The Last Undigitized Industry — And Who's Fixing It

Every major consumer industry of the past two decades has gone through the same transformation. In each case, a manual, relationship-driven business collided with cloud infrastructure and was rebuilt from the ground up, displacing the phone call, the paper trail, and the human intermediary that had held the whole operation together.

But finance, specifically institutional trading, has remained a relatively stubborn holdout, still running on bespoke data formats and on-premises servers that predate the smartphone.

For engineer Sangjun Yum, the under-digitization of financial markets is a set of concrete engineering challenges that have gone unsolved for too long, locking everyday investors out of products and services that a modernized infrastructure could open up to them. As a senior software engineer at FalconX, a crypto prime brokerage serving institutional clients, he's working on one of the most tangible expressions of that thesis: replacing the trading desk phone call with automated execution infrastructure built for the way markets actually move.

Why Finance Hasn't Been Modernized According to Sangjun Yum

The comparison Yum draws between finance and the industries that came before it is not rhetorical. It's structural. Industries like ride-sharing, food delivery, media streaming, and e-commerce all followed the same arc from high-touch, manual operations to automated digital platforms. Uber rebuilt the infrastructure around an industry that had been running on phone dispatchers and cash transactions. DoorDash digitized a business that had been running on paper menus and landline orders.

Finance, in Yum's framing, isn't a permanent exception to this pattern; rather, it's simply the next industry in line.

The reason for why it hasn't happened just yet, in his analysis, is multifaceted. Financial technology is deeply niche, largely on-premises, and not architected for the cloud environments that enabled distributed computing and, eventually, artificial intelligence everywhere else. Every other major industry that digitized before finance had incumbent relationship networks, regulatory considerations, and entrenched manual processes that made automation seem impractical, until it suddenly wasn't. The same conditions exist in finance today.

"I just found that traditional finance, especially on the trading floor, is not digitized nearly as much," Yum says. "I think that's the kind of problem Silicon Valley should be tackling."

Trading, particularly at the institutional level, still fits that pre-disruption profile almost exactly. When a hedge fund wants to execute a large Bitcoin order, placing it directly on a retail exchange like Coinbase is not a realistic option. The price slippage on a $1 million notional trade would be severe enough to render the execution economically irrational. Instead, the fund calls a trader at its brokerage. That trader manually routes the order across a network of liquidity partners.

The human in the middle is, in this case, the standard operating procedure. And for Yum, it's precisely the kind of problem that has already been solved everywhere else.

The Infrastructure That Never Got Built

To understand why finance has lagged, Yum points to a sequencing problem that most observers overlook. A technology like AI didn't emerge in isolation: it was enabled by an underlying system of cloud infrastructure, which itself was built over two to three decades by companies like Amazon. Distributed web applications became possible because the underlying cloud layer existed to support them, and AI followed from there.

Finance has, so far, skipped that transition.

"A lot of technologies in the finance world are really, really niche and not really geared toward the cloud," Yum says. "I think that's one of the core things that is blocking the finance world from coming into a more digitized world."

The practical consequence of this gap became concrete for Yum during his time at SIMON Markets, a Goldman Sachs spin-off that built a centralized distribution platform for structured investment products. Because of its origins, SIMON ran on Goldman's on-premises server infrastructure rather than standard public cloud environments. The gap between what Goldman's internal deployment systems could do and what an AWS-native architecture would have allowed shaped what could and couldn't be built, and Yum saw it firsthand.

Financial institutions looking to adopt AI tools face a version of that same wall. The tools themselves were designed to run on cloud infrastructure that most banks and brokerages have never built. Before the application layer can follow, the underlying infrastructure has to be constructed, and for much of the financial industry, that work is still at its earliest stages.

Decades of compressed catching-up, in Yum's view, stand between where most institutions are today and the kind of fully modernized environment that would let them participate in the same AI-driven future that has already reshaped every other major industry.

The Structured Products Problem

The clearest illustration of what under-digitization costs comes from the market for structured products, which are complex financial instruments that bundle together features like principal protection, dividend caps, and market-linked returns into a single security. These include products like covered-call baskets and market-linked certificates of deposit, which function more like investment contracts than traditional savings accounts.

These instruments are among the most widely used retirement vehicles for high-net-worth investors, yet accessing them required, until recently, physical meetings, manual prospectus reviews, and bilateral relationships maintained separately with each issuing bank.

The reason those manual processes persisted was not purely regulatory. There was no shared data standard for structured products. Every issuer, whether big corporations like JPMorgan and HSBC or smaller boutique investment banks, defined the product lifecycle differently, with no common schema or interoperable format to communicate with each other.

SIMON's attempt to solve this meant building a platform that could accept product data from dozens of issuers simultaneously, each with different requirements for how their products were structured and presented. Yum was among the earliest engineers on that backend team, and the core technical challenge fell to him: designing a unified structured product marketplace and portfolio management system that could accommodate radically different product structures without forcing any single issuer to rebuild their own data formats from scratch.

"Fitting our product to our issuers was really, really hard, as a lot of issuers were basically coming up with different requirements for us to upload their products," Yum recalls.

The parallel to earlier platform shifts is exact. Before DoorDash could automate restaurant delivery, someone had to solve the problem of digitizing restaurant menus into a common format. Before SIMON could automate structured product distribution, the underlying data had to be standardized. The infrastructure problem always precedes the application problem.

What Digitization Unlocks

As a current senior software engineer at FalconX, Yum believes the company represents the current frontier of Yum's argument in practice. FalconX's platform is designed to fully automate the execution of institutional crypto trades that, today, still depend heavily on human traders making phone calls and manually routing orders across a brokerage's liquidity network.

Basically, when a hedge fund places a large order, the old process required a trader to pick up the phone, work the order manually, and hedge the exposure across however many liquidity partners the brokerage maintained relationships with. FalconX's infrastructure eliminates that intermediary step.

What most retail observers don't realize is how much of the crypto market depends on this invisible layer. Retail platforms net out their users' positions internally, managing their own books to stay as neutral as possible, and then hedge their residual crypto exposure through the over-the-counter network rather than routing it onto public exchanges like Coinbase or Binance. FalconX sits at the center of that process as one of the primary counterparties handling that flow, a layer of market infrastructure that the average retail investor never sees and rarely considers.

Removing human intervention from trades of this scale carries consequences across pricing, execution speed, and systemic risk, the same categories of improvement that made automated market-making transformative in equities a decade ago. Just as algorithmic trading displaced the open-outcry pit, FalconX's infrastructure is displacing the trading desk phone call. The technology required to do this isn't new; it's the application of it to this particular corner of the market that has changed.

The people building this infrastructure are software engineers, and Yum is clear-eyed about what that shift means. Digitization at this layer does not just improve operational efficiency. It creates the conditions for finance to finally receive the same AI-driven tools that have already transformed every other major industry.

A Vision for Finance's Future Platform Shift

The same pattern that reshaped taxis, restaurant delivery, and media streaming is now working its way through the last major industry that resisted it. Sangjun Yum's work, from standardizing structured product data at SIMON Markets to automating institutional trade execution at FalconX, represents what that process looks like at the engineering level, where the abstract thesis of financial digitization gets translated into concrete infrastructure problems that someone has to solve.

The cloud layer has to be built before the applications can follow, the data has to be standardized before the platforms can scale, and the manual trade has to be automated before the phone call finally stops being the most reliable tool on the desk.