We started Marshmallow in 2017 for people like migrants and young drivers, who the market typically misjudges. Since then we have grown from 3 to 700+ people, turned profitable, insured millions of drivers and lent millions in car loans.
We are only just getting started. Over the next 10 years we want to become one of the largest financial services providers in the world, by scaling what we have and building what comes next.
That takes curious, commercially minded people who want autonomy and ownership, and who care more about winning together than being right alone.
Three years ago we brought claims in-house. Not just the team, the tooling. Unlike most insurers we own our own tech stack. That means fast learning, tight feedback loops and quick build cycles. And the people you design for sit in the same office.
That is what automating a single third-party workflow is worth.
Getting to high-quality accident data faster directly cuts solicitor fees and courtesy car days. Every hour you save the customer saves the business too.
Unlike other insurers we build on our own stack, so what you design ships in days, not quarters, and you learn from it just as fast.
The customer side is about meeting people at the point of distress, and capturing the signals they give off naturally instead of making them fill in forms. You need to be comfortable designing for someone's worst day.
Tools for our claim handlers, balancing partners and workflows. The long game is deciding, case by case, whether AI is a co-worker for handlers or replaces the workflow entirely.
The split shifts as you and your PM choose where the impact is each quarter.
Nobody hands you a roadmap. You and your PM decide what matters each quarter. Curiosity and a bias for action are the job.
Internal and external. Gather feedback, run your own studies and use insight to inform design decisions.
Use design principles, industry knowledge and best practice to make rational decisions that help customers when they interact with Marshmallow.
Design AI workflows that automate claim handling, and experiment with AI in your own process. Free rein with Claude, shipping internal tools straight to production.
Claims is a costly area with a lot of attention on it. Take stakeholders on the journey and keep them in the loop.
Pair with peers, hold design reviews, facilitate workshops and contribute to the wider design system.
A small, senior team that pairs on work, holds design reviews and runs its own research end to end.
Your cross-functional partners in the claims team, plus an operations floor of 200+ claim handlers to do research with and gain insights from.
Every Friday the team shares weekly wins, from hackathon medals to a test that moved cancellations by 4.2 points. And a few times a year we get out of the office for a team day: strengths, gaps, rituals and handwritten appreciation notes for each other.
A recent workshop: designing a claim snapshot that gets handlers up to speed in seconds. Everyone sketching, from engineers to handlers.
The Slack moment our first AI agent went to production, and the handler feedback that landed the same day.
From workshop sketches to an AI agent live on real claims, with handler feedback landing the same day. That is the loop you will work in.
The password is in the email we sent you.