As the cost of AI products falls, the quality of the problem they solve becomes a more important determinant of value. Yet many product teams still start with a mandate to use AI and then look for a problem to justify it. That’s the build trap. Product leaders must reverse the order. Validate the root cause, test solutions against it, and only then define the AI capabilities that the product needs.
Build The Product Around The Root Cause
The first report in this series covers how to define the problem and validate its root cause. Product and digital teams first generate and compare several solutions, without assuming that AI is the answer. They then translate the strongest solution into the product and enterprise capabilities needed to deliver it reliably, responsibly, and at scale. In practice, that means:
- Treat every solution as a hypothesis. Bring product, design, engineering, data, risk, and frontline teams together to generate options broadly, then narrow them on customer value, feasibility, commercial impact, and strategic fit. Netflix tackled viewer disengagement this way, developing multiple artwork variants per title, each a hypothesis about which visual cues signal relevance.
- Ship the capabilities, not just the feature. A feature that uses AI is not the same as the capabilities that make it reliable, scalable, trusted, and economically viable. Those capabilities span people, process, data, governance, architecture, and technology. For Project Nemo, which automates food spoilage claims after natural catastrophes, Allianz had to build several of them alongside its AI agents.
- Name how the product creates value. State the mechanism (productivity, conversion, retention, operating cost, or differentiation) before you build. Project Nemo cut Allianz’s claims processing and settlement time by 80%. Yet Forrester found that only a handful of organizations prioritize the strategic mechanisms that decide whether AI strengthens their competitive position, brand equity, or adaptiveness.
- Right-size each capability. Don’t default to the most sophisticated model or the highest accuracy, because thresholds for precision, speed, autonomy, explainability, and human oversight differ even within one product. Southern California Permanente Medical Group’s Kaiser Permanente Intelligent Navigator prioritizes precision for high-acuity risk and breadth for care navigation, achieving an approximately 54% booking success rate. Build the smallest viable capability configuration that produces the outcome and generates evidence.
Cheaper AI lowers the cost of building but compounds the cost when teams build the wrong thing. Read my latest report to move from a validated root cause to the capabilities that your AI product needs. Got a question? Book a guidance session with me.
