Aviva accelerates underwriting decisions on better data with hx

About Aviva

Aviva's Global Corporate & Specialty business prices and underwrites large corporate risk for the London and Lloyd's markets, across lines from property and construction to cyber and professional indemnity. It is a mature, sophisticated pricing operation, and it has run its pricing on hx for four years, since it moved 20 pricing tools onto the platform in nine months.

The Challenge for Aviva

Underwriters have priced the same way for decades

The way an underwriter approaches a submission has barely changed in a generation. A broker sends over a folder of documents, sometimes 20 or 30 Word files, PDFs, spreadsheets, and presentations, and the underwriter reads through, hunting for the same handful of data points on every risk, then keys them into a pricing tool by hand.

Manual entry costs quality as well as time

That manual step degrades the data itself. Joe Rogers, who co-led the integration, saw the cost land in three places: errors, incompleteness, and inconsistency. "A trailing zero on an exposure metric can really harm your rate adequacy," he says, and every error then flows downstream into Aviva's pricing metrics. Catching one was slow work: half a day to identify it, then emails back and forth with the underwriter to confirm and correct it.

A softening market raises the stakes

In a competitive market, that lost time carries a commercial price. Every hour spent retyping a submission is an hour not spent with a broker. As McMurrough puts it, "the speed of response could make the difference between winning an account or not." And some of the most valuable data never made it into pricing at all: claims history locked inside unstructured PDFs, too costly to extract by hand, so it was left on the table.

How hx Helps

Any submission, any format, straight into the pricing tools

hx's ingestion agent reads a broker's submission, whatever the format or structure, and pulls the relevant data straight into Aviva's existing hx pricing tools. "What we love about the ingestion agent is that it's sophisticated enough to work with lots of different types of documents regardless of format or structure," says Tabitha Ong, Head of Financial Lines, Technical Underwriting.

The system is designed not to be code-heavy: define which fields the agent should map, define what those fields mean, and overlay a playbook of instructions on top. "We had analysts on the team who were able to pick up prompt engineering really quickly," says Shyam Bhayani, Head of GCS Pricing, "and then they were off implementing it." And because the setup was tailored to Aviva's own document types, each of the 11 tools could be shaped to the line of business it served.

First tool live in five weeks, then two weeks per tool

Aviva started the project, with the first tool, motor, targeted to go live in five weeks. "A lot of people doubted we could do it," Bhayani says, "but we got it done." After motor the pace held at roughly two weeks per tool, until the ingestion agent was live across all 11 GCS lines.

Underwriters reach judgment sooner

With the same key information sitting in the pricing tool automatically, the pace of the work changed.

That speed runs straight through to the broker. Underwriting assistants spot data issues sooner, so underwriters can go back to the broker to resolve them; and with pricing completed faster, underwriters return to the broker sooner. In a soft market, that is the difference between competing for an account and missing it.

Better data, better prices

The decisions themselves improved, not just their timing. Cleaner, more complete, more consistent inputs mean Aviva prices more accurately, spends less time in the correction cycle, and can segment its portfolio with more confidence. The sharpest gain is data that used to be out of reach entirely.

For McMurrough, that is the part that matters most. "We're most excited about the quality of data we're now capturing, and the widening of the type of data we can ingest," she says. "It'll allow us to build better models that we can refine risk with, and come up with a better price to take to market. That, for me as an actuary, is really exciting."

The hx Difference

AI where the pricing already lives

Aviva chose not to build the capability in-house, and not to bolt a separate AI layer onto the side. Ingestion exists to feed the pricing tools, so it belongs at the point of pricing, in the environment the pricing team already uses day to day.

Keeping control inside the pricing team meant the people who know what to ask the model were the ones configuring it. The agent takes on the manual, laborious extraction, freeing underwriters to spend their time on the judgment calls that win accounts. "It's not about replacing expertise," McMurrough says. "It's about taking away the repetitive tasks."

A partnership with a track record

The confidence behind that five-week target came from experience. Four years earlier, Aviva had moved 20 pricing tools onto hx in nine months. McMurrough, newer to the relationship, still cites it: "nine months, which from my previous experience elsewhere is phenomenal."

The work has become a recruiting story too. "I still have people come to me in interviews or at events talking about the case study we did with hx four years ago," Bhayani says. The pace of change, McMurrough adds, is itself an attraction: "anyone wanting to join this market would be attracted by that."

Extending the approach

Aviva is already going further. Schedule-of-values data for the property lines, dense, unstructured, and slow to review by hand, is next for the ingestion agent. Alongside it, Ong's underwriting team is developing an adverse media tool that pulls information from many sources into one place for underwriters across GCS. And because the agent sits inside the pricing tools, each advance in the underlying models compounds without extra work from Aviva's side.