Software is Eating Labor

Episode Summary

Executive Summary: Alex Rampell argues that software’s next major frontier is not the $300B SaaS market but the $13T U.S. labor market. He traces software’s evolution from digitizing filing cabinets into databases to AI systems that can now perform end-to-end work, enabling outcome-based pricing, new business models, and expansion into labor-heavy industries worldwide.

Main Topics: Software’s shift from records to outcomes (Priority: 5/5): The talk reframes software as moving beyond storing information to actually completing tasks previously done by humans, such as support, sales, collections, and operations. The labor market as the real prize (Priority: 5/5): Rampell contrasts the size of SaaS with labor spend, arguing that AI vendors are competing for labor budgets rather than traditional software budgets. Historical pattern: filing cabinets to databases (Priority: 4/5): He reviews how major software categories emerged by digitizing physical record systems in travel, CRM, ERP, libraries, legal, accounting, health records, and HR/payroll. Outcome-based and usage-based pricing (Priority: 5/5): Because AI can do the work, pricing shifts from seats and records to completed tasks, calls, customers acquired, claims resolved, or collections collected. AI enables new markets and makes bad businesses viable (Priority: 4/5): AI reduces CAC, COGS, and staffing friction enough to make previously uneconomic marketplaces and niche vertical software opportunities possible. Operational advantages of AI workers (Priority: 5/5): Rampell highlights AI strengths in intermittent demand, demoralizing jobs, multilingual service, and regulatory consistency—areas where humans are costly or unreliable. Real-world examples from portfolio companies (Priority: 4/5): Examples such as Happy Robot and Salient show AI agents negotiating freight rates and collecting payments, demonstrating that these models are already operating in the market.

Key Arguments: The U.S. labor market is vastly larger than the global SaaS market, so the biggest opportunity for software is to capture spend currently allocated to labor rather than seats. Most legacy software categories began by converting physical filing cabinets into databases; AI now extends that transformation to executing the actual work around those records. If AI can complete a support, sales, collections, or operations task, charging per seat becomes obsolete; vendors should charge for outcomes instead. AI can be more efficient than humans in jobs with intermittent demand, high emotional toll, language complexity, or strict compliance requirements. Verticals that previously looked too small for software become attractive when AI lowers the cost of acquiring customers and delivering service. AI does not only replace jobs; it expands the number of solvable markets by making end-to-end service economically feasible. Software companies that already control the system of record can potentially move up the stack and own the workflow, not just the database. Human-centered pricing and staffing assumptions are being destabilized; companies may pay AI vendors more than old software fees but still far less than labor costs.

Data Points: Worldwide SaaS market: $300 billion per year - Used to contrast traditional software opportunity with the labor market Worldwide software market cap: $2.2 trillion - Cited as the value created by digitizing records and workflows U.S. labor market: $13 trillion - Presented as the true target software is now going after Nurses’ annual wages in the U.S.: $650 billion - Used as an example of a labor pool bigger than the entire worldwide software market Registered nurses in the U.S.: 4.5 million - Supporting scale of the nursing labor market Zendesk seat price: $115 per month - Example of legacy SaaS per-seat pricing Zendesk 1,000-seat annual software cost: About $1.4 million per year - Illustrates software cost versus labor cost in customer support Fully loaded human support cost: $75,000 per worker per year - Used in a 1,000-agent support-center example 1,000-person support-center labor cost: $75 million per year - Compared against software spend to show labor dominates Human cost per answer: About $37 - Roughly derived from support labor economics in the example Software cost per answer: 69 cents - Derived from Zendesk-style pricing in the support example Total cost per answer: $38 - Combined human and software cost in support operations Average annual pay for plaza lane optometry receptionist: $45,000 per year - Used to show a small business labor budget AI could undercut AI replacement offer for receptionist work: $20,000 per year - Illustrative price for an AI agent doing much of the job California job posting duration: 6 months - Shown to demonstrate persistent labor shortages in a small-business role AI collections example balance past due: 51 days past due for $825.35 - From Salient’s collections call demo Black Friday staffing issue: Intermittent demand - Explained as a key reason AI is useful for seasonal labor spikes

Pivotal Quotes: "“the labor market, this is almost obvious, is so much bigger than the software market.”" — Alex Rampell: Core thesis comparing software revenue to labor spend "“almost every software company has basically taken a filing cabinet and turned it into a database.”" — Alex Rampell: Describes the historical pattern of software category creation "“We’re not giving you software. We’re going to do a job for you.”" — Alex Rampell: Explains the shift from SaaS tools to outcome-based AI services

Implications: AI-native companies can sell outcomes, not seats, and target labor budgets in huge verticals. Expect pricing to move toward work completed, more automation in regulated and multilingual tasks, and a wave of new businesses previously blocked by labor economics.