The ongoing debate surrounding the economic consequences of artificial intelligence is increasingly shifting from theoretical projections toward rigorous empirical evidence. Two separate studies released in September 2026 provide granular insights into actual corporate practice across the United States. A large-scale empirical study analyzed real budget and staffing records from 21,559 US companies, while the Federal Reserve Bank of Philadelphia investigated operational adoption within smaller enterprises. Together, both publications paint a nuanced portrait that challenges widespread assumptions regarding imminent job cuts and immediate profit surges.
The empirical investigation into corporate spending and headcounts, published via SSRN on September 13, 2026, revealed an unexpected trend among frontrunners. Organizations categorised as high-intensity adopters, which make continuous and substantial investments into AI architectures, recorded an average of roughly 10 percent higher net employment growth than comparable control groups. This workforce expansion was concentrated in enterprises that integrated models deeply into core operations rather than limiting their efforts to superficial pilots. Within the information technology and information sectors in particular, sustained capital deployment consistently correlated with broad staff expansion.
Equally striking was the functional distribution of newly created positions across these expanding enterprises. Countering fears that automation tools primarily displace entry-level talent, the strongest personnel growth occurred precisely in entry-level roles. Alongside junior openings, the surveyed firms hired heavily in software engineering, commercial sales, and customer service departments. The researchers suggested that productivity enhancements unlock additional demand for specialized services, prompting firms to scale human capacity to deliver broader project scopes.
A contrasting perspective emerges from the Federal Reserve Bank of Philadelphia in a specialized survey authored by Theresa Dunne and Adam Scavette, published on September 22, 2026. Conducting a special evaluation under the Small Business Credit Survey framework, the authors examined firms operating within the Third Federal Reserve District. The adoption rate among these small businesses stood at 37 percent, lagging behind large corporate peers by roughly 10 percentage points. In terms of primary tasks, 79 percent of active adopters utilized systems for drafting text and marketing collateral, 56 percent for personal assistance and meeting summaries, and 51 percent for planning and research.
Crucially, the Philadelphia Fed survey revealed a stark divergence between perceived operational efficiency and audited balance sheets. While 71 percent of adopting small businesses reported subjectively higher day-to-day productivity, over 70 percent documented no measurable change in their labor expenses, external contractor fees, or top-line revenues. The efficiency gains largely remained confined to individual time savings during routine workflows, failing to translate into improved operating margins or tangible cost reductions.
Synthesized together, these findings indicate that the economic yield of modern software depends fundamentally on organizational integration depth. Large corporations implementing profound operational changes can generate substantial scale effects, resulting in net hiring increases across junior and specialized technical tiers. Conversely, smaller enterprises often remain trapped at an ad-hoc stage where personal conveniences do not alter economic fundamentals. Achieving demonstrable returns on AI expenditures therefore demands comprehensive process redesign rather than passive adoption of generic assistants.