Call it disruption by degrees. The past year delivered a steady drumbeat of headlines: AI is reshaping industries, investors are piling into anything with an algorithm, and futurists argue about whether human-level intelligence is already here. Yet when you scratch beneath the headlines and into the firms that actually run the economy, the picture looks more cautious — and uneven.
Employees get the tools; businesses hold the keys
Regional surveys from Federal Reserve banks paint a consistent early-adopter story. In the Richmond Fed’s December 2025 survey, 70 percent of respondent firms said they provide employees with AI tools — often public models or company-supplied assistants — but only 56 percent reported using AI inside core operations. That distinction matters. Giving staff access to a generative assistant to draft memos or produce charts is one thing; retrofitting inventory systems, risk models or production lines to run on AI is another.
The Richmond data also show firms largely called their own timing correctly. In mid-2024 many executives expected to automate tasks with AI by 2026; by the end of 2025 a substantial share had followed through. But those implementations skew toward task-level boosts (drafting, summarizing, graphing) rather than deep process overhaul.
Some marquee moves hint at broader change: enterprises are experimenting with models that reach into workflows and search tools built to mine corporate documents. Those push-and-pull efforts — novel assistants that read your inbox or search your drive — are already reshaping how employees find answers and make decisions, even if full operational integration remains a later-stage project. For examples of the kinds of workflow integrations companies are building, see how firms are folding AI into document search and productivity tools like Gemini Deep Research in workspace products Gemini’s Deep Research plugs into Gmail, Drive and Chat.
Who’s adopting, and where the change is concentrated
Adoption is not uniform. The Philadelphia Fed’s late-October 2025 survey of Third District firms (Delaware, parts of New Jersey and Pennsylvania) found roughly half of respondents using generative AI. Service firms — professional services, finance, healthcare, and similar industries — were likelier to have adopted generative tools than goods-producing firms. Larger firms are somewhat ahead, but small and midsize companies are not far behind: adoption rates were similar across sizes, with small businesses showing more variability in plans.
That unevenness shows up in the market, too. Public markets and sector analysts have rewarded companies that can claim AI-driven revenue growth or efficiency gains, creating a “winner-take-most” dynamic in certain corners of corporate America. Some companies are going further: a handful of high-profile firms are already public about aggressive automation roadmaps — for instance, plans to automate large slices of software QA — which signal where deeper operational change might first take hold Square Enix’s QA automation plans are a concrete example.
So far, fewer layoffs than feared
One of the most politically charged questions is whether AI is causing layoffs. The short answer from recent regional surveys: not in any broad or sustained way yet. Among Third District firms using generative AI, nearly 70 percent said adoption did not change their need for workers. Another 17 percent reported a change in the types of workers needed (different skills, more hybrid roles) rather than a change in headcount. Only a small share — roughly 8 percent in that sample — reported a decreased need for workers; 2 percent said generative AI increased their need.
The Richmond Fed found a similar pattern: firms report AI primarily as an efficiency tool, not as an immediate labor-reduction device. When firms weigh reasons for adopting AI, productivity and speed repeatedly top the list; cost-cutting via layoffs ranks lower. That’s consistent with other research showing AI often augments tasks — it shortens how long a job takes or changes required skills — before it becomes a clean replacement for entire occupations.
Still, the effects are uneven across occupations and career stages. Recent studies and firm-level reports suggest early-career workers in AI-exposed roles (certain software tasks, entry-level support functions) can be more vulnerable than seasoned professionals whose jobs depend on judgment, relationships, and domain knowledge.
The gap between hype and business reality
Investors and pundits talk like we’re past a tipping point. Technologists spar over whether human-level intelligence has arrived. That debate matters for the long-term narrative, but it obscures the near-term reality: firms are experimenting heavily with task-level AI and are cautious about upending complex operations overnight. The result is a patchwork of adoption, with pockets of fast change — where data is clean, processes are modular, and scale makes automation attractive — and broad tracts of gradual, supplemental use.
There are second-order effects worth watching beyond immediate employment counts. Firms that adopt AI early are more likely to retrain staff, redesign roles, and demand different skills; those investments can widen gaps between winners and laggards. Public pressure and investor expectations push some companies to announce ambitious automation plans, while many smaller businesses treat AI as another productivity layer to be trialed cautiously.
For readers tracking the grander philosophical debate about AI’s capabilities, the arguments over whether models are already human-level or not are part of the same ecosystem of claims and counterclaims that shape corporate strategy and public policy the debate among AI experts illustrates that gulf.
What to keep an eye on
Expect more of the same mixed signals: rising use of generative tools on desktops, slower migration of AI into mission-critical backend systems, and localized labor shifts that depend on industry, firm size, and the tasks involved. Watch which firms move from task automation to operations automation — that step, not the mere granting of access to an assistant, will mark a new phase in productivity and employment impacts.
AI’s spread through corporate America will be messy, unequal, and iterative. For policymakers and business leaders, the urgent work is pragmatic: map which jobs are exposed, invest in retraining where complements are strong, and design transitions that capture productivity gains without leaving large numbers of workers stranded. Meanwhile, market incentives will keep spotlighting the early winners, and companies that engineer AI into core processes — not just into office workflows — will likely define the next wave of competitive advantage.
(If you want specific reporting on how models are reshaping workplace search tools or examples of companies publicly pivoting to AI-heavy strategies, the pieces linked above are a useful next step.)