How Relm Insurance embeds AI across underwriting, leadership, and product design
Six years is short for a carrier. It was long enough for Relm Insurance to hit the limits of traditional operating models and put technology at the center of the business.
"We have seen over 13-14,000 opportunities in more than 40 countries," said Christian Davies, global head of distribution and innovation at Relm. "It was a scramble trying to build and build as the scale increased." Late last year, leadership paused and asked a blunt question: how much could be augmented, simplified, or sped up with AI and new tech? "The answer was almost everything," Davies said.
Drop the buzzwords. Use proven tech.
Relm took a practical view. Davies pushed back on the need for a special category of "insurtech." The team focused on bringing proven, native technologies into a 300-year-old industry instead of treating them as experiments.
AI and blockchain were treated as operating tools, not side projects. Relm has insured digital asset and emerging tech risks from day one, so applying the same tools internally was a logical step. As Davies put it: "We've been insuring this space for six years. There's no reason we shouldn't use it ourselves."
Start inside: private AI for everyone
Relm rolled out a private, corporate version of ChatGPT across the company. Staff got access to generative AI without data leaving the environment. No external training data, no data exposure.
From there, the team began building in-house models focused on underwriting efficiency and decision support. The goal: reduce manual work, improve consistency, and help people make better calls faster.
Learn more about enterprise-grade ChatGPT
Re-engineer underwriting workflows
Relm built a proprietary classification model at the front of the underwriting process. "We built our own GPT," Davies said. Every submission ran through it to identify the industry vertical and check fit against Relm's appetite guide.
The model also linked decisions to regulatory and licensing requirements, important for a group regulated in Bermuda by the BMA. That connection kept triage aligned with what could actually be written and where.
Underwriters received a "sector classifier assistant" with a two-page overview: what the sector is, appetite alignment, and flags based on public signals. It cut research time and made decisions more consistent without replacing judgment.
Bermuda Monetary Authority (BMA)
Test ideas. Move on fast.
Relm experimented with external tools but prioritized hands-on learning. "When we knew how to use them, we had a better understanding of how to underwrite them," Davies said.
One third-party knowledge base project failed a simple test: speed. Load time was ~10 minutes, so the team walked away. Instead, they built an internal "ask me anything" that pulled data across file systems, Outlook, Teams, and internal messages in real time. Speed and usability beat novelty.
Make innovation a leadership mandate
This wasn't delegated to a lab. It came from the top. "It was coming from our CEO and founder, Joe," Davies said. "It was embedded in the management framework."
For 2026, every divisional leader must identify at least two areas to augment with AI or other tech. It's written into leadership KPIs, with bonuses tied to day-to-day adoption. With that clarity, resistance never formed. "There was no one in the company who said 'no'."
What insurers can apply now
- Give everyone access to a private AI workspace; protect data from day one.
- Build a simple triage model to classify submissions and check appetite fit.
- Connect intake to regulatory and licensing rules to avoid dead ends.
- Ship a fast internal search assistant that pulls from email, chat, and files. If it's slow, it won't get used.
- Tie AI adoption to leadership KPIs and compensation. Make it a mandate, not a memo.
- Kill slow or clunky pilots quickly. Keep the ones that save time this quarter.
Bottom line
Relm's playbook is simple: prove value inside the workflow, move with speed, and make leaders accountable. That's how AI becomes part of underwriting, not a side project.
If you're skilling up teams on practical AI for insurance roles, here's a curated set of options by job function: AI courses by job.
Your membership also unlocks:
