Boards want productivity from AI and are tempted to get it the fast way: reduce headcount. Workers want AI to take the drudge work and open better paths. Managers sit in the middle, already overloaded. This is where the future of work will be decided. Employers say they plan to use AI both to expand capacity and to cut roles, and early evidence shows the brunt of disruption is landing on the most junior employees (Microsoft, 2025; WEF, 2025; Brynjolfsson, Chandar and Chen, 2025).
What the labor data already say
A large study using administrative payroll records finds that early‑career workers, ages 22 to 25, in occupations most exposed to AI saw a roughly 16% relative decline in employment. Experienced workers in the same jobs, and workers in less exposed fields, did not see comparable declines (Brynjolfsson, Chandar and Chen, 2025). The adjustment so far is happening through hiring and headcount rather than wages or compensation. The same analysis shows that employment losses concentrate where AI automates human tasks, not where it augments them. That distinction, automation versus augmentation, is the best predictor of near‑term impact in the data. It is a working paper, not yet peer‑reviewed, but it is based on large‑scale administrative data and the results remain robust to alternative explanations (Brynjolfsson, Chandar and Chen, 2025).
This creates a practical dilemma. If firms lean on AI to automate entry‑level tasks without designing new ways for juniors to learn and contribute, they risk a broken pipeline. The people who most need on‑the‑job practice and coaching are the ones whose tasks are easiest to hand to software. That is not a claim about AI in the abstract. It is a pattern visible in employment, right now, for the cohort that would normally be learning the ropes (Brynjolfsson, Chandar and Chen, 2025).
Management is the bottleneck
The same organizations that plan to expand AI use are not set up to absorb the change. In a global survey of nearly 10,000 business and HR leaders, managers reported spending almost 40% of their time on firefighting and administration, and only 13% on developing their people. More than a third feel unprepared for the people‑management part of the job (Deloitte, 2025). Meanwhile, employers in the United States were advertising 42% fewer middle‑management roles at the end of 2024 than in spring 2022, and Gartner, as cited by Deloitte, projects that by 2026, 20% of organizations will use AI to flatten their structures (Deloitte, 2025).
Flattening can remove bureaucracy. It can also remove the layer that teaches, sponsors, and corrects. If those coaching functions do not get rebuilt, two things happen at once. First, the early‑career cohort loses the human scaffolding that helps them turn AI outputs into sound judgment. Second, managers become the pinch point for change, because the administrative load that AI could lift is not systematically targeted, and their calendars stay full of low‑value work (Deloitte, 2025).
Frontier firms are rewriting the map of who owns work
Leaders are not shy about their intent. In a survey of 31,000 knowledge workers across 31 countries, 82% of leaders said they will use digital work agents to expand workforce capacity in the next 12 to 18 months. Almost half already use agents to fully automate workflows or processes. In companies that have deployed AI agents across the organization, 71% of workers say their firm is thriving, compared with 39% in the global average. The same report argues for replacing the traditional org chart with a dynamic "Work Chart" that is oriented to outcomes rather than job boxes (Microsoft, 2025). These are survey findings, supported by Microsoft 365 telemetry, and they come from a company with a clear commercial stake in AI. They still point to a practical shift worth heeding. The operating unit is becoming the workflow. Ownership of outcomes, human and agent roles, and measurement need to be explicit at that level (Microsoft, 2025).
That way of working would also square the circle with the employment data. Where AI augments, not replaces, teams can write down who is accountable for judgment and escalation, then let agents handle repeatable steps. Where AI automates, leaders need a plan for how juniors build experience anyway, because the entry‑level task is gone. In other words, the design of the "Work Chart" is not just a productivity play. It is a talent strategy (Microsoft, 2025; Brynjolfsson, Chandar and Chen, 2025).
Skills, at the necessary scale
The scale of skill change is sobering. Employers expect 22% of jobs to be disrupted by 2030, with 170 million roles created and 92 million displaced, for a net gain of 78 million. Nearly 40% of the skills required at work are expected to change by 2030. Sixty‑three percent of employers name the skills gap as the top barrier to business transformation. In response, 77% plan to upskill their workforce, 41% plan to reduce headcount as AI automates tasks, and about half plan to reorient the business toward AI‑created opportunities. The Forum estimates that 59 out of 100 workers globally will need reskilling or upskilling. These are employer expectations, not observed outcomes, but they indicate the direction of travel and the scale of preparation required (WEF, 2025).
Taken together, the evidence suggests four moves that are both urgent and doable:
- Map tasks by automation versus augmentation, then focus AI on the administrative and coordination work that clogs managers’ schedules. Freeing managers’ time to develop people is the fastest way to increase capacity without sacrificing the pipeline (Deloitte, 2025).
- Build protected on‑ramps for early‑career talent in AI‑exposed roles. Pair juniors with outcome ownership on real workflows and give them explicit decision rights alongside agents. The goal is to replace lost entry‑level tasks with supervised practice, addressing the employment pattern identified in the payroll data (Brynjolfsson, Chandar and Chen, 2025).
- Replace static org charts with a living work map. Document the owner of each outcome, the agent’s role, the human escalation path, and the metric that defines success. This helps avoid invisible overload on managers and turns AI from a headcount tactic into an operating system change (Microsoft, 2025).
- Fund skills at the scale implied by the forecasts, with a bias for internal mobility. Treat the 63% skills‑gap signal as an operating constraint, not a talking point, and align upskilling with the workflows where agents are being introduced (WEF, 2025).
The practical implication is simple. If you flatten with AI but do not rebuild coaching and skills, you may get a short‑term cost win and a long‑term capability hole. If you design for augmentation where it creates judgment and for automation where it does not, then pair both with a Work Chart that makes ownership unmistakable, you get a stronger organization. The employment data suggest that who learns and who owns work will decide where AI helps and where it harms. The firms that act on that now will not just have fewer layers. They will have better ones.
Sources
- Brynjolfsson, Chandar and Chen (2025). Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of AI. Early-career workers in AI-exposed jobs saw a ~16% relative employment decline, with adjustments via headcount and sharper impacts where AI automates rather than augments. https://digitaleconomy.stanford.edu/publication/canaries-in-the-coal-mine-six-facts-about-the-recent-employment-effects-of-artificial-intelligence/
- Deloitte (2025). 2025 Global Human Capital Trends. Managers are overloaded with admin, spend little time developing people, middle-management postings are down 42% in the U.S., and Gartner projects 20% of organizations will use AI to flatten structures by 2026. https://www.deloitte.com/us/en/insights/topics/talent/human-capital-trends/2025.html
- Microsoft (2025). Work Trend Index 2025 — The Year the Frontier Firm Is Born. Leaders plan to use AI agents to expand capacity, nearly half already automate workflows with agents, frontier firms report stronger performance, and the report introduces the outcome-oriented Work Chart concept. https://www.microsoft.com/en-us/worklab/work-trend-index/2025-the-year-the-frontier-firm-is-born
- World Economic Forum (2025). The Future of Jobs Report 2025. Employers expect 22% of jobs to be disrupted by 2030, net job gains of 78 million, major skill shifts, widespread plans for upskilling, and significant headcount changes tied to AI. https://www.weforum.org/publications/the-future-of-jobs-report-2025/