are far more important than choosing the most sophisticated tools or platforms. Many early AI failures occur because companies invest in technology before putting these fundamentals in place.

He concludes by encouraging brands to start small, prove value quickly, and iterate. When organizations align their questions, data, and people, AI becomes a powerful enabler of better decisions rather than a risky or abstract technology experiment.