The private equity landscape recently witnessed a financial seismic shift that has forced industry veterans to sit up and take notice. Last week, a Blackstone-led investment group quietly wrested control of customer-experience software giant Medallia from Thoma Bravo in a move containing a staggering five-billion-dollar write-off. Representing the second-largest loss in the entire history of private equity, the write-off was a stark reversal from 2021, when Thoma Bravo triumphantly took Medallia private in a transaction valued at $6.4 billion. Buoyed by temporary cheap, pandemic-era debt and a valuation trailing close to nine times forward revenue, the deal was built on the rock-solid assumption of perpetual high-velocity growth. Thoma Bravo’s founder, Orlando Bravo, has since termed the venture a “big mistake,” admitting his firm bet on rapid market expansion that failed to materialize. Yet, dismissing Medallia’s spectacular collapse as a mere byproduct of bad timing or poor financial underwriting completely misses the broader, structural transformation occurring under our feet. While analysts point to rising interest rates as the singular villain, the reality is far more disruptive. To understand this shift, examine businesses like Chegg. The academic homework-help platform possessed no crippling debt and operated as a highly profitable, public enterprise valued at $14 billion in 2021. For a decade, Chegg built an seemingly impenetrable moat by paying contractors to assemble a proprietary library of millions of academic solutions, renting them to students. Then, ChatGPT arrived—delivering answers that were free and superior. Within months, Chegg’s CEO confessed that the chatbot was eating into its core growth. When Google’s AI-generated search summaries systematically severed Chegg’s organic referral traffic, the company cut half its staff, ending with a market capitalization hovering at a heartbreaking $100 million. The 2021 underwriting model missed a brutal reality: when the marginal cost of producing an answer, summary, or codeline falls toward zero, a recurring-revenue moat built on producing those outputs completely collapses.
This sweeping industry-wide phenomenon has sparked an era of profound anxiety that technology insiders have dramatically dubbed the “Saaspocalypse.” It is an evocative term, yet it suffers from being far too generalized. The claim that artificial intelligence will unilaterally exterminate all software-as-a-service businesses misses crucial tactical defense: AI does not kill all SaaS, but it does ruthlessly gut specific, vulnerable categories. Customer-experience and feedback software reside squarely in this digital firing squad. This explains why Medallia’s closest peer, Qualtrics, was pulled off the public markets in an identical defensive maneuver, acquired by Silver Lake in a massive $12.5 billion take-private transaction in 2023. When a company’s value proposition is simply collecting qualitative feedback, synthesizing client sentiments, and outputting generic analytical reports, it is building its house on sand. Modern large language models can perform these tasks natively, instantly, and for a fraction of a cent. Therefore, the existential challenge facing tech founders, venture capitalists, and private equity operators today is dividing the vulnerable from the invincible: which software architectures possess a structural moat that actually hardens and expands when the cost of generation drops to absolute zero? Particularly within the highly complex world of financial technology, a durable business cannot rely on the superficial beauty of its interface or simple automation. Instead, truly resilient platforms must ground their survival in a tripartite framework of competitive advantages that I define as the “3Ds”: Distribution, Data, and Delivery. In this terrifying new paradigm, artificial intelligence does not render this classic framework obsolete; rather, it acts as a high-powered prism, splitting each of these three critical legs directly in half. One half represents the replicable, commoditized tasks that a cutting-edge frontier AI model can execute over a weekend, so that portion has evaporated. The remaining half contains the deeply human, operational, and structural complexities that no model can reach, making this half infinitely more valuable. The entire art of modern investing and company survival boils down to successfully telling these two halves apart.
To appreciate the fracturing of these competitive moats, we must first dissect the leg of Distribution, which counterintuitively represents the advantage that is actually strengthening in the AI age. We have officially entered an era of absolute hyper-abundance where anyone can conceptualize, code, and deploy a highly functional, beautifully polished software product over a weekend. Because of this, building software has ceased to be a scarce skill, shifting the competitive battlefield from product creation to customer demand and retention. The only forms of distribution that will survive the coming decade are those that cannot be easily conjured up by a smart algorithm. This means sitting deeply inside a client’s existing system of record, holding highly regulated banking licenses, possessing proprietary regional payment rails that a competitor cannot easily copy, and carefully cultivating the deep human trust required to convince a stranger to let you move their money. Chegg’s fatal mistake was relying entirely on “rented” distribution through Google search referrals. When Google modified its search engine algorithms to prioritize its own native AI summaries, Chegg’s primary customer acquisition channel went entirely dark, proving that rented distribution will vanish the exact second the channel owner decides to reprice or redesign their platform. Compounding this challenge, the buyer of these software services is starting to evolve from a human consumer into an autonomous artificial intelligence itself. Recently, Visa, Mastercard, Stripe, and Google have all quietly built and shipped dedicated transactional rails designed to facilitate payments made directly by AI agents. Visa frames the upcoming year as the crucial historical tipping point where autonomous AI agents transition from merely assisting humans to actively executing transactions on their own. McKinsey projects that this machine-to-machine economy could represent up to one trillion dollars in the United States alone by 2030. When an agent is the entity making the purchasing decision, traditional marketing strategies designed to appeal to human psychology—such as SEO, paid traffic, visually stunning user interfaces, and frictionless sign-up flows—become completely useless. Consequently, builders must learn to ask entirely new, highly technical questions: Are we the precise API endpoint that the autonomous agent is incentivized to choose? If so, why? Do we own the foundational payment rail that the agent settles its transactions on, and are those transactions legally protected?
The second leg of our strategic framework is Data, which represents the specific area where technology founders and venture capitalists deceive themselves most frequently and most disastrously. Before AI, possessing a massive, static archive of historical data was celebrated as an impenetrable moat, but today, these legacy silos are rapidly depreciating assets easily replicated by standard foundation models. In highly regulated and complex fields like fintech and healthcare, the data that survives this technological purge is not a dusty archive, but rather a live, exclusive, and regulated stream of real-time information that your system continuously generates and no competitor can buy. For example, extending credit to a merchant by tapping directly into their live, consented transaction history provides an incredibly powerful, real-time moat that is proprietary and continuously updated. Similarly, uncovering localized fraud and credit patterns tied to a specific geographic market remains highly defensible, simply because a generalized foreign model developed in Silicon Valley cannot cheaply or easily acquire that granular, localized context. Therefore, the modern founder must stop boasting about the sheer volume of historical data they have accumulated over the past decade and ask: Do we sit directly on top of an active, compounding flow of exclusive information that dynamically feeds our models every single second of the day? Within financial technology, the dividing line between defensible and defenseless data is remarkably clean. Proprietary, real-time data continues to dominate generalized models when it comes to credit underwriting and fraud prevention. This live data must be deeply woven into complex daily workflows that cannot be easily automated away by a generalized foundation model operating in a vacuum. Builders must look long and hard at their product and determine whether they are building a superficial application that AI will completely rebuild from scratch, or whether they are constructing an indispensable, data-rich workflow tool that AI is forced to leave alone.
The final leg of our framework is Delivery, an advantage that has historically taken many creative shapes in the technology ecosystem. For some companies, their delivery advantage lay in their ability to delight users with an intuitive and beautiful interface. For others, it was about achieving a highly specialized delivery mechanism, such as being deeply embedded within a merchant’s payment flow, obfuscating what was once a complex back-end experience. Today, however, clean UI and basic digital delight have been thoroughly commoditized. Because an AI-native competitor can casually stand up a polished, beautifully designed software product in any target market in a matter of weeks, a great user experience is table stakes rather than a moat. The only part of the delivery moat that remains absolutely durable is what sits quietly beneath that surface-level interface: highly trusted, deeply compliant, and legally accountable execution. Ultimately, when a transaction occurs, money must settle, compliance laws must be met, and a legally liable corporate entity must take responsibility when the model inevitably hallucinates or fails. This crucial dynamic explains my focus on the boring but essential spaces of trust, identity, and compliance verification. It is incredibly easy to build a beautiful chatbot that claims it can handle corporate tax filings or execute foreign currency transactions, but the moment the system makes an error, the chatbot cannot go to court, absorb financial liability, or negotiate with regulatory bodies. The businesses that enjoy massive advantages in the era of artificial intelligence are those that do not merely generate pretty outputs, but actively shoulder the legal, operational, and financial responsibility of executing those outputs in the real world.
Stepping back, the inescapable conclusion of this shifting paradigm is the paramount importance of “owning the flow” as the ultimate, enduring moat. If your software sits directly inside the core transaction and manages the physical distribution of goods or capital, the other advantages—your proprietary data streams and your legally accountable execution—will naturally arrive as a tightly bound, self-reinforcing package. Owning the flow acts as a powerful catalyst that rapidly accelerates your distribution networks; the raw transaction flow itself generates the exclusive, live data you need to survive, and that very same flow is the exact physical place where financial settlement and compliance tracking actually occur. Conversely, if you choose to merely rent the flow—operating as a cute, thin application layer that sits precariously on top of someone else’s infrastructure or rails—you actually hold none of these competitive advantages. You are not a real business; you are merely a transient feature, destined to be priced at exactly zero the moment the underlying foundation model beneath you undergoes its next routine product update. In a hyper-competitive global market where a hundred highly motivated teams are aggressively chasing the exact same software idea in the co-working spaces of San Francisco, while perhaps only three teams are quietly building it on the ground in São Paulo, Jakarta, or Munich, the winners will not be those with the flashiest AI models. The enduring champions of this new technological epoch will be the gritty, localized, and deeply embedded businesses that own their flow in physical markets where data, trust, and accountability are anchored deep into the ground. These are the robust, irreplaceable enterprises that artificial intelligence will not make obsolete, but will instead make infinitely stronger, more profitable, and entirely indispensable to the global economy.
