Carbon’s Ferrier on building a delegated underwriting “engine” as the cycle turns

Carbon Underwriting was last on the Voice of Insurance podcast five years ago, back when it was one of the first Syndicates in a Box at Lloyd’s and part of the early wave trying to modernise delegated authority underwriting.

It is a different business now.

Carbon has grown into a carrier with gross written premium just under $500m, a team of 78, and a proprietary data platform, Graphene, built specifically to deal with the messy reality of delegated portfolios at scale.

What makes the catch-up interesting is timing. Delegated underwriting has had a boom, results have been strong, and the market is starting to soften. That is usually when weak monitoring shows up. Jacqui Ferrier, Carbon’s co-founder and CEO, has built her pitch around not being surprised.

“A kid from Australia… who loves to learn“

Ferrier opens with a line that tells you a lot about her.

Those closest to her, she says, would describe her as “a kid from Australia, from the bush, who loves to learn and loves to live.”

She fell into insurance through necessity rather than a family connection. She took a short-term admin role at Swiss Re Asia Pacific to save for a masters in chiropractic science, then stayed because she was curious and stubborn enough to keep asking for more.

She started at what she calls “the bottom of the bottom”, doing filing and data entry. The deal with her boss was simple. Get the admin done, and she could spend time with underwriters. The rest, as she puts it, is history.

The “gold standard” moment that spooked them

Carbon’s origin story is not a VC pitch. It is a discomfort story.

Back in 2017 Ferrier and co-founder Nick Tai were writing a mix of open market and delegated business at Catlin’s The Box. Lloyd’s was deep-diving delegated authority, asking questions about how syndicates really underwrote and monitored MGA portfolios.

Ferrier says they had a process, but it was manual and spreadsheet-heavy. After the review they were told they were the “gold ticker”, essentially best in class.

“That was when Nick and I looked at one another and thought, this is the best? This is deeply concerning,” she says.

So they did the unfashionable thing. They did not go to private equity with a deck. They remortgaged their homes and started Carbon themselves, with help from Charles Taylor Insurance Services to build the platform around them.

Graphene, and why speed matters in delegated

Carbon’s core bet is that delegated portfolios are not inherently risky. They become risky when you cannot see what is happening quickly enough.

Ferrier talks about the problem in blunt operational terms. In the early years Carbon received around 250 bordereaux a month. Now it receives over 2,500, which she equates to roughly 30,000 bordereaux a year being ingested, cleansed and standardised automatically.

She says the Graphene team now represents about 20% of total headcount, down from roughly half in the early period. In other words, scale without adding bodies every time a new binder lands.

The bigger point is cycle management. Ferrier says the market is still set up to be reactive. In many delegated arrangements, reinsurers might not see meaningful data for six to nine months. That is far too late if a portfolio is drifting.

Carbon’s answer is to share the insights, not hoard them. Ferrier says Graphene is not an internal dashboard. It is a platform that is visible to coverholders, capital providers, managing agents and reinsurers.

The aim is simple. No surprises.

“Boring is beautiful“

The underwriting philosophy matches the tech.

“Boring is beautiful” is a line Ferrier says is used regularly inside Carbon.

The portfolio is built around SME commercial lines, spread across about eight core lines, with 125 risk codes approved at Lloyd’s. From a geographic standpoint the business is international, with a “core 10 countries” where Carbon concentrates rather than trying to be everywhere just because Lloyd’s has licences.

Her point is not that Carbon will never expand. It is that delegated authority only works when you understand what you are delegating. She says Carbon will not “dabble” in territories or lines it does not understand, even if “anything is possible at the right terms and right price.”

Selection is quantitative and qualitative

Ferrier splits partner selection into two buckets.

The first is quantitative. Can Carbon evidence what is in the book, where the profitability sits, and where the pain is concentrated? She repeats the point Mark makes about 80/20. Carbon sees similar patterns, where a minority of premium drives most of the losses.

The second bucket is qualitative, and Ferrier treats it as equally important.

She talks about alignment and honesty. What does the MGA want to build over three to five years? Do they want to invest in claims capability, in their own tech, and in understanding ultimate portfolio performance earlier?

She says many MGAs have never had a capacity provider sit down and talk through the areas that need improvement. It is usually numbers, not quality.

Carbon wants the partners who want to get better.

Claims codification, and the “words to numbers” problem

Ferrier also goes after a piece of delegated underwriting that many people avoid because it is messy.

Attritional losses.

The industry is good at codifying large events. It is much worse at understanding what drives high-frequency, low-severity loss creep. Ferrier argues that without that, you cannot properly manage performance through a cycle.

Graphene, she says, has built a way of translating free-text claims descriptions into Carbon’s own claims codes. She calls it a kind of “Google Translate” for claims.

The episode produces one of those only-in-delegated examples. Ferrier says Carbon saw a 4–5% deterioration in loss ratios in an Italian municipal book during Covid because wild boars moved into towns and caused a spike in traffic accidents.

It is funny, but it is also the point. If you can’t see the detail, you can’t learn from it.

Consortia, whole account, and scaling the model

Mark picks up on Carbon’s move into consortia.

Ferrier describes it as an extension of the model. If Carbon can ingest and analyse delegated portfolios properly, it can bring other market participants into the same view of the risk.

Carbon began with a property consortium, moved to a 10% whole account, and this year sits at a 32% whole account.

Ferrier says it validates their approach because blue-chip peers follow, not just for the underwriting, but for the data transparency.

She also stresses that Carbon does not currently support broker facilities, although she acknowledges Carbon would be well placed to handle them operationally given its ability to ingest data in different formats, including via API.

What comes next: more platforms, possibly managing agency status

Ferrier is “platform agnostic” in the sense that she cares most about security and sustainable capacity, not the label on the paper.

Carbon already writes US business via its MGA platform rather than the Lloyd’s syndicate, and the delayed Syndicate 5757 plan remains alive, with timing linked to capital appetite for US exposure.

She is also open about ambition.

Asked whether Carbon might become its own managing agent, her answer is an unambiguous “absolutely”. The logic is partly economic as scale increases, but also operational. Ferrier argues that Graphene is not a bolt-on analytics tool. It is the digital spine of the business, touching underwriting, finance, operations and compliance.

Not building for an exit

Carbon’s private equity partner, Apiary, comes up towards the end.

Ferrier says there is no process under way and no deal on the table. Apiary has helped Carbon professionalise leadership and governance, and Ferrier says she has enjoyed working with them more than she initially expected.

The focus, she insists, is still on execution.

The delegated market is likely to get more competitive as pricing softens. Ferrier thinks that will “sort the wheat from the chaff”. MGAs that do not add value will struggle. Those that can evidence performance and react early should survive.

And Carbon’s bet is that speed, transparency and a slightly obsessive focus on data quality will keep it in the right group.