Klarent Builds One Trusted View of the Insurance Customer From Lead to Claim

It’s time to plan next quarter’s ad budget. You can see what each click costs, but you have no way of knowing which clicks drive ROI. You guess — and it costs real money.

This was Klarent’s blind spot. Its parent company Telenav spent years building navigation software for tens of millions of vehicles and, in the process, developed expertise collecting and processing billions of automotive data points. Klarent drew on its parent company’s experience with data at scale for its end-to-end operating platform for personal-auto managing general agents (MGAs), which brings together smartphone-based telematics with customer, policy, billing and claims data. 

Novo Insurance, Klarent’s first customer, ran its Arizona personal-auto MGA business on the platform. As the platform and Novo’s book of business grew, the problem sharpened. Marketing data sat apart from policy data, and telematics and claims data sat somewhere else. Nobody could follow one customer across the entire journey. Each system also defined the same customer, policy and claim in its own way, so the same question got different answers.

The old tools produced reporting rather than answers. A data warehouse and two product analytics tools each described a slice — enough to see last quarter’s numbers but not enough to know which campaigns converted new customers. A separate query engine sat over a data lake in Amazon S3. The Spark pipelines ran on Amazon EMR, where the longest took 10 hours 13 minutes. Keeping those clusters alive cost real engineering time.

Klarent didn’t need a better dashboard. It needed one shared definition of the business, on a data and AI platform where anyone with a question gets the same trusted answer.

With multiple tools already in place and still no unified view, the question wasn’t which tool to add, but how to build the right data and AI foundation. Having built data products before data marketplaces even existed, the team wanted a single, cost-effective solution that could replace the complexity and expense of managing multiple tools and platforms.

Two Snowflake benefits settled it: Computing power could run beside the data, and a separate system for governance wasn’t needed.

Klarent migrated in two stages. First, it retired its data lake query engine in a few months, using Snowpipe to load data from S3. Early results with Cortex Agents moved the AI work forward before stage two. In stage two, Klarent moved its heavy Spark workloads off Amazon EMR with Snowpark Connect for Apache Spark™.

The end result was a unified governance layer on Snowflake: a shared dictionary that defines each customer, policy, trip and claim once, organized into domains for marketing, policy, telematics and claims. Every agent, report and application reads the same definitions. 

In Snowflake CoWork, a marketer or support lead using the Klarent platform can now ask questions in plain language while the master agent, built on Cortex Agents using Snowflake CoCo, answers, covering data from leads to claims. The same governed foundation powers the AI-assisted reporting portal that Klarent’s MGA customers use to ask plain-language questions about their book of business.

Development and quality assurance teams use CoCo to generate SQL, write agent instructions and semantic definitions, and get new hires productive in the codebase — delivering more work with the same headcount.

A marketer can now trace what it cost to bind a policy back to the channel that produced it, or where prospects fell out. With Snowflake, non-technical employees across product, marketing, business and customer support answer their own questions and get the same answer to the same question, and engineers deliver more without adding any headcount. Work that once consumed several days can now be completed in a single day.

Klarent’s migrated Spark pipelines, spanning more than 100 billion events a month, now run 3.6x faster at 60% lower TCO compared to Amazon EMR. Most internal tools now also run on Streamlit in Snowflake. The next step is to complete the migration — bringing every pipeline into Snowflake and creating a unified foundation where Klarent can govern data, build applications and scale AI with minimal infrastructure to manage.

Klarent’s rule was foundation first, agents second: Define the data once, govern it, then build agents on it. The team didn’t want fast answers from data nobody trusted. Thanks to its approach, the same marketer planning next quarter’s budget can now follow any customer from click to claim, and every employee gets the same trusted answer to the same question.