SARA (Socratic AI Reasoning Agent) Qorus-NTT DATA Innovation in Insurance Awards 2026

SARA – Socratic AI Reasoning Agent

SARA (Socratic AI Reasoning Agent) is a Generative AI reasoning partner for Income Protection (IP) assessors, designed to help customers return to work sooner through better and earlier intervention decisions.

In Irish Life, IP assessors actively manage claims with the objective of supporting recovery and helping claimants return to work. Improving the quality and timing of these decisions delivers meaningful customer benefits and significant financial value for the business.

Concept & Objectives

Income Protection claims are not simple binary decisions. They require continuous professional judgement as new medical, occupational and rehabilitation evidence emerges throughout the life of a claim.

SARA supports this process by combining two capabilities:

• An intelligent workflow that assembles the full case context, including medical evidence, timelines and extracted facts.

• A Generative AI–powered Socratic interaction model designed to strengthen human reasoning rather than replace it.

In practice, SARA provides a structured case workspace where assessors can review timelines, summarised evidence, contextual flags and supporting documentation in one place. The system generates an initial Next Best Action (NBA) and presents it in the workspace chat, where assessors can challenge, question and refine the recommendation through Socratic dialogue.

This environment allows assessors to explore reasoning, test assumptions and converge on the most appropriate intervention strategy for each claim.

What Makes It Innovative – The Socratic Partner

The core innovation is the Socratic Partner interaction model.

Rather than simply responding to prompts, SARA actively engages assessors in structured reasoning conversationally via the workspace chat. It challenges assumptions, asks clarifying questions, highlights trade-offs, surfaces alternative interpretations and encourages assessors to articulate priorities before deciding on an intervention strategy.

Grounded in corporate knowledge and live claims data, SARA nudges assessors from instinctive “System 1” decisions into more deliberate “System 2” thinking, while preserving diversity of thought. The goal is not uniform outcomes, but better-reasoned decisions.

This approach reflects a shift from traditional AI decision automation toward AI systems designed to augment human reasoning in complex judgement-driven environments.

Continual Improvement Through Context Mining

SARA is designed to improve over time through context mining.

During assessor interactions, the system captures the reasoning behind decisions, such as why particular actions were taken or rejected. These insights can be curated and approved as structured knowledge.

Over time, learning from real claims feeds back into the system, strengthening future recommendations while maintaining human oversight.

Reasons Behind the Initiative (Why Now)

Income Protection is a significant service and cost area. Irish Life currently pays approximately €75m in claims annually, expected to rise to €100m by 2026, with increasing claim volumes.

Reducing claim duration through earlier and better interventions improves customer outcomes and is financially material for the business. A 5% reduction in claim duration would deliver approximately €5m in annual savings, while most importantly helping thousands of claimants return to work sooner.

State of Competition / Sources of Inspiration

Many AI solutions in insurance focus on automating decisions using predictive models or rules-based systems. While these approaches can improve efficiency, they are less suited to domains such as Income Protection where complex medical, occupational and behavioural factors require expert human judgement.

Over-automation in these environments can create risks including reduced diversity of thought, weaker ownership of decisions and the gradual erosion of human judgement as people become passive recipients of system outputs.

SARA takes a different approach. Rather than replacing the assessor, it acts as a reasoning partner, using Socratic questioning to strengthen human judgement while maintaining accountability and diversity of thought.

The goal is not to automate decisions, but to improve how they are reasoned.

Departments Involved & Results So Far

SARA has been developed by the ExO Innovation Hub in collaboration with Customer Solutions and Income Protection assessors.

Assessors are currently reviewing Next Best Action recommendations and engaging with the Socratic interaction model as part of the claims workflow.

Whilst claims patterns can take years to emerge, analysis indicates that SARA improves how assessors structure their reasoning, identify potential interventions and plan rehabilitation or return-to-work pathways for claimants. Overall we have seen a 25% improvement in strategic decision making as a result of SARA

"SARA helps me assess claims more effectively. It saves time on unnecessary reviews and helps refine my decision-making. It’s valuable for both new and experienced assessors." Senior IP Claims Assessor