AI Agent Deployments Nearly Triple as Entry-Level Hiring Slows

The tripling is real, but it measures depth, not reach

The headline number is hard to miss. Salesforce's 2026 Agentic Enterprise Index draws on Agentforce usage data from February 2025 through April 2026. It finds that the average number of AI agents each organization deploys has nearly tripled over the past year.2 The skill count per agent went from about two in early 2025 to six. Employees' interactions with agents rose 300% over the same period.2 Put in absolute terms, the average customer went from 5 agents to 13 in about 14 months.4

Other data points the same way. Gravitee surveyed 750 executives and technical leaders. It found the number of agents running inside a typical enterprise roughly doubled between December 2025 and April 2026, with the most common deployment growing from 26–50 agents to 76–100.7 KPMG's third-quarter AI Pulse, which covers U.S. companies with at least $1 billion in revenue, found that the share building multi-agent systems quadrupled in one quarter, from 6% to 25%.8 Microsoft says active agents in Microsoft 365 grew 15-fold year over year, and 18-fold in large enterprises.16 Vendors are reporting the same pattern in their own numbers. ElevenLabs says weekly conversations on its agent platform more than tripled after February, passing 15 million.3

These figures need a careful reading, though. Most of the "tripling" data counts how heavily companies that already use agents are using them. It does not count how many companies have started. That matters for any claim about the workforce. Stanford's 2026 AI Index found that 88% of organizations use AI in some form, yet agent implementation is still below 10% in nearly every business function.10 Cisco surveyed major enterprise customers and found 85% experimenting with agents but only 5% running them in broad production by its stricter definition.9 McKinsey's survey is in between. About two in ten respondents say they are scaling agents across the company.9

The sources do not really contradict each other. They measure different things, and once that's clear the picture is simple. A group of mostly large companies is moving quickly while everyone else moves slowly. McKinsey found that 40% of businesses with more than $1 billion in revenue had reached the scaling stage in at least one function, up from 27% a year earlier. Smaller companies stayed around 22%.911 The tripling is a story about intensity among early adopters, and it should be read that way.

Broad adoption keeps rising

General AI use is also climbing quickly. The New York Fed's August 2026 regional surveys found 61% of service firms using AI, up from 25% in 2024. Among manufacturers the figure was 51%, up from 16%, which is roughly a tripling.56 A U.S. Chamber of Commerce report released October 2 found that 66% of small businesses use AI, nearly three times the 23% that did in 2023.1 Among workers, Gallup's May 2026 study of more than 22,000 employed U.S. adults found 52% using AI in their role and 15% using it daily.16

The New York Fed data also includes a warning that the headline figures leave out. Among service firms that use AI, the median company has only 17% of its workers on the tools. For manufacturers the figure is 7%.5 In other words, most firms have given access to a small group of employees rather than changing how the whole company works. Salesforce has a version of the same problem. Coverage of its Agentforce rollout reports that only 34% of its customers have fully adopted the platform, mostly because their data is messy or scattered.4

What the labor data shows so far

The labor data is where the agent story gets complicated. Companies are deploying more automation, but direct job losses have stayed small so far.

In the New York Fed panel, 4% of AI-using service firms said they had laid off workers because of AI, up from 1% in 2025. No manufacturer reported AI-driven layoffs in either year.5 McKinsey's 2026 survey found that 14% of respondents at AI-using organizations say AI contributed to a smaller workforce over the past year. A year earlier, 32% had predicted cuts, so actual reductions came in at less than half of what was forecast.1113 A Census Bureau analysis found AI-related employment decreases at just 2% of firms.11 The Yale Budget Lab's analysis through the first quarter of 2026 found no clear aggregate effect on employment or wages.12

Layoff announcements suggest more strain. Challenger counted 116,175 announced U.S. job cuts attributed to AI from January through August 2026, out of 529,914 cuts for all reasons.12 That is a large share, but it reflects the reasons employers give. It is not a verified count of jobs replaced by software.1215 Some companies may find it useful to describe ordinary cost-cutting as AI efficiency, so these numbers mark the upper end of the possible impact rather than a measurement.

The more convincing evidence of disruption is in hiring. In the same New York Fed panel, 15% of AI-using service firms said they had hired fewer people because of AI, compared with 4% that cut staff and 13% that hired more.512 Stanford Digital Economy Lab analysis of ADP payroll records through June 2026 found employment among 22- to 25-year-olds in highly AI-exposed occupations running about 19% below the trend of less-exposed peers. The gap came mainly from fewer hires, not more departures. Experienced workers showed no gap.1213 Anthropic's March 2026 labor research found the same pattern. Unemployment did not rise systematically among highly exposed workers, but there were signs that hiring of younger workers had slowed in exposed fields.1520

My reading is that the agent boom is not yet putting current employees out of work in large numbers. It is narrowing the path into white-collar work. A company that triples its agents does not need to lay anyone off to change the labor market. It only needs to stop filling some of the junior roles that used to be how people entered a profession.

The exposure is in white-collar jobs

The occupations where agents are spreading fastest are the ones showing the most strain. Agents have the clearest production use in coding, customer service and IT support. Those are areas where success is easy to check: code passes tests, and tickets get resolved.9 Anthropic's usage-based measure ranks computer programmers as the most exposed occupation, with 75% of their tasks covered, followed by customer service representatives and financial analysts near the top.1320 Intercom says its Fin agent averages a 76% resolution rate across about two million customer queries a week.9

That helps explain why economists are increasingly focused on degree-holders. Indeed Hiring Lab's third-quarter survey of more than 120 economists found growing expectations that AI will push down the median real wages of college-educated workers. The question about which group's wages AI would hurt more showed the largest quarter-to-quarter shift in the survey.14 The same panel had no consensus on net employment effects. 45% called AI a minor negative for jobs, 25% a minor positive, and 21% said it had no net effect.14 Panelists expected the biggest losses in administrative assistance, software development, and data and analytics. Software and data also appeared on their list of the biggest gainers.14 Those fields are changing internally, with some roles shrinking and others growing.

Some new jobs are appearing too

The employer surveys that report job creation deserve attention, but also some skepticism. In the Chamber's report, 47% of small businesses said AI is creating jobs today, while 6% said it is enabling headcount reductions.1 Earlier IDC and Salesforce research found that 23% of employers created new roles because of AI, against 14% that cut positions.2 Indeed reports that 5.9% of U.S. job postings were AI-related in June 2026, well above the 2022 peak of 3.3%.17 PwC puts the average wage premium for AI skills at 62%. PwC also sells AI consulting.1318

The new demand looks real, but it does not obviously go to the people losing out. A recent graduate who missed an entry-level analyst job is not automatically qualified for a role building agents.

Governance and management are behind

The adoption data also shows companies that are not ready for what they are deploying. Gravitee found that average monitoring coverage of agents rose only from about 47% to about 52% while the number of agents doubled. Only 9.5% of organizations secure more than 80% of their agents.7 Deloitte found that 74% of leaders expect to use agents at least moderately by 2027, but only 21% have a mature governance model for agentic AI.8 Microsoft's Work Trend Index found that organizational factors explain about two-thirds of the AI impact employees report. When managers openly use AI themselves, the value employees report rises by 17 points.16

That is why the idea of every employee becoming a manager of AI agents matters more than it might seem.2 The job that is growing is supervising agents: checking their output, catching errors, and deciding what to hand off. In Microsoft's data, 86% of AI users treat AI output as a starting point rather than a final answer, and half name quality control as a critical skill.16

Where this leaves workers

Public anxiety is moving ahead of the measured effects, but the direction of that worry fits the data. Pew found 71% of U.S. adults expect AI to reduce jobs over the next 20 years, up from 64% in 2024. Among adults under 30 the figure is 73%.13 McKinsey respondents expecting AI-related headcount cuts next year rose to 39% from 32%.11 Past expectations have run about twice as high as the cuts that actually followed, however.18

Overall, agent deployments have tripled where companies were already invested, while broad use across the economy is still shallow. The labor effects are real but limited, and they are showing up mostly as fewer hires for young workers rather than as visible layoffs. Watch entry-level hiring and the wage data for degree-holders in exposed fields. Those numbers will show whether the agent boom is shrinking the white-collar workforce or reorganizing it.