OpenAI recently released a new set of economic evaluation benchmarks called GDP-VAL, designed to measure how well AI performs on economically valuable, real-world tasks across different jobs — including finance and insurance.

What’s notable isn’t the GDP headline itself.
It’s the underlying shift in how AI impact is being evaluated.

From roles to tasks

Historically, conversations about AI in insurance have been framed around roles:
Which jobs will change? Which functions can be automated?

GDP-VAL reflects a different approach. Instead of evaluating AI at the role level, it measures performance at the task level — the individual units of work that make up a job.

That distinction matters in insurance.

How value is actually created in insurance work

Insurance work is rarely transformed by automating an entire role end-to-end. Most roles are composed of a mix of judgment-heavy decisions and manual, repetitive tasks that support them.

Value is created when AI:

• Removes the repetitive, time-consuming work embedded inside workflows
• Improves consistency and accuracy in how information is prepared and reviewed
• Frees professionals to focus on judgment, risk assessment, and decision-making

This is true across underwriting, claims, compliance, and operations. The opportunity isn’t to replace expertise — it’s to remove friction around it.

What we’re seeing in practice

Research like GDP-VAL reinforces a direction that’s already becoming clear in practice. Insurance teams are less interested in broad claims about automating jobs and more focused on identifying specific tasks that can be handled more efficiently and reliably.

That task-level lens aligns better with how insurance organizations actually operate:

• Workflows are layered and interconnected
• Responsibility is distributed across teams
• Judgment and accountability remain central

AI fits best when it’s applied precisely, not generically.

Why the task-level shift matters

Evaluating AI at the task level creates a more realistic framework for adoption. It allows teams to:

• Be deliberate about where AI is applied
• Maintain appropriate controls and oversight
• Measure impact in ways that map directly to business outcomes

In insurance, that shift — from roles to tasks — is where the real opportunity lives.

For those interested, the full GDP-VAL write-up from OpenAI is here.