Before Deploying AI, You Have to Decide What You're Trying to Make Possible

A growing body of AI investment data reveals a gap between what companies spend and what they get back — and a consumer advocate argues the reason is that most organizations are starting in the wrong place.

Writing in Insurance Business, Kiernan Green profiles Birny Birnbaum, executive director of the Center for Economic Justice and a longtime consumer representative at the NAIC, who argues that AI strategy must begin with a values question, not a capabilities question. His case: the same AI model, with the same data, can either expand access to insurance or accelerate the narrowing of who gets covered — depending entirely on what the organization was trying to optimize for in the first place.

BCG's 2026 analysis found that AI spending as a share of P&C revenue is expected to triple this year, yet only 38% of carriers are generating value at scale from AI in core workflows. Birnbaum's framework suggests the returns gap is a strategy problem, not a technology one.

Key Takeaways

• Start with the outcome, not the tool. Birnbaum's opening principle: "Before we choose our tools and our techniques, we must first choose our dreams and our values." An AI system built to identify vulnerable properties can be used to direct mitigation dollars or to deny coverage — the data is identical; the outcome is not.
• The boardroom test. When an AI use case reaches the executive table, ask: who benefits if it works? A project that serves only the carrier through better segmentation is a different kind of initiative than one that improves affordability, expands coverage, or reduces client losses.
• Governance is strategy. For C-suite leaders, the article reframes AI governance as a strategic decision made before implementation — not a compliance layer applied after the model is deployed.
• The availability crisis is real context. The US Treasury's Federal Insurance Office noted in January 2025 that homeowners insurance is becoming more costly and harder to procure for millions of Americans due to climate risk. AI can help address that problem or worsen it, depending on design intent.

The framework extends well beyond insurance. Any organization asking "what can we automate?" before answering "what outcome are we building toward?" risks the same misalignment.

Read the full article on Insurance Business.