Boards are buying copilots while quietly pulling rungs from the ladder. Leaders say they will use AI agents to expand capacity, yet early-career workers in AI-exposed roles are the first to feel the employment hit. Middle managers spend their days on reporting and firefighting, the exact work AI can absorb, and the market is calling their bluff. The choice is not whether AI will flatten the org. It is whether you will design for it or let it erode the work and the workforce.
What the labor market already shows
A large-scale analysis of payroll data finds a stark pattern. In occupations most exposed to AI, employment for early-career workers dropped by roughly 16 percent relative to peers, while experienced workers in those same occupations and workers in less-exposed fields held steady or grew (Brynjolfsson, Chandar and Chen, 2025). The adjustment is arriving through hiring and headcount rather than pay cuts, and the losses concentrate where AI automates tasks, not where it augments them (Brynjolfsson, Chandar and Chen, 2025). This is a working paper but it uses administrative data and stresses robustness to alternative explanations.
Employer expectations tell a parallel story of scale and strain. The World Economic Forum projects that 22 percent of jobs will be disrupted by 2030, with 170 million roles created and 92 million displaced for a net gain of 78 million, and it expects almost 40 percent of job skills to change in that window (World Economic Forum, 2025). These are surveys of employers, so they are expectations rather than observations, but the direction is clear. Seventy-seven percent of employers plan to upskill their workforce in response to AI, 41 percent plan to reduce headcount as tasks automate, roughly half plan to reorient the business toward AI-created opportunities, and 59 of every 100 workers globally will require reskilling or upskilling (World Economic Forum, 2025). The bottleneck is not interest in AI. It is the skills gap itself, which 63 percent of employers name as the top barrier to transformation (World Economic Forum, 2025).
Why the middle manager is in the blast radius
Deloitte’s global survey puts numbers on why the layer between executives and teams feels exposed. Managers report that nearly 40 percent of their time goes to administrative work and firefighting, only 13 percent to developing their people, and more than a third feel unprepared for the people side of the job (Deloitte, 2025). The market is moving accordingly. U.S. employers posted 42 percent fewer middle-management jobs at the end of 2024 than in spring 2022, and Deloitte cites a Gartner forecast that by 2026 one in five organizations will use AI to flatten their structures (Deloitte, 2025). Flattening is not a slogan here. It is a shift in what work managers do, away from status shepherding and toward clarity of outcomes and coaching.
Erik Brynjolfsson has long emphasized the difference between automation and augmentation. The employment data back him up: where AI substitutes for tasks managers have historically policed, headcount shrinks; where it complements human judgment, headcount holds up and value can grow (Brynjolfsson, Chandar and Chen, 2025). The implication is practical. Redesign roles so that AI takes the bureaucratic load and humans own judgment, coordination, and development.
The operating model that fits the tools
Microsoft’s 2025 Work Trend Index argues that the firms getting this right look different, and it supplies early indicators from a 31-country survey and Microsoft 365 telemetry. Eighty-two percent of leaders say they will use digital agents to expand workforce capacity in the next 12 to 18 months (Microsoft WorkLab, 2025). Forty-six percent report they already use agents to fully automate workflows or processes, and in “Frontier Firms” with organization-wide AI deployment, 71 percent of workers say the company is thriving compared with 39 percent globally (Microsoft WorkLab, 2025). Microsoft is a vendor with a stake in this shift, so treat the framing with appropriate caution, but the directional point is useful.
Two design ideas matter. First, swap the static org chart for a “Work Chart,” a dynamic map of outcomes, skills, and the agents and humans accountable for them (Microsoft WorkLab, 2025). Second, accept that new hybrid managerial roles are emerging. Someone has to specify the workflow, train the model, evaluate the output, and make the call. Think of an “agent boss” who orchestrates automated routines alongside people, not a supervisor who collects updates (Microsoft WorkLab, 2025). The more your managers cling to status by oversight, the more AI will hollow their jobs. The more they own problem framing, standards, and talent development, the more AI raises their ceiling.
How to keep trust and talent while you flatten
-
Start with a task-level audit anchored in augmentation. Use the critical distinction documented in the employment data: map tasks where AI is a substitute versus a complement, and redesign jobs so that substitution frees time for higher-value human work while complementing tasks gain explicit human ownership (Brynjolfsson, Chandar and Chen, 2025).
-
Rewrite the manager’s scorecard. If managers spend 40 percent of their week on admin, strip it and let agents handle reporting and status capture (Deloitte, 2025). Measure managers on outcomes delivered, decision quality, and time spent developing people, not on attendance at meetings. The point is not fewer managers, it is different management.
-
Build a Work Chart, not just an org chart. Make outputs visible, assign owners for human decisions, and catalogue which agents do which steps so work is observable without micromanagement (Microsoft WorkLab, 2025). This reduces the need for hierarchy as a control system and raises the need for clarity as an operating system.
-
Invest in skills at scale with a bias to the front door. Employers say they plan to upskill, and 59 of 100 workers will need it, but early-career workers are the ones losing ground first (World Economic Forum, 2025; Brynjolfsson, Chandar and Chen, 2025). Create protected on-ramps: rotational programs where juniors learn to supervise agents, structured apprenticeships tied to clear skill signatures, and mentoring that targets the judgment work AI cannot yet do. Without this, AI will widen the experience gap your future leaders need to cross.
-
Align workforce strategy with business model change. About half of employers expect to reorient for AI-driven opportunities, but most cite the skills gap as the top barrier (World Economic Forum, 2025). Tie reskilling to specific workflows and products, not generic catalogs. If you are automating case intake, teach claims staff to design prompts, diagnose model errors, and set escalation rules. If you are automating forecasting, train product owners to interpret uncertainty and make trade-offs.
The near term will be messy. Surveys are not destiny, and one working paper is not the final word. Yet the signals rhyme. Employers expect disruption on a vast scale, managers are doing work software can do, and early-career workers are the canaries. Treat flattening as a design choice. Give agents the paperwork. Give people the problems. Then measure leaders by how well they grow judgment and trust in a system that finally has fewer rungs and clearer work.
Sources
- Brynjolfsson, Chandar and Chen (2025). Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of AI. Early-career employment fell about 16 percent in AI-exposed jobs, with impacts concentrated where AI automates tasks rather than augments them. https://digitaleconomy.stanford.edu/publication/canaries-in-the-coal-mine-six-facts-about-the-recent-employment-effects-of-artificial-intelligence/
- World Economic Forum (2025). The Future of Jobs Report 2025. Employer expectations indicate 22 percent job disruption by 2030, a net gain of 78 million jobs, large skill shifts, widespread plans for upskilling, some headcount reduction, and 59 of 100 workers needing reskilling or upskilling. https://www.weforum.org/publications/the-future-of-jobs-report-2025/
- Deloitte (2025). 2025 Global Human Capital Trends. Managers devote about 40 percent to admin and 13 percent to development, feel underprepared, middle-management postings fell 42 percent, and Gartner forecasts 20 percent of organizations using AI to flatten structures by 2026. https://www.deloitte.com/us/en/insights/topics/talent/human-capital-trends/2025.html
- Microsoft WorkLab (2025). Work Trend Index 2025 — The Year the Frontier Firm Is Born. Leaders plan to use AI agents to expand capacity, many already automate workflows, Frontier Firms report higher thriving, and the report introduces the Work Chart and agent-boss concepts. https://www.microsoft.com/en-us/worklab/work-trend-index/2025-the-year-the-frontier-firm-is-born