10 Years Building Vertical Software: My Perspective on the Selloff

By Nicolas Bustamante (@nicbstme)

In the past few weeks, nearly $1 trillion was wiped from software and services stocks. FactSet dropped from a $20B peak to under $8B. S&P Global lost 30% in weeks. Thomson Reuters shed almost half its market cap in a year. The S&P 500 Software & Services Index — 140 companies — fell 20% year to date.

Wall Street called it a panic. I've spent the last decade building vertical SaaS. First at Doctrine, now the largest legal information platform in Europe, and then at Fintool, an AI-powered equity research platform in the US that competes directly with Bloomberg, FactSet, and S&P Global today.

I built the kind of software that LLMs are now threatening. And I'm now building the kind of software that's doing the threatening. I've been on both sides of this disruption.

Here's what I see: LLMs are systematically dismantling the moats that made vertical software defensible. But not all of them. The result is a redrawing of what makes vertical software valuable — and the multiple it deserves.

The Ten Moats of Vertical Software (and What LLMs Do to Each)

Vertical software is software built for a specific industry. Bloomberg for finance. LexisNexis for legal. Epic for healthcare. Procore for construction. Veeva for life sciences.

These companies share a defining characteristic: they charge a lot and customers rarely leave. FactSet charges $15,000+ per user per year. Bloomberg Terminal costs $25,000 per seat. LexisNexis charges law firms thousands per month. Retention rates hover around 95%.

There are ten distinct moats. LLMs are attacking some while leaving others intact. Understanding which is which is the entire game.

1. Learned Interfaces → Destroyed

A Bloomberg Terminal user has spent years learning keyboard shortcuts, function codes, and navigation patterns — GP, FLDS, GIP, FA, BQ. These aren't intuitive. They're a language. Once you speak it fluently, switching means becoming illiterate again.

"We're a FactSet shop." "We're a Lexis firm." "We're a Bloomberg house." These aren't statements about data quality. They're statements about software muscle memory. People have spent a decade learning the tool. That investment isn't transferable.

This was the most under-appreciated moat. Knowledge workers pay to not relearn a workflow they've spent a decade mastering. The interface IS a big part of the value prop.

Vertical software providers maintain armies of designers and customer success managers whose entire job is onboarding customers onto their interface. Every UI change is a project: user research, design sprints, careful rollouts, handholding. The interface wasn't a feature. It was the product.

At Fintool, we have no onboarding. No CSMs teaching navigation. Our users type what they want in plain English and get an answer — because it's what they're already used to with ChatGPT. There is no interface to learn because it's all chat. That entire cost center — designers, CSMs, UI change management — just doesn't exist. The chat interface absorbed all those scaffoldings.

LLMs collapse all proprietary interfaces into one: Chat.

Consider what a financial analyst does today on a Bloomberg Terminal: navigate to the equity screening function, set parameters using specialized syntax, export results, switch to the DCF model builder, input assumptions, run sensitivity analysis, export to Excel, build a presentation. Each step requires learned interface knowledge. Each step reinforces switching costs.

Now consider the same analyst with an LLM agent:

"Show me all software companies with over $1B market cap, P/E under 30, and revenue growing over 20% year over year. Build a DCF model for the top 5. Run sensitivity analysis on discount rate and terminal growth."

Three sentences. No keyboard shortcuts. No function codes. No navigation. The user doesn't even know which data provider the LLM queried. They don't care.

When the interface is a natural language conversation, years of muscle memory become worthless. The switching cost that justified $25K per seat per year dissolves.

2. Custom Workflows and Business Logic → Vaporized

Vertical software encodes how an industry actually works. A legal research platform doesn't just store case law — it encodes citational networks, Shepardize signals, headnote taxonomies, and the specific way a litigation associate builds a brief.

This business logic took years to build. It reflects thousands of conversations with domain experts. The hardest part was often not the technology — it was understanding how customers actually work.

LLMs turn all of this into a markdown file.

Traditional vertical software encodes business logic in code: thousands of if/then branches, validation rules, compliance checks, approval workflows — hardcoded by engineers over years. Not just any engineers. You need people who can write production code AND understand the domain, which is rare.

At Fintool, we have a DCF valuation skill. It tells an LLM agent how to do a discounted cash flow analysis: which data to gather, how to calculate WACC by industry, what assumptions to validate, how to run sensitivity analysis, when to add back stock-based compensation. It's a markdown file. Writing it took a week. Updating it takes minutes. A portfolio manager who's done 500 DCF valuations can encode their entire methodology without writing a single line of code.

Years of engineering versus one week of writing. That's the shift.

And it's not just speed. The markdown skill is better in important ways: it's readable by anyone, it's auditable, it can be customized per user, and it gets better automatically as the underlying model improves — without touching a line of code.

Business logic is migrating from code written by specialized engineers to markdown files that anyone with domain expertise can write. The accumulated business logic that took vertical software companies a decade to build can now be replicated in weeks.

3. Public Data Access → Commoditized

A massive portion of vertical software's value proposition was making hard-to-access data easy to query. FactSet makes SEC filings searchable. LexisNexis makes case law searchable. These are genuine services — SEC filings are technically public, but try reading a 200-page 10-K in raw HTML.

Before LLMs, accessing this public data required specialized software and significant engineering scaffolding. Companies like FactSet built thousands of parsers — one for each filing type, each company's idiosyncratic formatting. Armies of engineers maintained these parsers as formats changed.

At Fintool, we built none of that. Zero NER. Zero custom parsers. Zero industry-specific classifiers. Why? Because frontier models already know how to navigate a 10-K. They know that Home Depot's ticker is HD. They understand the difference between GAAP and non-GAAP revenue. They can parse a nested table of segment disclosures without being taught the schema.

The model IS the parser.

Feed it a 10-K and it can answer any question about it. The parsing, structuring, and querying that vertical software spent decades building is now a commodity capability baked into the foundation models themselves. The data isn't worthless — but the "making it searchable" layer, where much of the pricing power lived, is collapsing.

4. Talent Scarcity → Inverted

Building vertical software requires people who understand both the domain and the technology. Finding an engineer who can write production code AND understands how credit derivatives are structured is extremely rare. This scarcity created a natural barrier to entry that historically limited serious competitors in any vertical.

LLMs flip this moat entirely.

In any vertical, hiring was brutal. You didn't just need good engineers — you needed engineers who could understand the domain: how change orders cascade through a construction project schedule, how lien waivers interact with payment applications, what triggers a delay claim under AIA contract terms. These people barely existed. Vertical software companies built their own through months of internal training.

With LLMs, a domain expert who has never written a line of code can now describe a workflow in plain language and have it executed by an AI agent. The rare hybrid talent that was the bottleneck to building vertical software is no longer the bottleneck. Domain expertise, which exists in abundance in every industry, has become the primary input. The scarcity that protected incumbents has inverted into an advantage for outsiders who have domain knowledge but lack engineering depth.

Why the Selloff Is Structurally Justified but Temporally Exaggerated

The market is right that the moats are eroding. It may be wrong about the speed.

Enterprise software has long replacement cycles. Contracts run 3–5 years. Procurement processes are slow. IT departments are risk-averse. Legal and compliance reviews add months. Even if an LLM-native competitor delivers 10x the value at 10% of the cost, the incumbent still has 2–3 years of contracted revenue.

The structural threat is real. The timeline the market is pricing may be compressed. This creates an interesting asymmetry — the selloff is right in direction, potentially wrong in magnitude and timing.

The Real Threat (It's Not What You Think)

The narrative is that startups will eat vertical software incumbents. That's partially true but misses the deeper threat.

The deeper threat is that the category itself may compress.

When business logic lives in markdown files, when interfaces dissolve into chat, when parsing is free — the question isn't just "who wins the category?" It's "how large does the category remain?"

A Bloomberg Terminal at $25K/seat exists partly because of genuine data value and partly because of interface lock-in, workflow encapsulation, and parsing infrastructure. Strip out the artificial value and you're left with the genuine value: proprietary data, unique datasets, trusted calculations, compliance-grade outputs.

That genuine value is real but smaller. The premium on top of it — the interface tax, the workflow tax, the parsing tax — is what's being competed away. The total addressable market for vertical software may shrink even as the best players survive.

What Replaces Vertical Software

Three things emerge as the new defensible layers:

1. Proprietary data that LLMs cannot replicate. Real-time pricing feeds. Exclusive licensing agreements. Data generated by the software itself through network effects. If the data doesn't exist in any LLM's training set and can't be scraped, it retains value.

2. Trust and compliance infrastructure. In regulated industries — legal, finance, healthcare — outputs need to be auditable, explainable, and defensible. The compliance layer around AI outputs becomes a new moat. Who can certify that the AI's analysis meets regulatory standards?

3. Workflow orchestration at scale. The new competitive advantage isn't encoding the workflow — it's orchestrating hundreds of AI agents running thousands of workflows simultaneously, with the right guardrails, integrations, and audit trails. The platform that makes this manageable wins.

What Comes Next

The vertical software industry is entering a period of violent repricing. Companies with genuine proprietary data will find floors. Companies whose value was primarily interface lock-in and workflow encapsulation will face sustained pressure.

The winners will be those who recognize which of their moats are real and which are legacy artifacts of a pre-LLM world — and who move fast enough to rebuild around what actually remains defensible.

For founders and investors, the question is no longer "can you build a vertical software company?" It's "which layer of the vertical software stack are you actually competing in — and does that layer survive?"

The selloff is a signal worth taking seriously. Not as a prediction of imminent collapse, but as a structural reset of what vertical software is worth — and why.