The AI-reshaped organisation.

All parts and the executive summary

The technology function is in the middle of a structural rewrite, not a tooling refresh. It did not start changing because of one product launch. It is changing because a stack of compounding forces hit at the same time, and tools shift quarterly while forces shape the decade. This abridged reading covers the headline findings, the forces behind them, and the three scenarios they point to. The full note carries the role-by-role detail, the delivery model, the people decisions, and the companion playbook.

1. Six findings, one rewrite

The note opens with six findings that hold across surveys, filings and academic work.

Middle management is the first layer to thin out. Gartner projects that by the end of 2026, 20% of organisations will use AI to eliminate more than half of current middle-management positions. Amazon trimmed roughly 14,000 corporate roles in late 2025 explicitly to remove layers, and AI-attributed US layoffs in 2025 reached about 55,000. The pace hardened into 2026: by May, AI was the single most-cited reason for US job cuts for a third month running, and the year-to-date total had already passed the whole of 2025.

Senior individual contributors gain power; juniors lose entry routes. Entry-level hiring at the 15 largest tech firms fell roughly 25% between 2023 and 2024, developer employment for ages 22 to 25 fell around 20% from its late-2022 peak, and the staff-level AI premium widened to 18.7%.

Delivery is moving from writing code to reviewing it. The 2025 DORA report found 90% of developers using AI at work. The catch is that AI amplifies what is already there: strong teams accelerate, weak teams see instability rise. Spec-driven development and multi-agent review are emerging as the AI-native delivery pattern. Tiny teams are setting new revenue-per-employee records. And most enterprises remain stuck in the pilot loop, where more than 80% see no tangible EBIT impact, because the constraint is workflow redesign, not access to technology.

2. The four forces compressing the tech team

Underneath the findings sit four forces that reinforce one another and do not unwind.

The first is coordination compression. Recent academic work frames AI as coordination-compressing capital: each unit of agent capital reduces the number of human managers needed to supervise the same work. Spans of control widen. Microsoft is reportedly pushing toward a 10:1 engineer-to-manager ratio, roughly double the typical Big Tech ratio.

The second is economic compulsion. Salesforce has taken customer support from 9,000 people to about 5,000 since launching Agentforce. These are operating-cost decisions taken by listed companies with auditors and shareholders, not pilots. Once one competitor removes 20% of overhead with credible quality, the rest follow.

The third is the throughput multiplier, and it is more honest than the headlines. AI gives a real lift when a team has strong foundations and a smaller one, or even a drag, when it does not.

The fourth is talent reallocation. Demand is hardening at the top of the seniority distribution and softening at the bottom, pulling frontier compensation clear while entry routes narrow.

3. Three scenarios for 2026 to 2029

The note pairs an evidence-first base case with bold scenarios, and assigns no probabilities on purpose.

In Scenario A, conservative, the existing design absorbs the tools. Spans widen modestly to about 1:9, middle management shrinks 10 to 20% at the most exposed firms, and delivery looks like a faster, more reviewed version of today. The risk is that it underweights the cost case.

In Scenario B, moderate, which the most data points to, the pure-coordination management layer is thinned 30 to 50% by the end of 2028, spans widen to 1:10 or 1:15, and the work chart becomes the operating document while the org chart persists for HR and legal purposes.

In Scenario C, radical, the engineering-manager layer effectively dissolves at most net-new organisations. Companies are designed from day one as work charts, spans reach 1:20 to 1:50, and the leading edge crosses USD 10 million of revenue per employee. Scenario C arrives only if quality, governance and trust hold up.

---

End of the abridged reading. The full PDF continues with the role-by-role detail (Part 2), the AI-native delivery model (Part 3), hiring, levelling, performance and pay (Part 4), whole-company structure and the work chart (Part 5), startups versus enterprises (Part 6), the implications and the companion playbook for leaders (Part 8), the risks and disruption signals to watch (Part 9), and the full data appendix and citation set. It is sent by email to readers who request it using the form below.