URL: theguardian.com Date: 2026-05-15 Class: foundational (long-form feature anchored on multiple academic and industry voices advancing an "asynchronous, agent-driven management" thesis)
Summary
A May 15, 2026 Guardian technology feature documenting an AI-driven reduction of middle management across several large technology firms, drawing on internal org charts, public executive statements, labor-market data, and interviews with four academics, a Gartner analyst, a consultant, and a departing Meta engineering manager. The piece reports layoffs at Coinbase and Block, expanded manager spans of control, and the emergence of AI-mediated management practices, and frames these as changes to how firms coordinate work rather than as cost-cutting alone.
Across the cases the feature describes three recurring patterns: managers act as both supervisors and producers (writing code and contributing to deliverables); spans of control widen from a prior range of roughly 6–12 reports to anywhere between 15 and 175 reports; and asynchronous, agent-mediated management appears as a fallback where spans become too large for direct contact.
Company actions and labor-market data
The feature reports specific actions at four firms and two broader labor-market figures.
Coinbase laid off 14% of its workforce, ended what it called "pure managers," and required remaining managers to contribute code and to hold at least 15 direct reports, per a May 2026 tweet by chief executive Brian Armstrong. Armstrong described the goal as "rebuilding Coinbase as an intelligence, with humans around the edge aligning it."
Block laid off 40% of its workforce, with engineering managers holding up to 175 direct reports under a new AI-oriented structure, according to internal org charts reviewed by the Guardian. The feature reports a stated longer-term goal that all 6,000 employees eventually report directly to Jack Dorsey once management layers are removed, citing Dorsey's board reporting (via a carnage4life Bluesky thread). Block's framework holds that "there is no need for a permanent middle management layer," attributed to Dorsey and board member Roelof Botha.
Meta's January 2026 earnings call featured Mark Zuckerberg discussing a flattening of the management structure, following the company's January 2023 "year of efficiency" precedent, per the Meta Q4 2025 earnings transcript.
Amazon increased its ratio of individual contributors to managers by at least 15%, a goal Andy Jassy announced in 2024 and that the feature reports as achieved in 2025.
On the broader labor market, the feature reports that U.S. middle-manager job openings were down 42% versus their 2022 peak as of end-of-2025 data (Revelio Labs, cited via Business Insider), and that managers made up 13% of the U.S. workforce in 2022, roughly 21 million people (Harvard Business School).
| Company | Action | Source |
|---|---|---|
| Coinbase | Laid off 14% of workforce; ended "pure managers"; remaining managers must contribute code and have ≥15 direct reports | Armstrong tweet (May 2026) |
| Block | Laid off 40% of workforce; engineering managers up to 175 direct reports under new AI-oriented structure | Internal org charts reviewed by Guardian |
| Block goal | Eventually, all 6,000 employees report directly to Dorsey (minus management layers) | Dorsey board reporting (carnage4life Bluesky thread cited) |
| Meta | January 2026 earnings call: Zuckerberg discusses flattening management structure; January 2023 "year of efficiency" precedent | Meta Q4 2025 earnings transcript |
| Amazon | Increased ratio of ICs to managers by at least 15% (Jassy 2024 goal — achieved 2025) | Jassy 2024 announcement |
| U.S. middle-manager openings | −42% vs. 2022 peak (end of 2025 data) | Revelio Labs |
| Managers as % of U.S. workforce | 13% in 2022 (~21M people) | Harvard Business School |
New role types
The feature describes role-redefinition frameworks at Block and Coinbase. At Block, AI is described as primarily responsible for sharing information between managers, reports, and teams; "directly responsible individuals" (DRIs) oversee strategy and priorities; and "player-coaches" are individual contributors who also manage employee growth. At Coinbase, the parallel framework removes "pure managers" in favor of managers who directly contribute code or other work, with spans of 15 or more direct reports.
Asynchronous agent-driven management
The feature's account of AI-mediated management comes from Prateek Singh, a Meta software engineering manager who left in April 2026. Singh said he switched 1:1 meetings with his seven reports from weekly to every other week and communicated asynchronously via AI agents in between, with his agent connecting to his reports' agents. He described the result as "more asynchronous, agent-driven management — then people lose touch with all the benefits you get from face time."
Singh also raised a concern that managers under span pressure are tempted to "use AI for decisions and blindly submit flawed suggestions… could lead to data leaks, security holes, or even system outages."
Academic and analyst voices
The feature quotes several academics and analysts:
- Anastassia Fedyk (Haas/UC Berkeley): AI tools shift work from managers to reports, and the structural changes may be permanent.
- Raffaella Sadun (Harvard): tech companies are positioned to make these changes, but the cost of change is still real, including a work-coordination overhaul and role demotions.
- Emily Rose McRae (Gartner): the manager job is already a "drag," many managers would not take the role again given the choice, and reductions will worsen the squeeze on remaining managers.
- Matthew Bidwell (Wharton): skeptical the experiments will catch on, citing a "history of companies that have tried to break old hierarchies with new forms of management… often abandoned or serve as one-offs," and noting that removing managers loses a "layer of kicking the tires," moving faster but breaking more things.
- Amalia Goodwin (Slalom): removing managers requires redesigning how work gets done, giving more authority to lower levels and providing training and resources to make judgment calls.
Key claims
| Claim | Confidence | Notes |
|---|---|---|
| Block managers up to 175 direct reports | medium-high | Internal org charts (reviewed by Guardian, not public); plausible given Block's stated structure. |
| Middle-manager openings −42% since 2022 peak | high | Revelio Labs, cited via Business Insider. |
| Coinbase / Block / Amazon / Meta all explicitly flattening for AI | high | Public statements; multiple direct quotes. |
| AI agents now used for asynchronous management feedback | medium | Singh testimony; not yet measured in cross-company surveys. |
| Pattern will likely fail / partially reverse | low-medium | Bidwell's attributed position; not a measured outcome. |
Reception and contested points
The feature presents the management changes as contested rather than settled. Bidwell argues the experiments are unlikely to persist and that removing a "layer of kicking the tires" trades speed for more frequent failures; this is an attributed editorial position, not a measured outcome. Singh offers a qualified positive on agent-mediated management while also warning that it degrades mentorship and human judgment, and that span pressure can push managers toward unsafe reliance on AI. Sadun and Goodwin both stress that the change carries real costs and dependencies (a work-coordination overhaul, role demotions, and the redesign of how work gets done), while Fedyk suggests the shifts may be permanent.
The feature situates these dynamics relative to a broader "AI replaces workers" account of labor disruption (see Prepare for an AI jobs apocalypse (The Economist Leader, May 14 2026)). The Economist piece centers on automation of tasks producing layoffs across roles, whereas the Guardian feature centers on compression of coordination overhead, visible primarily through widening spans of control; Bidwell's view that the management changes may be self-correcting contrasts with the more general expectation that task automation is at most partially reversible. The two accounts are presented as complementary rather than mutually exclusive.
Relationships
- supports: Intelligence Replaces Hierarchy, AI Labor Disruption, AI and Productivity
- contradicts: "AI augments managers, doesn't replace them" framing; the documented spans-to-175 are incompatible with augmentation
- depends-on: Agentic AI, Agent Architecture Patterns (agent-to-agent communication required for the Singh scenario)
- related: Meta AI, Amazon, Coinbase (created), Block (Square / Cash App) (created), Jack Dorsey (created), Brian Armstrong (created), Mark Zuckerberg, Andy Jassy, Anastassia Fedyk (created), Raffaella Sadun (created), Emily Rose Mcrae (created), Matthew Bidwell (created), 'A' Grades Are Suddenly Everywhere Since the Arrival of ChatGPT (WSJ, May 13 2026) — Chirikov UC Berkeley study ("AI erodes signaling" parallel)
- regulated-by: Labor law (NLRB on management-IC distinction); employment-discrimination law (Mobley v. Workday adjacency)
Wiki Folding
Updates Intelligence Replaces Hierarchy (canonical reference + Block 1:175 spans + Coinbase / Meta cases), AI Labor Disruption (manager-layer-as-distinct-from-IC section), Labor Disruption Timelines: Who Predicts What and Why, creates entity pages for academic interviewees + Dorsey/Armstrong (if not already present), and creates Block (Square / Cash App) + Coinbase company pages.