How Firms Can Reallocate Work Without Cutting Total Hiring

The industry counterfactual shows that coder employment lagged the path implied by the industries where coders work. The next question is organizational: what could a firm actually do that produces slower coder hiring while leaving its total number of job postings unchanged? This chapter uses a stylized firm model to connect AI adoption, task productivity, demand, hiring frictions, and occupational mix. The mechanisms are plausible interpretations of the evidence, not coefficients directly identified by the two Federal Reserve studies; that distinction is essential when moving from an observed pattern to an explanation.

A firm does not hire “technology” or “employment” in the abstract. It hires people to perform a changing bundle of work under expectations about demand, costs, and future capability. AI can alter that bundle even when the firm’s total recruiting plan remains steady.

The firm can change its mix before it changes its total

Imagine a firm that expects to create twenty positions over the next year: ten involving software development and ten involving sales, operations, security, and customer support. After adopting AI tools, the firm may decide that its existing coding team can handle the planned product work with fewer additional hires. It might post six coding roles instead of ten and use the remaining four openings for security, implementation, customer success, or a new product team. Total postings stay at twenty while the path into coding becomes less favorable.

This is not a report that firms in the Federal Reserve sample made exactly that substitution. It is a way to interpret what a total-posting result leaves open. The FEDS Note finds no evidence that higher AI adoption reduced firms’ total online job postings through November 2025. The authors explicitly distinguish that outcome from the experience of occupations that may be particularly susceptible to automation. A stable total is compatible with a changing internal allocation.

The same result could arise without a deliberate transfer from one role to another. A firm might keep its coding headcount stable but reduce replacement hiring when a developer leaves. It might merge a junior coding position into a broader product role. It might ask an analyst, designer, or operations specialist to perform some AI-assisted coding that previously belonged to a programmer. The posting measure would show a smaller coding signal, but the total number of openings could remain unchanged or even rise.

A useful sequence is therefore:

sequenceDiagram participant F as Firm participant T as Coding tasks participant M as Management participant C as Coder hiring participant O as Other hiring F->>T: Adopt or plan AI use T-->>M: Tasks become faster or are reorganized M->>M: Assess realized or expected demand alt Fixed or weakly expanding coding demand M->>C: Slow or hold new coder hiring M->>O: Shift openings to other priorities else Expanding coding demand M->>C: Maintain or increase coder hiring M->>O: Add complementary roles end C-->>F: Occupational mix changes O-->>F: Total postings can remain non-negative M-->>F: Long-run demand expansion remains unresolved

The branches matter.