Orlando Bravo is half-right, which in the world of private equity, is the most dangerous kind of right.
The Thoma Bravo co-founder recently suggested that certain software companies deserve a valuation haircut because AI is "disrupting" them. It’s a comfortable narrative for the CNBC crowd. It suggests a neat pruning of the garden—clip the dead branches, and the rest of the tech ecosystem continues its skyward climb.
That is a fantasy.
The market isn't facing a "valuation cut" for a few laggards. We are witnessing the total evaporation of the "moat-by-friction" business model. For a decade, SaaS companies scaled by selling complexity disguised as "solutions." They charged for seats, for storage, and for the privilege of navigating a UI that required a six-week certification.
AI doesn't just lower the barrier to entry for competitors. It destroys the economic justification for the middle-manager software layer entirely. If you think the "valuation cut" is limited to legacy names, you aren't paying attention to the fundamental shift in how value is captured in a post-LLM economy.
The Margin Compression Trap
The "lazy consensus" says that AI will make software companies more efficient. The theory goes: engineers code faster, support tickets get handled by bots, and margins expand.
Wrong.
In a competitive market, efficiency gains are always passed to the customer. If it costs 90% less to build and maintain a CRM, the price of a CRM will eventually drop by 90%. We are heading toward a world of "Software at Cost."
The incumbents are terrified because their entire valuation is predicated on high-margin recurring revenue. But when the cost of production hits near-zero, "recurring" starts to look a lot like "charity." Why pay Salesforce $150 per user per month when a specialized agentic workflow can query a raw database and deliver the same results for the price of a few million tokens?
I’ve spent years watching enterprise sales teams justify six-figure renewals by pointing to "ecosystem lock-in." That lock-in is a wall made of ice in a warming climate.
The Fallacy of the Proprietary Data Moat
The second big lie is that these software companies are "sitting on a goldmine of data."
Bravo and his peers love this one. It’s the ultimate fallback. "Sure, the UI is clunky, but they have twenty years of proprietary customer data!"
Data is only a moat if it’s clean, structured, and unique. Most enterprise data is a swamp of duplicate entries, "Test Test" profiles, and fragmented silos. More importantly, LLMs are proving that they don't need your specific, dirty data to understand your business logic. They can infer it from general patterns.
The "data moat" is the new "patent thicket"—a legal and psychological comfort blanket that provides zero protection against a 22-year-old with an API key and a better way to solve a specific business problem.
Stop Asking if a Company Uses AI
People always ask: "How is this company integrating AI into their product?"
That is the wrong question. It assumes the product should exist in the first place.
The real question is: "Does AI render the core problem this software solves irrelevant?"
Take "Customer Success" software. For years, we bought platforms to help humans manage other humans. If AI handles the customer relationship directly, the management software is a vestigial organ. It doesn't matter how many "AI features" you bolt onto it. You’re putting a jet engine on a horse-drawn carriage. The horse is still the bottleneck.
The Death of the Seat-Based License
The most violent disruption won't be in the code; it will be in the billing.
The SaaS explosion was fueled by the per-seat model. It was a beautiful, scalable lie. It decoupled value from cost. But AI is an "un-personing" technology. If one person can do the work of ten using an autonomous agent, the software company’s revenue drops by 90% under a per-seat model.
To survive, software companies have to pivot to consumption-based or outcome-based pricing. But Wall Street hates consumption pricing because it’s volatile. It’s not "predictable recurring revenue."
This is the catch-22. To stay relevant, software companies must destroy the very pricing models that gave them their unicorn valuations. If they don't, a "born-in-AI" competitor will charge by the outcome and bankrupt them.
The "Service-as-Software" Pivot
We are moving from Software-as-a-Service (SaaS) to Service-as-Software.
In the old world, you bought a shovel (software) and hired a guy to dig (labor).
In the new world, you just buy the hole.
The companies that will actually hold their valuations are those that take full responsibility for the outcome. If you are a legal tech company, don't sell me a platform to manage contracts. Sell me the contract. If you are a marketing tech company, don't sell me a dashboard. Sell me the leads.
This requires a level of risk and liability that most "high-margin" software execs are too cowardly to touch. They want the 80% margins of a software business without the 20% headaches of a service business. Too bad. Those days are over.
The Reality of the "Valuation Cut"
Bravo talks about a cut. I’m talking about a reset.
The 20x revenue multiple was an anomaly born of zero-interest rates and a lack of technological competition. In a world where software is easier to build, easier to replace, and harder to monetize, the natural multiple for a standard SaaS company isn't 10x or 15x. It’s 3x or 4x.
It’s becoming a utility. And utilities don't get tech multiples.
If you’re holding a portfolio of "steady" software names because they have "great enterprise relationships," you’re holding a collection of expensive rolodexes. Relationships matter until the cost differential between the incumbent and the AI-native alternative becomes an order of magnitude. At 10x cheaper, the "relationship" is the first thing the CFO cuts.
The Counter-Intuitive Play
If you want to find the winners, look for the companies that are actively cannibalizing their own revenue.
Look for the CEO who says, "We are killing our per-seat pricing and moving to an outcome-based model, even though it will tank our revenue for four quarters." That is the only person who understands the gravity of the situation.
Everyone else is just rearranging deck chairs on a very expensive, very digital Titanic.
The "valuation cut" isn't a temporary dip or a sector rotation. It is a fundamental repricing of what human intelligence—and the tools that facilitate it—is actually worth.
If your software is just a middleman for data entry and retrieval, your valuation isn't "hit." It's gone. You just haven't realized it yet.
Go back and look at your "safe" software bets. Strip away the marketing jargon. If the core value proposition is "saving time for a human," and that human is no longer necessary to the process, you aren't investing in software. You're investing in a ghost.
The haircut is going to be a lot closer to the scalp than Orlando Bravo wants to admit.
Sell the "AI-enhanced" incumbents. Buy the companies that make software invisible.
