The Man Who Built InsurTech Before It Had a Name

In 1994, Jon Kelly was using Mosaic on a Stanford server, watching Yahoo run out of a dorm room a hundred yards from his own. He was an econ major surrounded by future Silicon Valley legends, interviewing with banks and consulting firms who dismissed his obsession with “this internet thingy” as a distraction from a real career.

By 1998, at 25 years old, he was co-founding eCoverage — backed by SoftBank, Accel, and RRE Ventures — with a one-line mandate: build an online version of Geico. Eleven months later, they issued their first policy. No AWS. No infrastructure to borrow. They built their own server racks and constructed a full-stack insurance carrier from scratch, becoming the first startup to sell car insurance online in 1999.

That is one exit. It is not even the most interesting one.

In Episode 164 of InsurTechTalk, Jon Kelly — now CEO of Modern Metric — walked me through three decades of building at the frontier of insurance and technology, from the dot-com collapse that killed eCoverage’s Series C to the invention of the click wall that became a billion-dollar lead generation category, to the six-week onboarding nightmare that led him to build the private client insurance platform he needed and nobody else had built.

About Jon Kelly

Jon Kelly is the CEO of Modern Metric, a software platform built for private client and high-net-worth insurance brokers. He co-founded eCoverage in 1998, the first venture-backed online auto insurer, which sold to General Motors in 2001 and became GMAC Insurance Online. He then co-founded SureHits, inventing what the industry now calls the click wall — a JavaScript plug-in that let insurance carriers bid for excess consumer traffic — which he sold to QuinStreet before running their insurance division. After a detour into social media (a company popular with teenage girls in Commonwealth countries that proved very hard to monetize), he founded Kelly Klee, a private client insurance agency built on the thesis that high-net-worth insurance brokers had a massive technology gap. He sold Kelly Klee to Foundation Risk Partners a few years ago — but kept the technology, which became Modern Metric.

eCoverage: Building a Full-Stack Carrier in 11 Months, With No Cloud

The origin story matters because it establishes something rare: Jon was not reacting to InsurTech. He was building it before the category existed, at a moment when almost no one in the business world took the internet seriously.

What eCoverage Actually Was

• A full mandate: build the online equivalent of Geico, backed by SoftBank, Accel, and RRE Ventures
• Built from scratch in 11 months — no AWS, no cloud infrastructure of any kind, physical server racks constructed by the team itself
• The first startup to sell car insurance online, launching in 1999
• Jon and his team bought the second-ever car insurance ad on Google — before AdWords existed, on a CPM basis rather than CPC, purchased through an insertion order. Progressive was first. eCoverage was second.

What Happened When the Bubble Burst

eCoverage began raising a $50 million Series C in December 1999, with Solomon Smith Barney running the private placement. Then a Barron’s article ran in March or April 2000, and — in Jon’s words — it was like a light switch turning off. Massive inbound interest evaporated overnight. The company that had been the clear category leader in online insurance could not close its round. eCoverage sold to General Motors, becoming GMAC Insurance Online. Jon was 27.

The lesson he carries forward: you can have something that looks like it is genuinely working, watch it fail anyway, and try again without being too attached to any single outcome. That is not a talking point — it is the operating principle behind three subsequent companies.

SureHits: Inventing the Click Wall

Jon’s second act came through a partnership with John Jones, a search agency operator working with GMAC Insurance. The origin of the click wall — now a foundational mechanism behind companies like MediaAlpha and QuinStreet — was almost accidental.

GMAC only operated in about 30 to 35 states. Their client wanted to monetize the traffic from states where they could not write business — redirect it to other carriers. Jon and John built a page to sell that excess traffic to competing carriers. Then someone suggested: why not have the carriers bid for it? That question became SureHits, and the JavaScript plug-in they built to power it became what the industry now calls the click wall — deployable on any website, not just their client’s excess traffic pages.

SureHits sold to QuinStreet. Jon ran their insurance division for a couple of years afterward.

The Complicated Reputation of Lead Generation

Jon’s explanation of why lead generation attracts so much distrust in insurance was one of the most candid parts of the conversation.

• The legitimate case: Most consumers genuinely do not know where to start. Ask 100 people whether Geico sells health insurance, and Jon’s estimate is that half would say yes. Getting a consumer from “I need insurance” to five qualified providers who actually write in their space and convert well is a real service — not a scam
• The bad-actor problem: Lead generation attracts people who can make fast, easy money before anyone catches on to what they are actually doing. When the 2008 mortgage market collapsed, a wave of mortgage lead generation operators — many of whom were, in Jon’s words, simply bad people — moved directly into insurance lead gen
• The opacity problem: Buyers cannot see under the hood. Stale leads, warm leads, recycled leads get sold interchangeably, and by the time a buyer realizes the quality does not match the invoice, the operator has often already moved on

Kelly Klee: Six Weeks to Get a Chubb Policy

After selling SureHits, Jon tried to buy a Chubb policy for himself. It took six weeks — and only six weeks, not six months, because he called his broker every three days asking why it was not done yet.

The broker kept discovering new questions mid-process. “Your house is in an LLC — why didn’t you tell me that mattered?” Jon’s answer: he did not know it mattered, because nobody had a system that surfaced the right questions upfront.

That experience became the founding thesis for Kelly Klee: personalized, high-net-worth insurance brokers had a massive technology gap — a thesis that turned out to be, in Jon’s words, a thousand percent true. He built Kelly Klee around solving it directly. Five years later, he sold the agency to Foundation Risk Partners. He kept the technology. That technology became Modern Metric.

Why the Technology Gap Exists at All

This was one of the sharpest structural observations in the conversation.

Direct-to-consumer carriers — Geico, Progressive, State Farm, Allstate, Farmers — built genuinely excellent technology because they had to. They deal directly with consumers, so their agent and consumer-facing tools evolved accordingly.

None of that investment carried over into the independent agency channel — and Jon’s explanation for why is specific: scale. Independent brokerage scale is a relatively recent phenomenon, built over the last 20 to 30 years through waves of consolidation. Marsh, Hub, Willis, Aon — the rollups that created enterprise-scale brokerages also created the first IT departments capable of evaluating and standardizing enterprise software across dozens of offices. Before that consolidation, agency operations were fragmented and ad hoc, with no infrastructure layer worth building for.

Agency management systems solve two problems well — accounting and system-of-record documentation for E&O protection — but were never designed to be client-facing. Jon’s blunt assessment: they look like Windows 95, and nobody would want to share that interface with a client. That left an entire missing layer in private client insurance that direct-to-consumer carriers had already solved decades earlier for the mass market.

What Modern Metric Actually Does: Discover

Modern Metric’s flagship product, Discover, is positioned as the actual workstation for a private client broker — not a replacement for the AMS, but the layer where the real work happens.

Two Kinds of AI, Doing Two Different Jobs

Jon drew a sharp distinction that is worth sitting with, especially for anyone evaluating AI tools in insurance broadly.

Data aggregation and ingestion:

• A private client broker operates like a detective — assembling data from the client directly, from financial advisors, from third-party assessor data, from declaration pages, driver’s licenses, and complex asset schedules (Jon described a 100-page fine art schedule as a real example)
• Discover’s proprietary data model brings all of this together, deduplicates it, organizes it, and surfaces it in a usable interface
• Modern Metric is approaching 100,000 documents ingested specifically within the private client space — a lead that compounds, because a generalist trying to solve this across all of insurance sees a fraction of the document volume and pattern density that a specialist sees

Analysis:

• Carrier underwriting guidelines are extraordinarily complex — Jon described Chubb’s New York underwriting guidelines as nested if-then logic requiring what he jokingly called a 130 IQ to parse manually
• An LLM can review a case against those guidelines and flag violations instantly
• Modern Metric’s “Proofreader” feature reviews a proposal and flags mismatches — a $10 million home with a $1,000 deductible, for example, where the client is needlessly overpaying instead of self-insuring the first $20,000
• This is not replacing the broker’s judgment — it is giving them superhuman visibility across an account that is too complex for any one person to hold in their head completely

Why AI Helps With Analysis But Not With Workflow

This was the most practically important insight of the episode, especially for anyone leading technology strategy at an enterprise brokerage.

Founders, solo operators, and technical people can build highly customized personal workflows inside an LLM — and it works well for them individually. But that does not scale. A brokerage with 500 private client producers cannot have each of them building bespoke LLM workflows to get through their day. They need to show up, log into a single platform, and see everything that needs to happen — proposals to generate, quotes to process, complex asset schedules to review — without reinventing their process every morning.

One of Modern Metric’s clients put it directly to Jon: she does one-off things in Claude here and there, but Discover is where she comes to work every day, because it is where she can see everything and know what needs to be done.

The Closing Answer: Operational Excellence

Jon’s answer to what the industry should be talking about more was not about AI, distribution, or capital. It was about process.

Private client insurance, in his experience, is full of extraordinarily smart, dedicated people who make things work through sheer competence and effort — not because the operational infrastructure supports them. There is no Costco or Amazon-style logistics discipline: defined, repeatable, improvable process. The result is enormous wasted time and effort on things that do not actually deliver value for the client.

His sharpest line: the secret to what makes Modern Metric’s platform actually work is not the AI tools. It is standardizing the process. The industry — across the board, not just at Modern Metric — does not talk nearly enough about the unglamorous discipline of operational excellence.

Key Takeaways

• Jon Kelly built the first venture-backed online auto insurer in 1998, before cloud infrastructure existed, and watched it die when the dot-com bubble collapsed his Series C overnight
• The click wall — now a billion-dollar mechanism across companies like MediaAlpha and QuinStreet — originated almost accidentally from a client’s request to monetize excess web traffic
• Lead generation’s complicated reputation comes from genuine value mixed with an industry that attracts opportunistic bad actors during every capital cycle
• The technology gap in private client insurance exists because brokerage scale — and the IT infrastructure that comes with it — is a relatively recent product of industry consolidation
• AI excels at data aggregation and complex analysis in insurance, but does not scale as a workflow replacement for large teams who need a shared, repeatable operating platform
• The industry’s biggest unaddressed gap is not technology adoption — it is operational excellence and standardized, repeatable process