Building a UX Operating Model for AI-Assisted Claims Operations at AIG

Helping a UX team bring structure, human judgment, and trust to AI-enabled insurance experiences.

AIG · 2026
Role: GenAI Design Strategist
Team: 4-person UX team
Focus: Claims, AI-assisted workflows, UX operating model

The Challenge

AI was moving faster than the UX process could support.


AIG was developing Claims Assist to help commercial claims professionals work through complex information and decide what to do next.

The opportunity was significant. AI could extract information from documents, identify potential duplicates and contact matches, summarize losses, and help connect claim information to policies and endorsements.


But introducing AI into a high-stakes workflow creates a different design problem.


It isn’t enough for the AI to be useful. People need to understand what it is doing, verify important information, correct it when necessary, and remain in control of decisions.


At the same time, the UX team was working in an environment where priorities and requests could move quickly, making it difficult to consistently create space for discovery, validation, and thoughtful handoff.


I saw an opportunity to address both problems.

My Role

I led UX for Claims Assist while also helping establish a more consistent way for the team to approach AI-enabled product work.


My responsibilities included:

  • Setting UX direction for Claims Assist
  • Leading discovery and cross-functional alignment
  • Guiding product and interaction design
  • Establishing a structured UX workflow
  • Leading UX validation and testing
  • Introducing a human-AI trust review
  • Exploring information needs across Claims and adjacent insurance functions
  • Coaching and directing a four-person UX team
  • Partnering with Product, Engineering, Research, and business stakeholders


The goal wasn’t to add process for the sake of process. It was to make it easier for the team to make good design decisions at the right time.

The Interventation

I created a practical UX operating model.

I introduced a clearer path for moving work from an initial request to delivery:

Intake → Discovery → Design → Validation → Handoff

Each stage answered a different question.


Intake

What problem are we actually trying to solve?


Discovery

What do people need to accomplish, and what constraints shape the experience?


Design

What should the experience be?


Validation

Does it work for the people who will use it?


Handoff

Have we resolved enough detail for the experience to be built as intended?


This gave the team a shared language for discussing UX work and made gaps visible earlier.

But AI required something more.

Traditional UX evaluation wasn’t enough for AI-enabled workflows.


An interface can be easy to use and still create problems if people don’t understand where an AI-generated answer came from or whether they should act on it.


So I introduced a Trust-Pattern Review as part of the design process.


I asked six questions.

  • Can I see where this information came from? / Traceability
  • Do I understand how certain the system is? / Uncertainty
  • Can I understand what the AI did? / Explainability
  • Can I correct the AI? / Human control
  • What happens when the AI is wrong? / Recovery
  • Can someone understand what happened later? / Auditability


These became practical design questions rather than abstract principles.

One Design Principle

Let AI do more of the information work. Keep people in control of the decision.

AI

Finds
Organizes
Summarizes
Suggests

↓

ADJUSTER

Reviews
Verifies
Corrects
Decides


This distinction became a useful way to think about where automation belongs and where human judgment matters.

Impact

A more consistent way to do UX.

The operating model created clearer alignment across UX, Product, Engineering, and business stakeholders.


25% reduction in design-to-development rework

The improvement came from addressing questions earlier, validating important decisions before handoff, and creating greater clarity around the work.


I also established a repeatable approach for evaluating human-AI interactions beyond traditional usability.

Key Takeaways

AI UX is not just about making AI useful.

It is about designing the relationship between:

People + Information + AI + Decisions


And sometimes the most important design isn’t another screen.

It’s creating the conditions for a team to consistently make better decisions.


Note: Details have been generalized to protect confidential business information.