2026-01-23 04:46:55
Context graphs have become the new battleground in enterprise software. @JayaGup10 and @ashugarg argued that the next trillion-dollar platform opportunity isn’t in systems of record. It’s in capturing the decision traces that systems of record miss. That’s true, but it’s missing the key point that vertical software companies have been building these context graphs for over a decade.
A context graph is a queryable record of business logic: the reasoning, precedents, and decision traces that explain why things happened, not just what happened. Every company has one in theory. It’s the accumulated knowledge of how the business operates: the exceptions that get approved, the precedents that govern decisions, the tribal knowledge in people’s heads.
Agents need this to move from automation to autonomy. An agent can run a workflow, but it can’t handle exceptions or apply precedent without access to the reasoning behind past decisions. The context graph separates an agent that follows rules from one that exercises judgment.
In most companies and products, the context graph exists in theory but not in practice. It’s scattered, implicit, and inaccessible because:
No one logs the reasoning behind decisions. The VP approved the exception on a Zoom call but never recorded why.
Systems capture outcomes, not context. The CRM shows the final discount, not the service outages or churn threat behind it.
What context exists is siloed. It’s scattered across tools that don’t share a worldview.
These aren’t product bugs. They’re structural features of horizontal SaaS. Generic platforms use flexible abstractions that can model any business, which means they model no business precisely. Humans must bridge the gap between “what the system captures” and “how decisions get made.” But humans don’t leave audit trails.
The trillion-dollar question: who fixes this? The prevailing thesis is that agent startups have a structural advantage. They sit in the execution path at decision time, so they can capture context that systems of record never see. The assumption is that the context graph needs to be built from scratch. But that’s not entirely true.
The debate has a blind spot: it focuses entirely on horizontal enterprise software.
Think about the workflows and software stack of a typical company. It’s a patchwork of horizontal tools, each built for a broad use case, and none designed to really work together. Sure they have APIs and integrations, but even the best integrations drop tons of context as data and actions move through them. The context graph fragments because the tools don’t share a worldview.
But vertical software is different. The difference starts with something I’ve written about before: the data model.
Horizontal platforms like Salesforce use generic abstractions (”accounts,” “contacts,” “opportunities”) that can model almost any business. That flexibility helps breadth but hurts depth. The ontology doesn’t map to how any particular industry works. It’s a blank canvas customers must configure into meaning.
Vertical software starts from the opposite premise. The data model isn’t flexible. It’s opinionated, purpose-built, and mapped to that industry’s reality.
Look at Toast’s data model. You’ll see objects like Order, MenuItem, Customer, PrepStation, etc. These aren’t generic transactions dressed up with custom fields. They’re first-class entities with built-in relationships: orders connect to customers, menu items, locations, and payments as native concepts. The ontology is the domain.
Compare that to Salesforce’s data model: Account, Opportunity, Lead, Contact, etc. Powerful abstractions, but they describe a law firm, restaurant, or rocket manufacturer equally well. They describe none of them well. Configuration, integrations, and human memory must bridge the gap.
In vertical software:
The reasoning gets recorded because the system is built around actual business decisions. When a construction platform tracks change orders, it captures why the change happened (weather delay, subcontractor default, scope change) as structured data, not notes in a generic “activity” field.
The context isn’t siloed because vertical platforms consolidate functions into one system. The construction platform isn’t just project management; it’s also scheduling, procurement, billing, and field operations. One system, one data model, one view of reality.
The full decision trace is reconstructible because the ontology supports it. When everything from estimate to invoice lives in a system built around menu items or job sites or patient treatments, you can trace how a price changed, why a timeline slipped, or what drove a discount.
There’s a deeper point about how context graphs work. The original context graph thesis assumes decision traces need explicit capture: an agent logs “I did X because of Y” at the moment of decision. But there’s another way the “why” becomes accessible: through inferential density. When the graph is sufficiently rich, the reasoning can be reconstructed from relationships between nodes.
Consider an order in Toast. Alone, it’s just a transaction with attributes. Connect it to inventory and you know which items are selling faster than expected and what needs restocking. Connect it to customer data and you know who your most valuable customers are, what they order, and when. Connect it to payments and you know revenue by payment method and which orders are outstanding.
Now when you see a spike in refunds, you don’t need someone to log what went wrong. You can see that the spike correlates with a specific menu item, the inventory system shows that ingredient was substituted due to a stockout, and the affected customers skew toward your highest-value segment. The reasoning is embedded in the relationships.
Each new function adds inferential power to everything already in the graph. Inventory data makes orders more meaningful. Customer data makes inventory decisions more meaningful. Payment data makes customer relationships more meaningful. The context graph doesn’t just get wider; it gets denser. Density makes the “why” reconstructible without explicit capture.
This is the structural advantage vertical platforms carry into the age of agents.
Vertical software companies didn’t build these context graphs by accident. They were forced to go deep because they couldn’t go broad. A construction platform or dental practice system faces a ceiling: there are only so many construction companies, only so many dental practices. If you can’t grow by adding customers across industries, you grow by capturing more value from each customer within your industry.
Embedded payments was the proof point. When a vertical platform embeds payments, such as with Rainforest, it isn’t just adding a monetization layer. It’s extending the context graph into new categories. The platform knows not just what transactions happened, but the full economic reality: payment velocity, cash flow patterns, default rates. Every transaction adds to the accumulated knowledge.
The same logic now drives the next wave. Vertical platforms have spent a decade consolidating the operational core. But activity at the edges (customer acquisition, purchasing, workforce management) still happens outside the platform. Marketing in Mailchimp. Procurement in email threads. Accounting in QuickBooks. Each represents a gap in the graph.
The foundation exists. Vertical platforms have domain-specific ontologies, core operational modules, and deep customer relationships. They understand what a job site is, what a patient treatment plan looks like, what a menu item costs. This is the precise business logic edge functions need to plug into.
The financial layer is in place. Embedded payments gave vertical platforms visibility into the money moving through their customers’ businesses. That financial context is essential for procurement and accounting, which live at the intersection of operations and money.
AI changes the economics. Marketing, procurement, and bookkeeping require labor. They’re judgment-intensive and context-dependent. AI can now handle workflows that previously required humans. The vertical platform, with its deep context graph, is the ideal substrate.
A few categories we’re tracking here:
Marketing pulls customer acquisition into the graph. Reach routes customer communications and other marketing data through the vertical platform itself, connecting acquisition cost to the customers it produced.
Accounting pulls financial categorization into the graph. With Layer or Tight, transactions get categorized within the platform rather than exported to QuickBooks, closing the loop between operations and the books.
Procurement pulls purchasing into the graph. Companies like Sticker and Faliam bring supplier relationships out of email threads and into the system of record, making spend visible alongside operations.
Each follows the same pattern: absorb an adjacent function, blend context with financial data, deploy AI, and deepen the graph.
Horizontal SaaS will lose ground. These companies have brand equity, data gravity, and deep workflow integration. But they’ll lose share to agent-first startups smart about bootstrapping context graphs, and to vertical incumbents picking off customers tired of stitching together horizontal solutions.
Agent-first startups face a cold start problem. The context graph thesis is right that agents can capture decision traces. But vertical software already owns the workflow for millions of businesses. An agent startup building for restaurants competes with Toast’s distribution, data, and decade of accumulated context. A hard gap to close.
Vertical software incumbents are undervalued. The market prices vertical software on revenue and growth. It doesn’t price the context graph. As horizontal SaaS struggles and agent startups face distribution challenges, vertical platforms with dense context graphs have a structural advantage the market undervalues.
Embedded service providers are an emerging category worth watching. Not every vertical platform will build the graph extensions in house. Embedded solutions will be more attractive. marketing, procurement, and accounting in-house. Their growth is complementary to vertical software’s expansion. As vertical platforms absorb more functions, the infrastructure providers that enable it become more valuable.
The trillion-dollar context graph opportunity is real. But it won’t be captured by whoever builds the best agent. It will be captured by whoever already has the deepest graph and knows how to extend it.
My name is Matt Brown. I’m a partner at Matrix, where I invest in and help early-stage fintech and vertical software startups. Matrix is an early-stage VC that leads pre-seed to Series As from an $800M fund across AI, developer tools and infra, fintech, B2B software, healthcare, and more. If you’re building something interesting in fintech or vertical software, I’d love to chat: [email protected]
2026-01-13 03:15:44
Risk, like energy, cannot be destroyed: only transferred or transformed.1 Call it the thermodynamics of financial services.
Every financial transaction involves risk: uncertain exposure that someone must bear. A loan might not be repaid. A payment might be fraudulent. A counterparty might fail. Someone is always holding this exposure.
Risk is balanced by cost. Interest rates on loans. Interchange fees on payments. Spread on foreign exchange. You take the risk, you get paid for it.
Risk is also balanced by friction. The three-day settlement window. The paper application. The in-branch visit. These delays make risk more manageable. Friction is another form of cost (implicit rather than explicit, paid in time and conversion rather than dollars).
Fintechs reduce both cost and friction. Lower fees. Faster approvals. Seamless onboarding. The pitch is simple: same product, better experience.
But reducing cost and friction doesn’t reduce risk. That’s the thermodynamics at work. The risk remains, conserved in the system. The friction you removed was managing something. Where does that exposure land?
Risk takes many forms, including:
Each requires different capabilities to manage. Being good at credit risk doesn’t make you good at fraud risk. Operational excellence doesn’t help with concentration.
This creates an opportunity. A process change, a new data source, or a different product structure can shift risk from one form to another. A company with a specific edge in data, technology, or operational discipline can deliberately transform risk into a form they’re better equipped to manage. The goal isn’t to avoid risk. It’s to hold the right risk.
The difference between success and failure is whether this transformation is intentional. Reducing cost and friction without understanding the new risk you’re holding is how fintechs blow up. The ones who win recognize the transformation and build specific capabilities to handle it.
In each of these examples, notice the trade: what friction came out, what risk moved in, and what capability the winner built.
Square (underwriting risk → fraud and chargeback risk)
Traditional payment processors manage risk through friction: site visits, reserve accounts, manual underwriting. The merchant is vetted before they ever process a transaction. Square removed that friction, letting anyone accept payments instantly. But the risk didn’t disappear; it transformed. They won because they built real-time detection systems designed for fraud and chargebacks. They deliberately traded a risk managed through slow process for a risk they could manage through technology.
Toast (credit risk → operational and concentration risk)
Traditional lenders manage credit risk through friction and cost: lengthy applications, document collection, credit checks, high interest rates. Toast removes that friction for restaurants on their platform. A restaurant can get capital with minimal paperwork, underwritten on real-time transaction data flowing through its point-of-sale system. Concentration risk cuts both ways: their entire loan book is restaurants, so when COVID hit, the whole portfolio was correlated. But focusing on a single sector means deeper knowledge, purpose-built software, and better underwriting for that specific business. They win when depth of insight outweighs concentration exposure.
Buy now, pay later (consumer credit cost → provider credit risk)
Traditional card payments balance risk through cost: merchants pay interchange, consumers pay high APRs if they revolve. BNPL changes the equation. Merchants pay a higher fee (4-8% vs 1.5-3%) but shed chargeback exposure and gain conversion lift. Consumers get interest-free financing. The BNPL provider absorbs credit risk that was previously managed with high APRs. Companies like Affirm and Klarna are betting they can manage this through short durations, transaction-level underwriting, and merchant-funded economics. They win when underwriting keeps losses below the merchant fee. They lose when credit losses spike and the math stops working.
If you’re building a fintech that reduces cost or friction, ask yourself: what risk are you now holding? The friction you removed was managing something. Do you have data, technology, or expertise that makes you better at managing the new form? Or are you just hoping it doesn’t materialize?
Risk can’t be destroyed. But it can be transformed, and the best fintechs turn that transformation into their advantage.
My name is Matt Brown. I’m a partner at Matrix, where I invest in and help early-stage fintech and vertical software startups. Matrix is an early-stage VC that leads pre-seed to Series As from an $800M fund across AI, developer tools and infra, fintech, B2B software, healthcare, and more. If you’re building something interesting in fintech or vertical software, I’d love to chat: [email protected]
For the physics-minded: diversification and better information can reduce risk at the portfolio level (that’s the point of insurance). But compression in one dimension creates exposure in another. Diversify across a thousand borrowers and you’ve traded idiosyncratic credit risk for systematic risk and operational risk. The risk isn’t gone; it’s reshaped. Squeeze it in one place and it bulges out somewhere else.
2025-10-15 00:49:21
Product market fit is the startup holy grail. “Product” and “market” are essential, but a startup’s data model is the dark matter that holds them together.
“Data model” refers to what a startup emphasizes in its product, i.e., which parts of reality matter most in how the product represents the world. It’s the core concepts or objects a startup prioritizes and builds around, the load-bearing assumptions at the heart of their strategy and worldview. It’s partially captured in the database architecture (hence the name), but it shapes everything from the UI/UX to the product marketing, pricing model, and GTM strategy.
This shows up differently depending on the layer. In the database, it’s which tables are central and how they relate. In the product, it’s which UI elements dominate and what actions are easiest. In pricing, it’s what you charge by. In GTM, it’s the workflow or pain point you lead with. But they all stem from the same choice about what deserves to be the center of gravity.
Every founder has a data model, whether they realize it or not. Either you choose it explicitly or it gets inherited from whatever you’re copying. Most founders never articulate it. By the time the architecture solidifies around these implicit choices, it’s nearly impossible to change.
And that’s generally fine, because most companies shouldn’t innovate on their data model. Customers have existing mental models and workflows built around incumbent tools. Fighting that is expensive and slow. But at the extreme ends of markets—where you’re toppling multi-billion-dollar incumbents or creating entirely new categories—a distinctive data model becomes a critical and non-obvious edge.
The biggest breakout companies of the last decade often trace their success to an early, non-obvious data model choice that seemed minor at the time but proved decisive. Consider:
Slack’s persistent channels vs 1:1/group messages: While Yammer and HipChat replicated email’s ephemeral group messages, Slack made persistent, searchable channels the atomic unit. This created organizational memory—every decision, discussion, and document lives forever in context. Incumbents couldn’t match this without rebuilding from scratch.
Toast’s menu-item-centric architecture vs generic POS SKUs: Toast makes menu items first-class objects with embedded restaurant logic—prep times, kitchen routing, and modifier hierarchies built in. Generic point-of-sale systems treat menu items as retail SKUs, requiring third-party integrations for kitchen workflows. Toast’s model enables native order routing and real-time kitchen management, plus natural extensions like ingredient-level inventory and prep-based labor scheduling—creating a locked-in ecosystem that becomes the restaurant’s operational backbone.
Notion’s blocks vs Google’s documents: Google Docs gives you documents; Notion gives you Lego blocks. Every piece of content can be rearranged, nested, or transformed into databases, kanban boards, or wikis. This modularity collapses entire tool categories into one system. Traditional tools can’t compete without abandoning their document-centric architecture.
Figma’s canvas vs files: Photoshop and Sketch are built on local files. Figma is built on a shared web canvas where everyone sees changes instantly. This eliminates version conflicts and “final_final_v2” chaos. Adobe couldn’t respond without deprecating their entire desktop-first ecosystem.
Rippling’s employee data model vs siloed tools: Rippling treats the employee record as the lynchpin connecting HR, IT, payroll, and finance. Not separate products sharing data, but one product with multiple views. Each new product module is automatically more powerful than standalone alternatives because it inherits full employee context. Competitors remain trapped in single categories or attempt inferior integrations.
Klaviyo’s order-centric data model vs email-centric tools: MailChimp optimizes for email campaigns. Klaviyo optimizes for customer lifetime value by making order data a first-class citizen alongside emails. This lets e-commerce brands segment by purchase behavior, not just email engagement. Generic email tools can’t match this without rebuilding for vertical-specific data.
ServiceNow’s connected services vs standalone tickets: Traditional help desks treat tickets like isolated emails. ServiceNow links every ticket to a service map—showing which system is down, who owns it, and what it affects downstream. This transforms IT from ticket-closing to problem-preventing, making ServiceNow irreplaceable once companies reorganize operations around this model.
The importance of a differentiated data model is rising dramatically. AI is commoditizing code. Technical execution is table stakes rather than a competitive advantage. AI can generate code, but it can’t refactor the organizational reality customers have built around your architecture—the workflows, integrations, and institutional muscle memory that compound over time.
Meanwhile, many markets have become so crowded that single-product companies can’t survive. This is particularly true in vertical markets, where companies are expanding into adjacent software products, embedding payments and other financial products, and even competing with their customers’ labor and supply chains with AI and managed marketplaces.
This all points to the same conclusion: when code is cheap, competition is fierce, and vertical depth matters, your data model is the foundation of your moat. The companies that win won’t be those with the most or even the best features. AI will democratize those. The winners will be built on a data model that captures something true about their market, which in turn creates compounding advantages competitors can’t replicate.
Consider how this plays out. Rippling’s employee-centric model made it trivial to add payments, benefits, and spend management. Each new product inherits rich context, making it instantly more powerful than standalone alternatives. Toast’s menu-item architecture naturally extended to inventory, labor, and supplier management. The data model wasn’t just their first product decision. It was their platform destiny.
The path to a differentiated data model depends on your market. The more horizontal you go, the more your moat comes from technical and interface innovation. The more vertical you go, the more your moat comes from elevating the right domain objects with the right attributes.
Horizontal tools serve broad use cases where underlying concepts are already familiar. Leverage comes from changing how the product is built or experienced. Notion reimagined documents as composable blocks. Figma rebuilt the foundation entirely as a multiplayer web canvas.
Vertical tools serve specific industries with deep domain complexity. Leverage comes from what you choose to emphasize. Toast elevated menu items—not transactions—with prep times and kitchen routing as first-class data. Klaviyo promoted order data to equal status alongside email metrics.
A good place to start is by looking for model mismatches in existing successful products. Where are incumbent products forcing an incorrect or outdated model on their customers? Where are customers using workarounds—spreadsheets, low/no code tools, extensive in-product configuration—to make the product match how they think and work?
Despite all the emphasis on data models, start with the workflow, not the technical implementation. Don’t ask “what data do we need to store?” Ask “what’s the atomic unit of work in this domain?” For restaurants it’s the menu item. For design it’s the canvas. For employee operations it’s the human.
If you’ve already built a product, you can audit how powerful and correct your data model is. Open your database schema and see which table has the most foreign keys pointing to it. Is that the atomic unit your customers actually think in? List your product’s core actions. Do they all strengthen one central object, or are you building a feature buffet? What would break if you deleted your second-most important table? If the answer is “not much,” you probably have the wrong data model.
Test whether your data model creates compound advantages. When you add a new feature or product, does it automatically become more powerful because of data you’re already capturing? If your answer is “we’d need to build that feature from scratch with no inherited context,” you don’t have a compounding data model—you have a product suite. The right model creates natural expansion paths that feel obvious in retrospect but were invisible to competitors.
Your data model is your destiny. The paradox is that this choice happens when you know the least about your market, but that’s also why it’s so powerful when you get it right. Competitors who’ve already built on different foundations can’t simply copy your insight. They’d have to start over, and by then, you’ve compounded your advantage.
My name is Matt Brown. I’m a partner at Matrix, where I invest in and help early-stage fintech and vertical software startups. Matrix is an early-stage VC that leads pre-seed to Series As from an $800M fund across AI, developer tools and infra, fintech, B2B software, healthcare, and more. If you’re building something interesting in fintech or vertical software, I’d love to chat: [email protected]
2025-08-13 21:34:38
You might’ve heard the joke: airlines are just credit card companies that happen to own planes. It isn't far from the truth. Loyalty program fees generate the bulk of US airline profits.1
Airlines represent an increasingly common pattern: X is just a [bank/lender/issuer] that sells [unrelated, non-financial product].
Examples include auto and equipment manufacturers like Toyota and John Deere, which have extensive financing arms, homebuilders like D.R. Horton, which offer vertically integrated mortgage origination, and wireless carriers like AT&T, which earn hefty fees from device insurance and financing.
Modern software platforms have transformed this pattern into a powerful strategy that’s upending financial services and software alike. Restaurant owners no longer go to Wells Fargo for their card payments, they go to Toast. Landlords no longer wait in line at Chase to deposit checks or get loans, they use Yardi, etc.
These platforms—from the multi-billion-dollar pioneers like Toast, Shopify, and ServiceTitan to the crop of upcoming challengers—all represent the new strategy of embedded fintech, or when a non-financial company offers financial products that are highly contextual, curated, and customized for the company’s existing customers.
To appreciate the power of embedded fintech, consider the contrast between customers in line at the airport and those at the bank. Airline passengers may be very different from each other—different ages, different income levels, different reasons for traveling, different travel frequencies, and so on. But they all have one thing in common: a need to fly. Meeting that need is the airline’s main job, and they spend aggressively to fill seats on their planes. But airlines are notoriously low-margin businesses, and now that they have a captive audience, they see a cross-sales opportunity too good to miss. Airline credit card and loyalty schemes are designed to cross-sell relevant, complementary products to this captive audience.
Contrast the airport line with the line at your local bank. The people here are more diverse than at the airport, as are the services they’re waiting for. You may be next to a restaurateur setting up card payments, a construction company owner applying for a loan, new parents opening a 529 account, and a local lawyer opening a checking account. To appreciate the diversity, just look at the online product offering from Chase: everything from checking accounts to Disney Visa Cards and private banking to home mortgages and cross-border payments.
Both airlines and financial services are highly competitive industries. Both spend aggressively to acquire new customers, and both try to sell new customers as many products as possible to cover their acquisition costs and increase retention. The difference is that bank customers are so diverse that it’s impossible to customize an experience for them. Airline customers, by contrast, have just enough in common that it’s possible for airlines to customize experiences for them. That ability to customize is what gives embedded fintech its power.
Platforms with embedded financial services (PEFS) have been at the forefront of refining embedded fintech. PEFS like Shopify ($190B market cap), Toast ($25B), and ServiceTitan ($10B) are nominally software companies that serve well-defined customer segments or needs: Shopify is the one-stop shop for ecommerce merchants, Toast for restaurateurs, ServiceTitan for field service businesses, etc.2
But all earn 25% to 50%+ of revenue from financial services—everything from payments and lending, to payroll, spend management, and more. The revenue share is even higher for newer PEFS that are more deliberate about their financial product strategy. Let’s discuss some of the ways they’ve refined their product strategies.
PEFS target well-defined customer segments or needs, rather than serving a wide range of customers. They don’t try to be everything to everyone; they instead try to be everything to one type of customer. Contrast the generic, impersonal experience a lawyer, restaurateur, and plumber will get from a Chase branch with the highly customized, trusted experience they will get from a Clio, Toast, and ServiceTitan, respectively.
Targeting a specific audience enables PEFS to solve—and monetize—multiple problems for that one audience. Call this focus on being everything to one type of customer a maximalist product approach. The maximalist approach is best described by Parker Conrad's Compound Startup strategy, which leverages a single platform and shared infrastructure to launch a compelling bundle of products quickly to increase LTV and retention.
The maximalist approach and the compound strategy both extend to financial products, which PEFS treat as strategic priorities rather than afterthoughts. The compound strategy naturally complements many B2B workflows (see “B2B payments aren’t payments, they’re workflows”). Financial products have the added benefit of transactional business models rather than SaaS, which lowers the platform’s nominal usage cost for customers.
These financial products are deeply integrated from the start, so they create a unique flywheel that increases adoption and reduces risk. Beyond the benefits of zero CAC and high cross-sell rates, risk is an underappreciated benefit of embedded fintech. Most financial products require some risk assessment and underwriting. This is much easier when underwriting a specific type of user and business, and when you have a broader, longitudinal view of their activity via software.3
PEFS play a critical role in acquiring users and customizing both financial and non-financial products for their specific customers. But financial products come with onerous regulatory, compliance, risk, and capital requirements. Those requirements are why PEFS rely on a rising number of embedded fintech vendors (EFV) to embed and monetize financial products. Examples include Rainforest for payments and Unit for banking.
EFVs serve as a critical abstraction layer between software platforms and financial infrastructure providers like sponsor banks, card issuers, and lenders. Those providers don’t have the technical chops or embedded expertise to support fast-moving software platforms, and the individual platforms are unlikely to have the scale or expertise to work directly with a financial infra provider.4
EFVs meet the challenge of making financial infrastructure easy to embed, customize, and use across diverse platforms and use cases. Their success depends on mastering both the financial primitives they're built on and the unique requirements of each vertical market they serve. Most importantly, they need to deliver an exceptional experience both for the developers who integrate their solutions and for the end users who interact with them.
EFVs have made the wide array of financial products and services that banks offer embeddable within software platforms. Major categories like payments and lending have existed for a while, with companies like Stripe and Fundbox, but modern challengers like Rainforest, Unit, and Pipe have extended and accelerated adoption. Now nearly every product category has an embedded option:
Although EFV is a broad and fast-moving model, there are a few noticeable trends. EFVs are supporting a broader sliding scale of integration options, from fully built-out white-label components to raw APIs and platform ownership of critical functions like risk and underwriting.
Many PEFS are launching multiple embedded products and want the ability to pick and choose the best vendors for different functions like payments and lending. This development is leading EFVs to be more composable and interoperable—witness the integration of Rainforest with Unit. Many scaled EFVs are themselves going multi-product in an attempt to capture multiple financial use cases.
Finally, many EFVs are taking a cue from the platforms they serve and providing tools that complement their financial primitives. Think, for instance, of a card issuer providing embedded spend-management tools. Embedded tools like these offer higher margins and increase adoption.
Embedded fintech is so powerful because it changes the way consumers and businesses access financial services. It reflects the broader trend of offering more convenient, customized, and contextually relevant products and experiences enabled by software and the internet. Giants like Stripe and Toast have shown the power of embedded fintech. But we're only at the end of the beginning. New waves of natively embedded companies, particularly in vertical software, are building even larger and more ambitious companies with the model.
My name is Matt Brown. I’m a partner at Matrix, where I invest in and help early-stage fintech and vertical software startups. Matrix is an early-stage VC that leads pre-seed to Series As from an $800M fund across AI, developer tools and infra, fintech, B2B software, healthcare, and more. If you're building something interesting in fintech or vertical software, I'd love to chat: [email protected]
This isn’t just a B2B strategy. For example, Uber offers a range of curated, customized, and complementary products to drivers, from a card with high cash back on gas and EV charging, car rentals and sales, and even bundled insurance.
This is a critical topic that we don’t have enough space to cover here, but if you’re interested in learning more, check out “The best companies turn risk into leverage” and “Price and shape risk that others can’t or won't".
For more on embedded payments in particular, check out “Payfac in 1,000 words”.
2025-05-13 21:35:07
Imagine you’re the world’s most entrepreneurial dentist, the Mark Zuckerberg of molars. You’re great at cleaning teeth, but you’re also great at running a practice. You have your own playbook for the dental business.
Thanks to your playbook, your practice is larger and more profitable than the market average. You’ve surveyed your dentist friends. You’ve learned that you’re the best at acquiring patients, running your back office, configuring your CRM to minimize cancellations, and getting paid quickly. But you aren’t satisfied with cleaning teeth all day. You want to build a dental empire. The gap between the average business and what’s possible with your playbook is your opportunity.
How do you start your empire? Until recently, there were two broad and distinct paths: private equity (PE) and venture capital (VC).
In the PE model, you’d raise money to acquire a controlling stake in one or more practices. Then you would have full control to implement your playbook and manage those practices day-to-day.
In the VC model for B2B software, you’d raise money to start a company that builds software that codifies your playbook for any practice to implement. Maybe it’s a suite of CRM, scheduling, and billing tools, with opinionated workflows and design choices specifically for dental practices. You wouldn’t own or manage any practices directly, but your software would help them run better.
Both models start with the same principles: (1) the average company in a given vertical isn’t run as effectively as it could be, and (2) you have the insights and playbooks to run the average company more effectively. You believe you can close the gap between the blue and green dots, and get paid handsomely for it:
However, the PE and VC models take very different approaches to point #2. The tradeoffs of each model are most obvious in their approaches to control, concentration, and value capture.
The PE model is high control, high concentration, and high value capture. It says, “This specific business can be run better, and so it’s likely undervalued. I’m going to buy and run it, applying my playbook. As the business grows and becomes more efficient, it will become more valuable. I’ll benefit from the increased equity value.”
The VC model (at least within B2B software) is low control, low concentration, low value capture relative to an individual business. It says “There are lots of businesses in this market, and most of them could be run better. I’m going to build software that helps them do that. I’ll charge a small fee, but will sell it to many businesses.”
The PE and VC models are extreme ends of a spectrum, rather than distinct models. PE and VC firms are both in the business of generating returns for their investors. They take capital, combine it with their belief in the superiority of their playbook, and then implement their strategies and playbooks in a given vertical. This may involve buying businesses or building tools to enhance their performance, generating profits, and returning the profits to their investors.
Until recently, you’d be forgiven for thinking PE and VC are effectively distinct entities. However, headlines like these are making it more apparent that they’re just a spectrum of strategies and that there’s a lot of white space in between them:
Why are VCs (and the companies they fund) adopting PE-like strategies? Let’s explore.
In seeking outsized returns, VCs and venture-funded startups are venturing beyond their previously narrowly defined model. They’re getting more creative and aggressive in exploring the messy middle, the whitespace between the previously distinct PE and VC ends of the spectrum.
Several factors are behind this move towards the messy middle. Some are pushes from the traditional VC model: the classic, almost boutique approach of minority investments in asset-light, high-growth, all-or-nothing, power-law-seeking, software-first-and-only startups. As this model gets more competitive, smart founders and investors realize they must try something different. At the same time, several powerful factors are pulling toward these new models: new technologies and business models seem to enable venture-like returns from traditionally non-venture-type businesses.
The push side is well-documented: the traditional VC asset class has become saturated, especially in B2B SaaS. Nearly a trillion dollars have flowed into the venture industry in the last decade. This has led to a proliferation of companies serving every vertical and every niche. Some market maps are so crowded that you need a microscope to make out a single logo:
At the same time, AI promises to further reduce the cost of software development. That isn’t to say that the B2B SaaS market is going to zero, or that there won’t be another generational B2B SaaS business. But the noise and saturation make it harder for these companies to grow quickly while retaining customers and high margins. There aren’t many land grab opportunities in pure SaaS like there were in the 2010s.
Even for companies with great products, it’s getting harder to sell to businesses with SaaS fatigue. To return to the original dental example: suppose you (the entrepreneurial dentist) build the best vertical software for dental practices. Your target customers are already inundated with such pitches. The same is true if you’re pitching the most well-oiled acquisition strategy for underperforming practices. It’s hard to sell them a genuinely better product, and it’s also hard to convince them to sell their businesses.
That brings us to the pull factors. What makes the non-pure-software approach so attractive? First, new tech and business models enable startups to monetize more than just software subscriptions. I talked about this in “Invisible asymptotes in vertical software”. Products like embedded payments with Rainforest* or embedded marketing with Reach* allow software companies to capture a variable portion of their customers’ success, upside that grows as their customers grow, and not just a fixed fee for software.
Another is the rise of AI as a credible substitute for certain types of work. This has two direct effects:
First, it allows significant costs to be taken out of a business and/or for labor to be added to places where it previously wasn’t economical. For example, an accounting firm that previously needed a large back office team for mundane, repetitive tasks can offload more of that to AI and reduce headcount. A small field services business can now use voice agents to answer the phone at a fraction of the cost of a full-time employee.
Second, AI allows companies to implement a broader, more consistent, and more scalable version of playbooks. If PE involves implementing playbooks for teams to follow, while VC involves building software that encodes playbooks into workflows and data models for teams to follow — both are vulnerable to human error. In both cases, the people involved need to actually follow and implement the playbooks. AI takes that out of human hands.
So, VC funds and venture-backed companies can and must explore strategies in the messy middle. They both (1) must move away from old and less effective models, and (2) can adopt new models with higher expected returns. That’s why seemingly unconventional company structures have emerged.
To improve businesses and generate returns, the VC model uses code as its primary source of leverage, while the PE model uses capital. The new models mix code and capital in unique ways to achieve the same result: improve how a set of businesses operates, then harvest upside. The latest mixtures of code and capital allow investors and entrepreneurs to work with more businesses. They also make it possible to create and capture more value than was previously possible in a pure VC or PE model.
The new models seem to fall into three categories: the crown jewel, the roll up, and business-in-a-box.
In the crown jewel model, a VC firm or VC-funded company acquires a one-off, large, strategic, and non-tech asset. The crown jewel might be a hospital or health system, a professional service firm in law or accounting, a call center or BPO, or a parking lot operator. The acquirer doesn’t simply acquire and run a business marginally better (as in traditional PE). The acquirer may already have a proprietary tech stack or strategy that it’s applying to the crown jewel asset. Or the acquirer may build such technology as it operates the new business, with the crown jewel asset as its ideal and captive first customer. Under this model, the acquirer applies proprietary tech to an already at-scale business where it will have the greatest and fastest ROI. Crown jewel examples include the AI-powered parking software company Metropolis acquiring a century-old, publicly-listed parking lot operator SP Plus for $1.5 billion, or General Catalyst acquiring a large healthcare system in Ohio for $500 million.
In a venture roll-up, a company acquires multiple non-tech assets and then applies its proprietary technology and strategies. Think of a roll-up in contrast with the opportunistic, often one-off, and strategic M&A of the crown jewel model. In the venture roll-up model, the M&A is continual and applied to multiple businesses. Traditional PE often involves a low-tech roll-up approach. The venture roll-up model differs from the PE roll-up because it involves applying a full stack of proprietary, purpose-built software for that vertical. The venture roll-up may also build in-house technology to assist and scale the M&A process, such as proprietary data and underwriting to identify, underwrite, and value the best acquisition targets. Examples include Cabana, Pipe Dream, and Long Lake in field services, as well as OLarry and Multiplier in professional services.
In the last model, business-in-a-box, the company actually helps create and start new businesses in a vertical. Like the roll-up model, it brings proprietary technology and works with multiple end businesses in parallel. But it creates these businesses rather than acquires them. It can create businesses under unique individual brands (business-in-a-box) or under a common brand (modern franchising). These can include purely digital services offered at least partly through a managed marketplace. Examples include Fora for travel agents or Headway for therapists, as well as brick-and-mortar businesses, such as Moxie for med spas. Sam Gerstenzang, the founder of Moxie and Boulton & Watt, is the expert on the model and wrote a great piece on which markets are most amenable to it.
These models are evolving quickly and aren’t mutually exclusive. For example, biz-in-a-box vendors may execute some opportunistic roll-ups or vice versa. Similarly, incumbent vertical software companies may begin adopting these new models. To return to the PG tweet above, if a vertical software provider finds its market or margins tapped by just selling software and services, it may be compelled to start operating businesses in their verticals. Alongside the list of products vertical SaaS companies offer, such as CRM and payments, it’s increasingly common to see an option to “Start your business with us” or even “Sell your business to us.” For example, Slice (“Start your pizzeria”) or MoeGo Care. This creates some interesting territory to navigate — including the potential for channel conflict and competing with existing customers.
The biggest companies are built and the best investments made when things are in flux, when generally accepted models have been exhausted, but no obvious replacement has risen. That flux is happening in VC and PE. The classic VC/PE investment model, as well as the traditional strategy of VC-backed, SaaS-centric, B2B software companies, is approaching a saturation point. It’s still early as funds and companies explore new combinations of code and capital, of software and financial engineering, in this messy middle. Interestingly, it isn’t just a game for new funds or companies. Some of the earliest examples of this trend are established investors (e.g., General Catalyst / Summa) and software companies (e.g., Metropolis / SP Plus). But the model is evolving, and there will be many new opportunities to build generational companies by creatively mixing code and capital.
My name is Matt Brown. I’m a partner at Matrix, where I invest in and help early-stage fintech and vertical software startups. Matrix is an early-stage VC that leads pre-seed to Series As from an $800M fund across AI, developer tools and infra, fintech, B2B software, healthcare, and more. If you're building something interesting in fintech or vertical software, I'd love to chat: [email protected]
2025-03-26 22:30:51
Forget being dismissed as a ChatGPT wrapper. Anyone building vertical AI should fear being a wrapper around a system of record.
The loud debates over “AI versus software” frame the choices poorly. They obscure the right strategy for each player in the emerging stack: incumbent vertical software, emerging vertical AI, and general model companies. Let’s look at how that stack started and is evolving.
The classic software stack has three layers:
This architecture applies to single software products and companies' broader software stack. Businesses need reliable sources for important data like customers, expenses, and revenue. Employees generate this data in fragmented and sometimes conflicting ways. For example, information about a customer may be spread between the teams and tooling in sales, implementation, and billing—but it must ultimately be synthesized and reconciled into a single place.
The relative value of these layers has changed as software has evolved. In the early software age, only a handful of vendors existed. The winning ones developed all-in-one solutions and systems of record. They centered on vital business resources: employees (human capital management, HCM), customers (customer relationship management, CRM), and revenue/expenses (enterprise resource management, ERP). In this model, most of the value accrued to databases like Oracle, SAP, and Salesforce.
SaaS made software cheaper to create and distribute. This broke down interaction systems into many specialized apps. During this time, there were notable SOR IPOs like Workday, ServiceNow, and Veeva. However, SOE companies like Docusign, Box, Zoom, Slack, and Smartsheet really defined the era. They created value by digitizing and streamlining employee workflows.
The latest evolution of this stack is vertical software (vSaaS). It combines the SOI and SOR into one product designed for a specific vertical. For example, Clio for lawyers, Housecallpro for field services, Brightwheel for schools.
A single product in a vSaaS bundle is likely weaker than a general option. For example, Clio’s CRM isn’t as strong as Salesforce’s. However, lawyers still prefer Clio because it’s purpose-built for them and integrates with all the other modules that Clio provides. For example, a prospect in Clio’s CRM can be turned into a client, so lawyers on Clio can track time, assign work, bill hours, and get paid.
Vertical AI is shaking up the stack again. It comes in many flavors. Co-pilots augment labor and workflows. Agents offer a more invisible “do it for me” value proposition. AI is also making software easier and cheaper to build.
Intelligence is becoming cheaper. The software for interacting with it is also more affordable. Everyone thinks this means every business will build their own software. Or that they won’t need to build software per se because the intelligence at the interaction layer will be nearly free.
What becomes more valuable if intelligence is cheaper and the interactions with it are cheaper? It’s the data model for a vertical, the data for a specific business that fits its vertical data model, and the workflows and controls necessary to use that data to fit that specific business or industry. Or—put another way—a system of record.
Let’s take a hypothetical example in the roofing market. “ShingleSoft” is a vSaaS for roofing companies. It has a full suite of software products: CRM, estimates, scheduling, project management, inventory management, time tracking, payroll, billing and payments, accounting, etc.
“RoofGPT” is an AI-focused startup. It knows that many roofing companies can't hire full-time sales, marketing, and operations employees. Many sales calls to roofers go to voicemail while the employees are up on roofs. So RoofGPT offers a voice agent. It picks up on the first ring and qualifies customers. It’s the equivalent of having infinite smart and polite salespeople ready near a phone 24/7.
The problem is that these “salespeople,” no matter how smart or skilled, are disconnected from the business. They’re perfectly capable of requesting numbers for a callback, and answering deterministic questions about operating hours. But they struggle when they’re asked about something that an employee sitting in the roofer’s office would know.
“Can I reschedule my inspection?” requires access to the roofer’s CRM to look up the customer record and the employee schedule.
“Can I pay my bill over the phone and can you send me a receipt?” also requires access to the CRM, invoicing, and payments system.
If all the roofing companies use a fragmented suite of tools (e.g., CRM from Hubspot, scheduling from Calendly), RoofGPT could build API integrations with all of them. However, for many verticals, these functions have been consolidated into a single vSaaS, like ShingleSoft. This is a double-edged sword for RoofGPT. They can interact with a single platform to create and update records across multiple functions. But they’re also beholden to that single vendor. Suppose they choose not to integrate. Suppose they focus on providing excellent voice functionality without direct access to the SOR. In that case, they’ll face commoditizing pressure from increasingly capable horizontal AI tools (e.g., OpenAI’s Operator).
So RoofGPT is facing two pressures. First, the pressure of increasingly capable and commoditized intelligence. Second, the pressure of increasingly centralized, rare, and controlled data and interaction with the vSaaS providers.
ShingleSoft isn’t an AI-native company, so they may be slower to build AI features. They may already have a robust API and partner ecosystem, so they may support RoofGPT initially. However, the core strategy of vSaaS is clear: bundle, bundle, bundle. The more products they offer, the more roofers they can acquire and the more they can charge. As outlined in “Invisible asymptotes in vertical software”, successful vSaaS know they need to (1) solve as many problems as possible and (2) find monetization models beyond SaaS that both drive growth/success of their customers and capture more of that upside and growth.
“Pretty good” products that are deeply embedded into the core vSaaS product are likely to beat “Excellent” products that lack deep integration and lack platform distribution.
Put another way, if I’m a roofer, which of these options would I prefer?
(A) the world’s best salesperson, but they’re stuck in a room alone with no access to live info about my business
(B) a decent salesperson who is deeply familiar with business, services, schedule, and employees
Most would probably choose (B). That’s especially true if the gap between the quality of the salespeoples’ standalone skills keeps shrinking every month.
Every business is a wrapper. To be flippant, SaaS companies just wrap cloud providers, which wrap cloud hardware providers, which wrap semiconductor foundries. Being a wrapper is not in itself a bad thing. What you wrap and how you wrap it makes it bad or good.
Especially if you’re wrapping a commodity, you must give customers a reason to buy from you versus going direct. You can do this with product (by improving or adding value to it). You can also do it with distribution (selling it cheaper and/or more effectively than customers could get it themselves). Having one is good, having both is better.1
Today vertical AI agents are primarily wrapping product, building something novel enough that it’s getting them powerful distribution. However, there’s a limit to how much value they can add without owning or accessing the SOR. At the same time, “good enough” intelligence is being commoditized. I doubt intelligence will ever be so commoditized that the end B2B user relies exclusively on agents. But I think it will be sufficiently commoditized enough to seep into adjacent software products. It’s already happening. Successful vSaaS companies are expert bundlers. They’ve mastered multi-product motion. They also have existing distribution through their install base. And because they are the SOR, they can build an integrated experience much better.
So what’s the right strategy here? Will vAI or vSaaS win?
Emerging vAI companies have a harder path to success than they thought. Early fast growth will taper off as they, their customers, and vSaaS partners realize the value of integrating with the SOR (and limits of not being integrated with it). vAI companies should ignore the FUD around the “end of software” and have a clear plan to displace the SOR. This may apply less to vAI companies in verticals without a modern vSaaS incumbent. They’ll have an easier time starting with an AI wedge and building out the SOR rather than displacing another one.
Existing vSaaS companies are net beneficiaries of AI commoditization. It’s likely a sustaining rather than disruptive innovation to them. They have more time to make strategic moves in offering their customers AI products, whether by building, buying, or partnering.
Embedded AI for vSaaS is a newer category, empowering vSaaS to add vAI features faster and more successfully. This is different than simply embedding AI. It involves configuring AI to help vertical-specific workflows, such as marketing and customer support (e.g., Reach). Think how companies like Rainforest or Stripe enable vSaaS to embed payments rather than become payfacs themselves.
The dynamic between vAI and vSaaS is reminiscent of the early days of the streaming wars. There was competition between Netflix (a streaming platform that licensed content) and HBO (a content company without a streaming platform). In 2013, a Netflix exec famously said, "The goal is to become HBO faster than HBO can become us." The right answer wasn’t streaming or content but both together.
It’s worth noting that most of the major streaming services today are “full stack.” They own the content production and tech. They’re about evenly split. Around half started with streaming and got into content production (Netflix, Amazon Prime Video, Hulu, Apple TV). The other half started as content producers and got into streaming (HBO, Disney, Paramount, Peacock). I expect something similar will play out in B2B software, where the winners in vSaaS will own the IP (customer data and workflows), not to mention existing brand and distribution, and layer on the new tech.
The dichotomy of AI versus software is a false one, especially in B2B verticals. They complement each other, and each vertical’s winner will thoughtfully combine both.
My name is Matt Brown. I’m a partner at Matrix, where I invest in and help early-stage fintech and vertical software startups. Matrix is an early-stage VC that leads pre-seed to Series As from an $800M fund across AI, developer tools and infra, fintech, B2B software, healthcare, and more. If you're building something interesting in fintech or vertical software, I'd love to chat: [email protected]
Think of cloud storage, something rare that turned into a commodity. Dropbox started off making this rare/difficult-to-use resource usable to the average person. It’s still a great company today, but as storage became commoditized, the winners in B2B/B2C are the ones that had an existing distribution advantage because of existing customer relationships and/or had products that enhanced the use/value of commodity storage (Apple’s Photos, Google’s Gmail and Photos and Drive).
