Why a Former Treasury Minister Just Bet Millions on the Most Boring Corner of Finance

Why a Former Treasury Minister Just Bet Millions on the Most Boring Corner of Finance

The water rises slowly at first. You do not hear a roar when a pipe bursts behind the drywall of a third-floor flat in Manchester at two in the morning. You hear a soft, persistent hiss, like rain against a windowpane you forgot to latch. By the time the ceiling plaster sags and gives way, ruining three generations of family photographs and soaking the floorboards down to the joists, the real disaster has not even begun.

The real disaster arrives six weeks later, wearing a lanyard and carrying a clipboard, speaking a dialect of exclusions, sub-limits, and actuarial caveats that nobody outside of a wood-paneled room in the City of London actually understands.

Insurance is supposed to be a promise made in daylight to protect you in the dark. Yet for anyone who has ever stood in a ruined kitchen or watched a small business stall while waiting on a payout, the industry feels less like a safety net and more like an endurance test. It is an engine built on friction. Paperwork piles up. PDFs bounce between brokers, underwriters, and loss adjusters like ping-pong balls in slow motion. Weeks stretch into months while human beings wait to rebuild their lives.

So when a former British government minister—someone who spent years sitting at the nerve center of Westminster’s financial policymaking—decides to walk away from the corridors of power and raise millions of pounds to fix insurance with artificial intelligence, you have to ask a simple question.

What did he see from the inside that convinced him the old machinery was beyond repair?

The View From Horse Guards Road

To understand the bet behind i23, the artificial intelligence start-up co-founded by former City Minister John Glen, you first have to understand what a City Minister actually does.

Officially, the Economic Secretary to the Treasury is the bridge between the British government and the square mile of financial institutions that pump oxygen into the UK economy. During his tenure—the longest of any City Minister in modern history—Glen sat through the aftershocks of Brexit, the emergency liquidity scrambles of the pandemic, and the frantic race to rewrite financial regulation for a digital age. He listened to bank chief executives, fintech founders, and insurance titans explain why London was either on the verge of a golden era or slipping quietly into irrelevance.

Most politicians leave office and take predictable paths. They join legacy boards, write memoirs, or give speeches about regulatory reform at conferences where the coffee costs twelve pounds a cup. They stay safely inside the velvet ropes of the establishment.

Glen went into the trenches instead. Alongside co-founders with deep roots in data science and commercial risk, he stepped into the arena to build i23, securing millions in early backing from investors who recognized that the most lucrative revolutions rarely happen in flashy consumer apps. They happen in the plumbing.

And insurance, for all its trillion-dollar balance sheets and historic Lloyd’s of London underwriting rooms, is running on Victorian plumbing.

The Anatomy of a Broken Promise

Consider a hypothetical scenario—let us call her Clara, the owner of a mid-sized precision engineering firm in the West Midlands. Clara employs forty people. She buys a specialized German milling machine worth four hundred thousand pounds, and she does the responsible thing. She calls her broker to insure it.

What happens next is an invisible relay race of human misery.

Clara fills out a thirty-page proposal form. Her broker manually re-keys that data into three different portals to get quotes. At the insurance carrier, a junior underwriter—a smart twenty-five-year-old who studied history at university—downloads Clara’s PDF, opens an Excel spreadsheet that was built in 2008, and spends four hours cross-referencing safety certificates, flood maps, and historical claims data. Halfway through, the underwriter notices a missing detail about the factory’s sprinkler system. An email goes back to the broker. The broker emails Clara. Clara is on the factory floor and misses it for two days.

Multiply Clara by ten million businesses and a hundred million households.

The numbers behind this friction are staggering, even if the industry rarely advertises them. Across global commercial insurance, expense ratios—the sheer administrative cost of acquiring, underwriting, and servicing a policy before a single penny is paid out in claims—regularly hover between thirty and forty percent. Think about that. Out of every pound paid into the system to protect against catastrophe, roughly a third disappears into friction, administration, and manual data entry.

I remember sitting across from a veteran claims adjuster in a cramped coffee shop near Fenchurch Street a few years ago, watching him rub his temples as he described his Tuesday morning. He had spent four consecutive hours reading a single commercial property policy—a hundred and forty pages of dense, cross-referenced legal prose—just to determine whether a specific type of water ingress was covered under section four or excluded under appendix B. He was brilliant, empathetic, and utterly defeated by the sheer volume of unstructured text standing between a grieving business owner and a check.

"We don't have a risk problem," he told me, stirring a cold espresso. "We have a reading problem."

Why AI Failed Insurance the First Time

If artificial intelligence is the obvious cure for a reading problem, why hasn't it fixed insurance already?

Because for the past decade, the industry tried to solve a messy, nuanced human puzzle with rigid, brittle tools. First-wave "insurtech" startups arrived in the mid-2010s promising to disrupt everything. They built slick mobile apps with pastel colors and chatbots that worked brilliantly until a customer had a question that fell slightly outside a pre-programmed decision tree. When real-world complexity hit—a multi-vehicle pileup, a nuanced supply-chain interruption, a bespoke commercial liability policy—those early algorithms choked.

Large insurers spent hundreds of millions on "digital transformation" projects that amounted to putting a shiny web form on top of a mainframe computer from the 1980s. Beneath the surface, human beings were still copying and pasting text from one window to another.

It was like bolting a Formula One steering wheel onto a horse-drawn carriage and wondering why you weren't winning at Silverstone.

Then came the shift in modern generative AI and specialized machine learning—models capable not just of matching keywords, but of comprehending context, parsing messy unstructured documents, and reasoning across thousands of pages of regulatory and actuarial rules simultaneously.

This is the exact fault line where i23 was born.

What i23 Is Actually Building

Strip away the venture capital jargon, and the thesis behind i23 is deceptively simple. Insurance runs on unstructured data: emails, broker slips, engineering reports, medical records, legal contracts, and regulatory filings. Historically, turning that chaos into a structured decision—Should we insure this? At what price? Is this claim valid?—required armies of highly paid specialists doing repetitive cognitive labor.

i23 builds artificial intelligence designed specifically for the high-stakes precision of the insurance lifecycle. Instead of forcing insurers to rip out their legacy core systems—a multi-year nightmare that most chief executives would rather retire than endure—intelligent AI layers sit directly inside the workflow.

Imagine Clara’s milling machine again, this time running through an AI-native underwriting engine.

Clara sends over her factory documents, safety audits, and existing policy PDFs in whatever messy format she happens to have on her desktop. Within seconds, not weeks, the system ingests the documents, extracts the exact technical specifications of the German machinery, cross-checks the West Midlands postcode against real-time environmental and crime datasets, flags the missing sprinkler certification automatically, and drafts a tailored policy wording for the human underwriter to review.

The underwriter does not disappear. That is the crucial distinction that outsiders often miss. In high-value commercial insurance, human judgment remains irreplaceable—the gut instinct of a seasoned professional who knows when a factory owner is cutting corners on maintenance or when a tech startup’s liability exposure is hiding in a footnote.

What disappears is the drudgery. The underwriter stops being a data-entry clerk and finally becomes a risk strategist.

The Invisible Stakes of the Treasury Bet

Still, a skeptic might look at John Glen’s transition from Treasury minister to AI entrepreneur and raise an eyebrow. Politics and venture capital speak different languages. Westminster rewards caution, consensus, and the careful management of downside risk; Silicon Valley—and its London counterpart around Old Street and Shoreditch—rewards speed, obsession, and the willingness to break established norms.

Is a former minister truly equipped for the brutal, unforgiving pace of an early-stage tech company?

It is a fair doubt. Many political figures who enter the startup ecosystem serve merely as decorative advisors—names printed on a pitch deck to impress institutional investors and open doors at regulatory agencies. They rarely stay in the building when the servers crash on a Sunday night or when an enterprise client demands a custom integration by Friday.

Yet Glen’s specific background carries an unusual weight in this particular sector. Insurance is not like social media or food delivery. You cannot "move fast and break things" when you are holding the pension funds, medical liabilities, and disaster recovery reserves of an entire nation. Every line of code that touches an underwriting decision or a claims payout sits directly under the microscope of the Financial Conduct Authority and the Prudential Regulation Authority.

When an insurer adopts AI, its biggest fear is not whether the software works. Its biggest fear is the "black box" problem. If an algorithm rejects a small business’s insurance application or reduces a claim payout, and the regulator asks why, saying "the neural network decided" is a fast track to a massive fine and a front-page scandal.

Having spent years overseeing those very regulators, Glen understands something that pure technologists often learn too late: in financial services, trust is the only currency that scales. Explainability, auditability, and regulatory alignment are not bureaucratic hurdles to be cleared after the product is built. They are the product itself.

The Silent Race for London’s Future

There is a quieter, larger drama playing out beneath the millions raised for i23, one that stretches far beyond a single company’s balance sheet.

For over three centuries, London has been the undisputed capital of global risk. When a satellite is launched in French Guiana, when a container ship navigates the Strait of Malacca, or when a film studio insures a blockbuster production in Los Angeles, the risk almost always finds its way back to a glass tower in EC3.

That dominance is no longer guaranteed. New York, Bermuda, Singapore, and Zurich are fighting aggressively for market share. Worse, the nature of risk itself is mutating faster than traditional actuarial tables can track. Cyberattacks cascade across global supply chains in milliseconds. Extreme weather events—wildfires, flash floods, heat domes—are rendering fifty years of historical climate data obsolete.

If London’s insurance market clings to manual spreadsheets and six-week turnaround times, it will slowly become a museum of financial history—admired for its heritage, bypassed for its speed.

To survive, the industry has to compress time. It has to turn weeks of administrative friction into minutes of transparent, auditable clarity. And when that happens, the savings do not just pad corporate profit margins; they flow outward into the real economy. Cheaper, faster insurance means a young founder can afford liability coverage to launch a robotics company. It means a coastal town can still insure its high street shops. It means a family waiting to rebuild after a midnight flood gets a decision before the plaster has even dried.

Think back to that quiet hiss of water behind the drywall at two in the morning.

Nobody lies awake wishing for better enterprise software or dreaming about regulatory frameworks on Horse Guards Road. You just want to know that when the worst happens, the promise you paid for holds firm—without a fight, without a labyrinth of paper, and without waiting half a year for a human being to finish reading the fine print.

The machines are finally learning to read the contracts. Now comes the real test: making sure they never forget who the contracts were written to protect.

JH

Jun Harris

Jun Harris is a meticulous researcher and eloquent writer, recognized for delivering accurate, insightful content that keeps readers coming back.