Beyond Layoffs: The Most AI-Savvy Companies Are Adding People
IDC ·

The last year of conventional wisdom on AI and enterprise software fits in one sentence: AI does the work, companies need fewer people, and software seats shrink along with the workforce. That assumption is now colliding with the data. In IDC’s Worldwide Future of Work Survey (May 2026), 68% of Leaders (organizations reporting a positive […] The post Beyond Layoffs: The Most AI-Savvy Companies Are Adding People appeared first on IDC .
The last year of conventional wisdom on AI and enterprise software fits in one sentence: AI does the work, companies need fewer people, and software seats shrink along with the workforce. That assumption is now colliding with the data.
In IDC’s Worldwide Future of Work Survey (May 2026), 68% of Leaders (organizations reporting a positive AI ROI of 50% or more) expect the number of entry-level roles in their organizations to increase over the next 18 months. Among Laggards (organizations with no positive ROI yet), that figure drops to just 40%, and Achievers, the middle group, fall in between at 51%.
It’s hard to ignore the trend line, which clearly suggests that organizations extracting the most value from AI are planning to add people, not simply cut them. While companies across all three of these groups are also cutting certain roles, the seat thesis that AI success means fewer people is clearly being disproven. Our data show that most Leaders expect to grow their workforce.
The pattern isn’t just aimed at entry-level positions. It holds at every level. Asked what impact AI (including agentic AI) will have on the number of roles over the next 18 months, 63% of Leaders expect middle management roles to increase, versus 39% of Laggards. And for senior leadership, the split is 59% versus 36%. The most telling data point: Leaders are the least likely to cut entry-level roles, and no more than 12% expect decreased headcount at any level, with only 3% in middle management or senior leadership. While companies in all three buckets are cutting some roles, most are adding them, as these are not mutually exclusive decisions. Rather, they reflect a shift of investment into newly developed roles and responsibilities.
IDC’s Economic Impact of AI: 2026 Edition reaches the same conclusion from a different direction: AI will primarily replace work, not workers. Layoffs are real, and entry-level hiring in AI-exposed occupations has weakened. But the leaders of this market are treating AI as a capacity multiplier. They see more work to be done, not less need for people to do it.
IDC’s Two Years In — A Longitudinal View of AI-Enabled Work 2025–2026 survey (Sept 2026) finds that workload is being spread across three labor pools: permanent staff, contingent labor, and AI agents. The planning challenge is not only how many people to employ. It also includes workload orchestration: deciding which work goes to which pool, at what cost, and under what oversight. This is not a new question, but it takes on particular urgency as more organizations begin to use the language of employment when discussing investment in AI agents. While it may seem like a small semantic matter, treating “managing agents” as if they were employees is actually a much more serious concern. Why? Because referencing agents as an equivalent to human headcount discounts workers as a line item in a budget. Agentic AI is a valuable resource, but counting agents as headcount equivalents measures the wrong thing. Agents are better understood as capacity than as colleagues. The real question is how work gets done, and how work gets allocated across the three pools, not who or what performs each task.
The software industry has been here before. When the cloud arrived, the model shifted from perpetual licenses to subscriptions, and the seat became the unit of value because the product was something a person logged into. That logic held for more than two decades. It is now breaking down because a growing share of the work is being done by agents that never take a seat.
Meanwhile, the application stack still assumes one labor pool at a time. Workday manages employees. SAP Fieldglass manages contingent workers. Agent platforms such as Salesforce Agentforce manage digital labor. Each is strong inside its own boundary. None was built to answer the question leaders now face: what work needs doing, who or what should do it, what it costs, and how to optimize across all three?
This is precisely why HCM, workforce management, workflow automation, and agent management are converging. The vendors that connect them will not be selling a feature. They will be defining a category. This is an emerging opportunity to double down on cross-functional workflows that pull focus from an individual doing a task to a cohort of interdisciplinary teams achieving a business outcome.
Pricing follows the same logic. If Leaders add people and deploy agents at the same time, seat count becomes a weaker proxy for the value software delivers. Swapping per-seat fees for per-agent fees does not solve the problem either, because it only changes what is being counted. The durable metric is work completed. Vendors that can measure and price against workload, across humans and agents alike, will hold the advantage. The important thing is to support clients’ focus on long-term value: driving revenue gains via topline growth, not just productivity and short-term efficiency gains.
Vendors:
Tech Leaders:
Headcount will continue to dominate the headlines, but it is not the most important number to watch. Instead, pay close attention to the distribution of work: what share goes to permanent staff, what share to contingent labor, what share to AI agents, and what each unit of that work costs and delivers. IDC data show that most Leaders expect an increase in roles at every level, and they are redistributing work and instituting governance that enables greater oversight.
Leaders are building the organizational structure around AI, not just deploying it. Among Leaders, 63% have a fully operational AI center of excellence, compared with just 15% of Laggards. They are also investing in new job roles, and they trust agents with more: 33% of Leaders run agents fully autonomously within guardrails, against 20% of Laggards. Governance comes first, and autonomy follows.
That is not a staffing adjustment. It is a structural change in how enterprises operate, and the software that runs them has not caught up. The window to claim the orchestration layer is narrowing. Vendors that keep counting seats will be measuring a number their most AI-successful customers are already moving past.
Count the seats, and you see the past. Count the work, and you see the market.
The post Beyond Layoffs: The Most AI-Savvy Companies Are Adding People appeared first on IDC .