Cutting quote time by 75% with agentic underwriting
Services
Product design · UX research · Interaction design · Cross-functional facilitation · Change management
Description
Built a zero-to-one agentic underwriting system designed to aid high-stakes decision-making. The product surfaces insights and recommended actions while keeping people in control of the final call
Impact
~75% faster quoting
~40% reduction in decision-making time
~60% less manual data entry
The starting point wasn't a blank page, it was an overbuilt prototype and a team that hadn't yet been sold on why design mattered.
01
An Overbuilt Starting Point
Brought in on a rough, overinflated vibe coded prototype that had run ahead of its own purpose. Team fell into a fidelity trap and were attached to the prototype due to it's polish. The first job was stripping it back to the real value proposition and core user needs,
02
Design's Value Wasn't Understood Yet and Trust was Missing
The team hadn't fully grasped the value of design, or the adoption risks and change management the project needed. Trust, both in the AI and in design's role was the real barrier to progress.
03
Speed Without Losing Rigour
With engineers now shipping faster through vibe-coding, design had to move at the same pace without skipping validation. That meant pairing speed with sharp insight to prove where design decisions were actually earning their keep.
MVP/POC team makeup guidance and right, is documentation behind AI patterns and experience principles
Building a set of AI patterns for the design system, accompanied by AI experience principles. In order to accelerate building trustworthy AI products.
How I created value
Moved the work from an unproven prototype to a trusted, adopted system then made the patterns reusable beyond this one project.
01
Research & Design
Ran research to map existing underwriting workflows and tested prototypes to surface trust and usability gaps. Designed interaction patterns balancing AI autonomy, human control, and feedback, then partnered with engineers and data scientists to ship production-ready workflows.
02
Building Trust to Drive Adoption
Built trust through visible rationale and traceability, plus an interim human-in-the-loop review phase to prove extraction capability before full handover. Ran a champions programme and applied Lewin's change model to win underwriters over from familiar, established processes.
03
Scaling the Way of Working
Shifted the team from a proof-of-concept to an MVP mindset, and helped shape a resourcing model that could scale with the agentic AI roadmap. Formalised reusable patterns into the design system and built Claude agent skills so future teams could build faster while staying aligned.
