Why this entrepreneur and AI strategist thinks AI's next casualty isn't jobs, but applications

TEXT START: As anyone working closely with AI will know, conversations around autonomous digital workers outright replacing human employees are rife right now.

The Dissection

This text is recategorizing the battlefield. It shifts attention from workers to software interfaces, then positions “AI applications” as the next strategic product category. Its strongest insight is real: systems that deliver conclusions and outcomes directly can hollow out interface-heavy SaaS, dashboards, and workflows before every underlying process is fully autonomous.

But an AI application that performs a business function is not outside labor replacement. It is the delivery mechanism for it. The interface may vanish first; the human function attached to it follows.

The Core Fallacy

The central error is treating application obsolescence and job obsolescence as competing outcomes. Under the Discontinuity Thesis, they are coupled. A sales application that continuously identifies at-risk deals and prescribes actions automates analysis, reporting, and increasingly management judgment. A finance application that diagnoses margin decline automates substantial portions of analyst and planning work.

The article calls “clarity” and “better decisions” a bottleneck, as though automating those outputs preserves human necessity. Under P1, decision-making is cognitive work and therefore automatable. Under P2, firms cannot preserve stable human-only domains at scale when AI applications are cheaper and superior. Under P3, productive participation collapses. The author has identified a new format for the same substitution mechanism, not an escape from it.

The headline is therefore marketing camouflage: the application is the shell being shed; the job is part of the organism inside it.

Hidden Assumptions

• Human oversight and interpretation will remain economically necessary rather than shrinking to exception handling.
• Business data will be clean, accessible, integrated, and reliable enough to support autonomous conclusions.
• “Outcomes” are measurable and unambiguous, despite the political, legal, and strategic disputes embedded in many decisions.
• Human desire to understand a business implies continuing demand for human employment. It does not.
• The distinction between agents and applications is substantive. In practice, it is largely a product taxonomy: applications can orchestrate agents beneath the surface.
• The cited spending and adoption figures prove capability. They prove investment and attention, not durable superiority or realized productivity.
• Productivity gains will be broadly redistributed rather than captured by owners of AI capital.

Social Function

Classification: partial truth, transition management, prestige signaling, and elite self-exoneration.

It is partial truth because the application layer is a genuine target. It is transition management because it reframes functional replacement as software modernization rather than mass labor displacement. It is prestige signaling because it elevates a new category and claims strategic foresight. It is elite self-exoneration because dispossession is presented as a cleaner interface and better decision-making.

This is not pure copium. It sees the machinery moving. It simply describes the blade as a new kind of handle.

The Verdict

The article is directionally right and systemically evasive. AI applications may kill traditional software interfaces, dashboards, and seat-based SaaS. That is not a substitute for job destruction; it is how job destruction becomes operationally invisible.

Under DT, functions are automated, the software shell is unbundled, and wages lose their anchor. Sovereigns controlling models, data, compute, distribution, and the Energy–Logistics–Maintenance trinity capture the gain. A smaller Servitor class handles exceptions, trust, and infrastructure. The majority are left with consumption needs but declining productive necessity.

The text identifies the first corpse and mistakes it for the whole autopsy.