Summary

AI turns software from a claim on IT budgets into a claim on labor spend because the product can now perform the work. David Haber’s defining example is software that speaks 50 languages, compliantly, 24/7; Alex Rampell’s is even broader: “I’ve never been able to hire somebody for a dollar. Now I can hire software for a dollar.” That should create new consumption rather than simply eliminate jobs.
The AI capability differentiates a product, but it does not by itself defend the company. Haber’s durable moats remain familiar: owning the end workflow, owning the context in which it is applied, becoming the system of record, generating network effects, and embedding deeply enough that the customer depends on the product. “AI is an incredible tool for differentiation,” but its ubiquity makes it weak as a standalone moat.
AI lowers software-production costs while making the race to defensible scale more brutal. Rampell’s anti-fraud analogy: four customers versus three proves little, but four billion observed customers versus one billion can produce a real data advantage. With “nine million ankle biters” competing around obvious ideas, momentum matters because it offers the best path to “gravitational scale.”
Per-seat SaaS faces a pricing-model problem, not necessarily an extinction event. Adobe or Zendesk may sell fewer seats when AI reduces the labor attached to them, yet they could potentially quadruple revenue by charging for outcomes. The more credible disruption is concentrated where wall-to-wall licenses are expensive and underused; software whose payment is tied directly to actual usage, like payroll, is much harder to rationalize away.
The best entry markets combine greenfield customer creation with patient founders. ADP and Paychex inhabit a “Goldilocks zone of irrelevance”: payroll fees are too small relative to payroll itself to justify switching, making entry difficult and retention excellent. A new EHR faces the opposite problem—almost no new hospital systems are created—so even superior software has no clean beachhead.
AI features can reach meaningful revenue unusually fast, but they must still backfill into products and companies. An orthodontic receptionist may look like a feature layered on existing software yet command $20,000 annually because it replaces labor; Rampell’s warning is that “the feature has to backfill product, backfill company as quickly as possible.” Haber’s “messy inbox” wedge shows the path from ingesting email, fax, and phone data into owning downstream scheduling, benefits, and eventually the system of record.
Platforms and incumbents remain advantaged, but the vertical opportunity is too broad for one provider to absorb. OpenAI can pursue five billion ChatGPT users, the developer back end, coding, and large-enterprise deployments without building every obscure vertical workflow. Consolidation should still punish undifferentiated number-three-through-number-100 players, while incumbents that preserve distribution and adopt AI may turn labor replacement into higher margins rather than disruption.

Deep dive

1. AI changes the market size, not the anatomy of a moat

Haber’s opening distinction is load-bearing: “AI is an incredible tool for differentiation,” but the AI itself is not defensibility. Speaking 50 languages compliantly, 24/7, beats human labor; owning the workflow, context, system of record, network effect, and customer dependency keeps competitors out.
The structural break is that software can now perform the work. Its addressable market is therefore “no longer just IT spend. It’s largely labor,” opening categories whose customers historically lacked meaningful software budgets but already carried substantial operating costs.
Rampell’s anti-fraud analogy explains why many data moats remain invisible early. Gravity technically exists at the atomic level but becomes noticeable only at planetary scale; likewise, seeing four customers rather than three means little, while seeing four billion customers rather than one billion can materially improve fraud decisions.
That creates the zero-to-one trap: AI makes obvious software easy to produce, filling markets with “nine million ankle biters,” yet defensibility often emerges only at mega scale. A startup needs differentiation and momentum before it possesses the scale advantage it ultimately promises.

2. Per-seat pricing is exposed, while workflow replacement embeds deeper

Rampell sees two fears behind public software weakness: AI may reduce seat counts, and customers may vibe-code substitutes. Adobe needs fewer designer seats if fewer designers are hired; Zendesk needs fewer support seats if software answers the tickets.
The first fear is real but does not imply collapse. Companies could charge per outcome and potentially quadruple revenue; the problem is replacing the psychologically accepted “tall, grande, venti model” of per-seat, per-month pricing with something customers still perceive as fair.
The build-it-yourself thesis has produced less evidence. Even feature-bloated incumbents contain edge cases customers do not anticipate: “Why don’t you grow your own food or weld your own aluminum or build your own house?” Comparative advantage still favors buying mature software despite Salesforce’s cited 80% gross margin and the argument that a customer’s ability to build software is the opportunity in that margin.
Haber adds that software replacing a team may become more deeply embedded. Whether switching that software is harder than rehiring the team remains “an open question,” but the customer now depends on the product to operate the business, not merely to assist employees.

3. The Goldilocks zone protects incumbents and constrains greenfield attacks

Rampell’s “janitorial services problem” captures the ideal incumbent position: offering a giant company 9% cleaner toilets and 1% savings is too immaterial even to route internally. “The problem is it’s hard to get in. The good news is it’s hard to get out.”
ADP and Paychex fit that “Goldilocks zone of irrelevance.” Payroll involves taxes, counties, time spent in New York, garnishments, and child support; a paltry per-person fee—Rampell guesses $50 per month, perhaps $100—is small relative to the overall payroll, so customers rarely switch.
Seat licenses invite more scrutiny. A company shrinking from 1,000 employees to 200 sees 1,000 Salesforce licenses at $100 monthly as $1.2 million annually; unlike payroll delivery, payment is not inextricably linked to actual usage, so wall-to-wall licenses become early cost-cutting targets.
Torenberg’s pushback—where can new software actually replace incumbents?—draws two conditions from Rampell: patient founders and abundant new customers. A new payroll company needs a high rate of new-company creation to work; new EHR vendors face almost zero newly created hospital systems and must pursue $5 million deals with hospitals already using Epic or Cerner.

4. Frontier fluency gets attention; industry context turns it into a company

Haber’s steelman for speed and brand starts with noise: standing out matters more when many teams can build. Younger, more technical founders may lack industry roots but can live on the model frontier; they must then hire context early because “context is king.”
Eve is his example. Its founders came from Rubric rather than plaintiff law, then applied document extraction, voice, and LLMs to plaintiff-law workflows while hiring plaintiff attorneys on staff to understand each new model’s impact on drafting and case reasoning.
Eve’s contingency-fee market also aligns technology with economics. In other areas of legal, making an employee 50 times more efficient can erode billable-hour revenue; a firm paid only when it wins can turn 5x efficiency into 5x the clients without that conflict.
Rampell’s complementary steelman is brand plus scale. Momentum is not itself a moat, but it offers “the highest chance of getting you to gravitational scale”; with 20 equivalent competitors, a weak slope is fatal because “you can’t hand-crank out the Cheerios” while the leader builds the factory.

5. Labor economics let narrow features become substantial businesses

Haber notes that a “GPT wrapper” is risky when model and application capabilities overlap. Yet AI makes previously unattractive verticals investable because the budget comes from labor: plaintiff law and nonbank auto-loan servicing can support companies even without historical millions in IT spend.
Rampell’s old feature-product-company hierarchy still applies, but feature revenue has changed. A front-office agent for an orthodontic clinic can command $20,000 annually because it fills a job; customers buy the immediate fix, so “the feature has to backfill product, backfill company as quickly as possible.”
Salient carries the economics: voice agents service auto loans across 50 states in 50 languages, compliantly and 24/7, while collecting meaningfully more than the displaced labor. The capability begins as a feature atop existing systems but taps an operating budget far larger than a conventional software add-on.
Haber’s “messy inbox problem” supplies a repeatable expansion path. His example starts above the system of record, extracting patient data from email, fax, and phone into an EHR, then moves into scheduling, prior authorization, and eligibility and benefits—potentially using that wedge to become an end-to-end platform and eventually, perhaps, the system of record.

6. Platforms threaten applications through competition and arbitrary taxation

Haber argues OpenAI is unlikely to build every obscure vertical workflow, such as a dental-clinic assistant, and that vertical applications may add value by orchestrating work across several model companies. That plural model layer may limit any one foundation company’s ability to absorb the entire application stack.
Rampell’s platform warning has two branches: the owner can compete, or it can tax applications “at my fancy,” moving from 10% to 40%. Multiple foundation models improve the position versus Windows’ dominance, but applications must still judge whether their workflow is important enough for the platform to enter.
His spreadsheet history shows the danger: VisiCalc held 100% in 1979; Lotus 1-2-3 reached roughly 70% by 1985; Microsoft released Excel for the Mac that year. Because spreadsheets were central to buying business computers, Rampell uses the example to illustrate how a platform owner can win.
Yet large platforms prioritize nearby “gold bricks.” A Facebook executive rejected Rampell’s payments pitch because Facebook had hundreds of easier opportunities at its feet; obscure vertical workflows can therefore remain available for years, and AI makes those distant gold bricks larger because they substitute for labor.

7. Scale, consolidation, and incumbent distribution still govern the endgame

Rampell’s OpenAI playbook is to become both the consumer default and the back end for builders: take ChatGPT from 800 million weekly active users toward five billion and serve developers. Even a Gemini 3 that might be five times better could still struggle to dislodge an established habit.
Haber expects selective horizontal expansion: coding and IDEs, including Google’s Antigravity launch, plus Palantir-like forward-deployed work for large enterprises that understand AI’s potential but not where to start. Anthropic’s interest in financial services is an early signal of that consultative path.
Consolidation remains conventional. With 20 equal competitors, prices approach zero—or “the price of electricity”—until the bottom 15 go bankrupt or leaders acquire rivals; Rampell invokes Jack Welch’s rule that there is no value in being number three through 100. Model providers face the harshest version when they are “state-of-the-art minus minus minus,” though Rampell leaves room for specialization by modality and customer.
AI’s consensus status favors incumbents more than cloud or mobile did: nobody is scoffing, and every system-of-record vendor can add a button. Tata, Wipro, or Infosys might automate a 100,000-person call center while preserving the customer integration—or lose it to a startup—but Rampell’s broader point is that when software costs a dollar, businesses buy tasks they never economically hired humans to perform.

China's AI Upstarts: How Z.ai Builds, Benchmarks & Ships in Hours, from ChinaTalk Bitcoin Reclaims $90k: Have We Bottomed?