By Novan Le, Head of Finance at Payabli
For most of the SaaS era, building a great software company followed a relatively understandable playbook. Find an underserved workflow, build a better product around it, establish a recurring subscription relationship, and continue adding features until the platform becomes difficult to replace. Product breadth mattered because software was expensive and slow to build. If it took a competitor several years and dozens of engineers to recreate what you had built, a strong product lead could itself become a meaningful competitive advantage.
AI is changing that equation faster than I think most of us fully appreciate.
As a finance leader, I tend to look at technology shifts through a fairly simple lens: what changes the long-term quality of a company’s revenue, margins, retention, competitive position and ultimately its enterprise value? From that perspective, the important question isn’t whether AI will create more software. It almost certainly will. The more interesting question is what kind of software will remain valuable when creating software itself becomes dramatically cheaper.
Investors and software buyers are already wrestling with that question. PwC described the software market in early 2026 as undergoing a selective repricing, with a growing distinction between durable platforms with defensible moats and products whose value resides primarily in surface-level functionality that AI can increasingly reproduce. A recent 2026 survey of more than 200 private equity investors, strategic acquirers and SaaS CEOs found that 85% viewed AI-driven commoditization as the biggest risk facing SaaS. At the same time, buyers were still willing to pay premium valuations for businesses with durable growth, strong retention, deep workflow integration and credible AI strategies.
That distinction is important. AI is not necessarily reducing the value of software. It is changing where the value resides.
From owning features to owning the workflow
One of the more useful frameworks I’ve seen recently comes from Foundation Capital’s work on what it calls “context graphs.” In a January discussion with Foundation’s Ashu Garg, Altimeter’s Jamin Ball and Animesh Koratana, the group argued that the constraint on enterprise AI increasingly isn’t model intelligence. The leading models are already sufficiently capable for many business use cases. The harder problem is giving AI the context to understand how a particular business actually operates.
Businesses make thousands of decisions that never make it cleanly into a database: why a customer received an exception, why a transaction was flagged, why one invoice was handled differently from another, which policy governed a decision, what happened afterward and whether the outcome was ultimately good or bad. Foundation calls these histories “decision traces.” Their argument is that applications sitting directly in the execution path of work have an opportunity to capture those traces and turn them into a proprietary understanding of how a business operates.
I think that concept gets much closer to what an AI-era moat actually looks like.
The traditional SaaS advantage often came from owning a system of record. The next generation of durable software will increasingly combine that system of record with a system of action. Instead of merely recording that an invoice was paid, software can determine how it should be paid, execute the payment, handle an exception and reconcile the result. Instead of simply displaying merchant risk, it can identify an anomalous pattern, determine whether the behavior falls outside policy, request additional information and route the account for review.
That matters because a product that participates in the decision and execution process sees information that a product sitting outside the workflow never sees.
Bessemer has arrived at a similar conclusion through its work on vertical AI. Its framework differentiates between AI functionality that simply improves productivity and products that own increasingly complete, high-value workflows. Bessemer specifically warns against easily commoditized capabilities such as basic extraction, matching or reconciliation in isolation. The stronger opportunity is combining AI with deeply integrated, sector-specific workflows where the application delivers economic outcomes rather than another feature.
This is also why vertical software may be particularly well positioned. A generic model can understand accounts receivable. It is much harder for a generic model to understand the complete collection workflow of a property manager, dental practice, field-service company or HOA, including the industry’s terminology, business rules, customer relationships, exceptions, payment behavior and regulatory constraints.
The vertical platform already has much of that context. The strategic question is how much more of the workflow it can own.
Why payments become more important, not less, in an AI world
This is where I believe embedded payments become particularly interesting.
Payments are sometimes discussed primarily as a monetization strategy for SaaS. That’s certainly part of the attraction, but it understates their strategic value. When payments are deeply embedded, a software company moves from helping its customer run a workflow to actually participating in the economic transaction that completes that workflow.
Consider what that means from a data and context perspective. A scheduling platform may know that an appointment was booked. A platform that also owns the payment experience can know whether the customer paid, which payment method they selected, whether the transaction succeeded, how long settlement took, whether it was refunded or disputed, and how the merchant ultimately reconciled it.
The same logic applies on the outbound side. Software may know that a vendor invoice exists, but a platform that manages payables can also see who was paid, when they were paid, which payment rail was selected, whether the payment succeeded and what happened when it didn’t.
These are not merely additional data points. They are closed-loop outcomes.
That becomes extremely valuable in an AI environment because intelligence improves when it can connect a decision to its eventual result. A system can learn that a particular risk signal preceded a chargeback, that a particular merchant onboarding pattern correlated with higher activation, that certain invoice characteristics increased the likelihood of late payment, or that vendors in a particular category were substantially more likely to accept a virtual card.
This is why I would be careful about simply calling transaction data a “data moat.” Data by itself is rarely enough. Large models are improving rapidly, third-party datasets continue to expand, and many supposed proprietary datasets turn out to be less differentiated than expected.
The more defensible asset is the combination of vertical context, workflow position, transaction history, decision-making and observable outcomes. When those components live together inside a platform, each transaction can make the system incrementally more useful.
Agentic AI makes the financial layer even more consequential
The emergence of AI agents strengthens the argument further because an agent that actually performs work eventually needs the ability to take real-world actions.
That became a recurring theme across this year’s payments and commerce conferences. In sessions featuring Amazon, AWS, Mastercard, Klarna and other software and commerce leaders, the conversation had noticeably moved past whether AI agents would participate in commerce and toward harder questions around authorization, identity, fraud, permissions and trust.
Amazon’s agentic-payments team described merchant conversations moving from “What is agentic commerce?” to “How do we get ready for it?” More importantly, merchants are beginning to ask who actually authorizes a transaction when the customer isn’t sitting in a checkout screen. AWS and other infrastructure providers separately discussed the rise of agent-to-agent commerce and the infrastructure necessary to put spending limits, authorization rules and payment credentials around autonomous software. Those same discussions noted that while consumer applications remain relatively early, B2B software use cases are already progressing toward agents operating with persistent context and delegated authority.
Mastercard made essentially the same point at its 2026 Connections conference: in a world where agents transact with other agents at machine speed, payments infrastructure has to become a layer of trust and control rather than simply a mechanism for moving money.
For vertical SaaS companies, this creates an interesting strategic advantage. The platform that already understands the customer, the workflow and the money movement is much better positioned to determine what an AI agent should be allowed to do.
Imagine a future field-service platform where an agent doesn’t simply tell a contractor that an invoice is overdue. It determines which invoices require follow-up, contacts the customer, offers an appropriate payment method, executes the payment when authorized and reconciles the result. Or a property-management platform where an agent identifies a vendor invoice, verifies the vendor, evaluates payment options, executes the payment and handles reconciliation without a person manually moving between four systems.
AI is what makes those experiences possible. Payments are what allow those experiences to finish the job.
That distinction between advising and executing may become one of the most important dividing lines in enterprise software.
The CFO case: embedded payments can change the economics of the company
There is another reason payments matter, and it is much more familiar territory for a CFO: the economics.
Traditional SaaS monetization is usually bounded in some way by subscription pricing. A business can increase prices, sell more seats, add modules or move customers into larger tiers, but monetization is still largely connected to software access.
Payments introduce a revenue stream tied to the economic activity of the customer’s business. If the customer processes more transactions, the platform has the opportunity to participate in that growth without having to acquire another customer or add another seat.
Recent platform operating data supports the significance of that model. Median payments adoption among vertical SaaS platforms increased from 27% in 2024 to 40% in 2025, while leading platforms achieved adoption rates of 80% or more. Tidemark research estimates approximately $4,200 of incremental annual recurring revenue for each customer adopting embedded payments, and the same operating data shows lower churn among platforms with embedded financial products and materially faster growth among multiproduct platforms.
What I find more interesting than the incremental revenue, however, is the second-order effect. Payments can increase revenue per customer, but they can also deepen the customer’s dependency on the platform in a constructive way. Once customers use the same system to manage the workflow, accept money, make payments, reconcile transactions and increasingly access adjacent financial products, the software becomes more central to operating their business.
That creates the potential for better retention and net revenue retention at the same time that it adds a new gross-profit stream. In financial terms, you’re improving both the quantity of monetization and the durability of the customer relationship.
That combination is important to investors because enterprise value isn’t simply a function of revenue. Buyers ultimately underwrite the durability and growth of future cash flows.
What investors and acquirers are actually looking for
There is a temptation in fintech to argue that adding payments automatically creates a valuation premium. I don’t think that’s the right way to think about it.
A poorly implemented payments business with low adoption, weak unit economics and limited strategic integration will not suddenly transform the valuation of a software company. In fact, it may create operational complexity without enough economic benefit to justify it.
Payabli’s recent roundtable discussion on how buyers evaluate vertical SaaS companies with embedded payments made this point well. The consensus was essentially that traction matters more than the story. Buyers will examine payment penetration, TPV, net take rate, cohort adoption, gross margin and the impact payments have on NRR and GRR. A company saying it has a large payments opportunity is different from a company demonstrating that its customers have adopted payments and that the resulting revenue behaves predictably over time.
That view is consistent with the broader M&A market.
· Kroll’s 2026 software M&A analysis found strategic buyers accounting for 76% of transactions in the period it analyzed, with buyers particularly focused on AI capabilities, data infrastructure and workflow automation. SEG similarly describes buyers becoming more selective about targets while concentrating competition around companies with strong retention, durable revenue and clear competitive positioning.
· PwC uses a term I like for this: workflow gravity. The most defensible software has enough domain expertise, proprietary context and ecosystem entrenchment that AI makes the platform more valuable rather than easier to replace.
· Embedded payments can contribute to that gravity because payments touch the customer’s workflow, data, economics and operating processes simultaneously.
· A June 2026 discussion between L.E.K., CFO Dive and Worldpay went even further. Their research found valuation premiums as software platforms progressed from software-only businesses into embedded payments and then broader embedded finance. Their illustrative model suggested embedded payments could contribute meaningfully to enterprise value over a five-year investment period. I would treat any exact modeled uplift as scenario-specific rather than a universal rule, but the underlying mechanism is compelling: payments add recurring transaction economics, deepen financial workflows and create logical pathways into products such as payables, working capital and other embedded financial services.
· The strategic ceiling of the software company gets higher.
The opportunity isn’t to become a payments company
None of this means every vertical software company should build payment infrastructure itself.In fact, I think the opposite is usually true. Payments contain an enormous amount of complexity that end users rarely see: merchant underwriting, card-network requirements, risk monitoring, tokenization, compliance, settlement, reconciliation, pricing, disputes, money movement and support. The more deeply payments become embedded into a platform, the more operational infrastructure exists beneath the surface.
The software company’s scarce resources are usually better spent on the things that are unique to its market: understanding the vertical, controlling the customer experience, designing the workflow and building the intelligence that makes that workflow better.
The role of an infrastructure partner should be to make the underlying financial layer possible without forcing the software company to recreate a payments organization from scratch.
That is the philosophy behind what we’re building at Payabli. Payabli gives software platforms infrastructure across Pay In, Pay Out and Pay Ops so they can own the embedded financial experience and participate in its economics while we handle much of the complexity underneath it. Increasingly, that infrastructure is intelligent as well. Amigo and Payabli’s broader intelligence platform bring AI into onboarding, payments analytics, risk, vendor enablement and other operational workflows, while agent-facing tools are being built to allow software agents to query, authorize, reconcile and ultimately execute payments natively.
I think this architecture is particularly relevant to the AI era. A vertical software company should not have to choose between becoming an AI company and becoming a fintech company. It should be able to use AI to make its vertical workflow more intelligent while using financial infrastructure to make that workflow economically complete.
Building for enterprise value in the next generation of software
We’re still early enough in the AI transition that anyone claiming to know exactly how software businesses will be valued five or ten years from now is probably overconfident. Even today, investors disagree about which application-layer moats will persist, how quickly agents will replace traditional interfaces, and whether many current software categories expand or collapse. But the market is already giving us useful signals.
The cost of creating functionality is falling. Buyers are becoming more skeptical of products differentiated primarily by features. At the same time, they continue to place significant value on companies with durable revenue, embedded workflows, proprietary context, customer trust and clear economic value.
That suggests that the best response to AI isn’t simply to ship more AI features.
For vertical software companies, the more durable strategy is to move deeper into the customer’s business: own more of the workflow, participate in more of the actions, understand more of the context and connect those actions to measurable outcomes.
Embedded payments can be an unusually powerful part of that strategy. Done well, payments add gross profit and expansion revenue, create a more integrated customer relationship, produce high-value transaction and outcome data, open the door to broader embedded finance, and provide the trusted execution infrastructure that increasingly autonomous software will require.
The software companies that are most interesting to me aren’t becoming less important because of AI. They’re becoming more deeply embedded in how their customers actually operate.
And when the software owns both the workflow and the movement of money through it, AI doesn’t have to erode the moat. It can make the moat deeper.