AGENTIC PRODUCT MANAGEMENT.
I'm Michael Lo — Product Manager in Generative AI. I solve complex real-world problems in financial services and insurance by building high-precision AI, tailored to each domain.
FROM COPILOT TO AGENT ORCHESTRATION.
Today, our generative AI layers work as a copilot for frontline claims consultants - not a replacement for them. When a consultant needs to know what the PDS says about an exclusion, or what our processes are, the agent surfaces the answer instantly. That takes the cognitive load of manual lookup off a human's plate, so they can focus on judgment calls instead of search.
"If an agent can curate the right information, answer the question, and enforce the guardrails around it - the human in the loop becomes optional."
The vision: Moving from an agent assisting a claims manager, to a network of autonomous agents coordinated through Agent-2-Agent protocols.
We can STP claims management by orchestrating Agents into a cohesive automated end-to-end process. Each agent stays narrowly focused on one domain task — contract interpretation, knowledge retrieval, compliance checks.
TAILORED FOR EACH DOMAIN.
I've always worked in financial services businesses across Sydney. Over the past few years, I've led regulatory, transformation, and Generative AI initiatives across Australia's leading financial and insurance institutions.
"Getting an AI agent to read a PDF contract is a solved problem from three years ago. The real engineering challenge is contextualizing domain nuances and enforcing guardrails so recommendations don't fail in high-stakes fields."
Consider a simple question: "Is damage to a water tank covered?" On paper, it sounds straightforward. But in general insurance, what is a water tank?
-
01
Residential Rainwater Tank 1,000L to 2,000L tank in a backyard.
-
02
Commercial Farm 40,000L agricultural water tank on a rural property.
-
03
Vehicle Water Tank IBC mounted on a ute/truck.
-
04
UTV/Polaris Mobile Tank Water carrier used on commercial properties or farms.
These are all legitimate uses for a water tank, yet each has vastly different PDS coverage types. Building fit-for-purpose AI products requires embedding these domain guardrails directly into the system - ensuring AI tools never make insurance decisions using their context, instead using your context.
PRODUCTION DEPLOYMENTS.
CLAIMS MANAGEMENT AND KNOWLEDGE ENGINE
The central system managing organisational knowledge for claims management at Australia's largest insurer. Giving users a single source of truth for how we execute processes, our business rules, FAQs - all surfaced instantly.
DOMAIN-AWARE PDS REVIEW TOOL
An advanced AI product designed to read and contextualise Product Disclosure Statements (PDS). Incorporates deep insurance domain guardrails to evaluate nuanced coverage rules accurately, mitigating LLMs misinterpreting our insurance specific context.
- Guardrails inserting property, motor, and liability claim context into the agent
- Prevents misrecommendation where context is crucial (e.g. differing water tanks, how motor vehicles can be an asset in property or motor insurance)
- Used by hundreds of insurance expert users every month across IAG
READY TO BUILD?
Based in Sydney. Available for any chats about building Generative AI products, and financial services transformation.