AI Risk and Governance

AI Is Not Coming for the Workers First. It Is Coming for the Layer Above Them.

The first roles AI thins out are not front-line workers but the coordination layer above them. What 2025-2026 data shows, and what leaders should do.

By Leon Soliman · 2026-06-18 · 2 min read

Key takeaways

Gartner predicted that through 2026, 20 percent of organizations will use AI to flatten their structure and eliminate more than half of current middle management positions.
AI mostly replaces coordination work (scheduling, status reports, performance dashboards, approval routing), not the builders who write the code, close the sale, or run the machine.
In a September 2024 memo, Amazon CEO Andy Jassy set a target to raise the ratio of individual contributors to managers by 15 percent by Q1 2025, explicitly to remove management layers.
Reported AI-linked workforce reductions ran past 150,000 roles in 2026, and across 2025 and 2026 firms including Intel and Google cut management headcount specifically, not only front-line staff.
The middle layer also held institutional knowledge and judgment under ambiguity; flatten it without naming where that judgment now sits and you move an unowned risk up, not away.

Which jobs does AI actually take first?

It takes the coordination layer first, not the front line. The roles thinning out fastest are not the people writing the code, closing the sale, or running the machine, but the managers whose day was assigning, monitoring, summarizing, and reporting on that work. Gartner predicted that through 2026, 20 percent of organizations will use AI to flatten their structure and eliminate more than half of current middle management positions. The reason is plain: scheduling, status reports, performance dashboards, and approval routing are exactly what a model does well, and they were most of what a coordination layer did all day.

Is this already happening at real companies?

Yes, and it started at the top of the market before the AI wave was obvious. In a September 2024 memo, Amazon CEO Andy Jassy set a target to increase the ratio of individual contributors to managers by 15 percent by the first quarter of 2025, explicitly to strip out management layers. Reported AI-linked workforce reductions ran past 150,000 roles in 2026. Across 2025 and 2026, companies including Intel and Google cut management headcount specifically: Intel set out to eliminate management layers, and Google reduced its manager, director, and VP roles. The pattern is consistent: the builders stay, the people who used to route information between them do not.

Why should an owner or family office care about this distinction?

Because the coordination layer was not only overhead; it was also where institutional knowledge, judgment under ambiguity, and quiet relationship management lived. A dashboard can reproduce the reporting. It cannot reproduce the manager who knew which client was about to leave and why. If you flatten the structure without naming where that judgment now sits, you have not saved cost. You have moved an unowned risk one level up, into your own calendar.

What should a leader do about it?

Decide deliberately which parts of the middle layer were overhead and which were judgment, before a vendor or a quarterly cost target decides it for you. Keep and elevate the judgment, automate the routing. Most organizations are doing the reverse by accident: they cut on cost and discover the judgment is gone only when something breaks. Treat the org chart as an attack surface of its own, and decide where accountable judgment lives once the routing is automated. Servola advises on AI risk and governance for owners weighing exactly this.

Frequently asked questions

Does AI replace workers or managers first?

Current 2025-2026 data points to the coordination layer first. The roles being automated fastest are managerial routing tasks such as scheduling, reporting, and performance monitoring, while the people doing the core work tend to remain.

What did Gartner predict about middle management and AI?

Gartner predicted that through 2026, 20 percent of organizations will use AI to flatten their organizational structure, eliminating more than half of current middle management positions.

If middle management is overhead, why not just cut it?

Because part of that layer is overhead and part is judgment, institutional knowledge, and relationship management. Cutting on cost alone tends to remove the judgment by accident, which is why the safer move is to separate the two before reorganizing rather than after.

The layer most exposed to AI is also the layer that quietly held the business together, and the firms that come out ahead will be the ones that knew the difference before the cuts, not after.

AI Future of Work Management AI Risk & Governance

More from the Servola Journal

AI Risk and Governance

If Apple Decided It Could Not Build Its Own AI, Why Do You Think You Can?

2026-06-21 · 3 min read

Read the article →

AI Risk and Governance

Your AI Vendor Just Became Your Competitor. What Now?

2026-06-21 · 3 min read

Read the article →

AI Risk and Governance

Can 13 Words Poison What AI Says About Your Company? Reportedly, Yes.

2026-06-20 · 2 min read

Read the article →

Servola

If you are weighing where AI thins your org chart and where accountable judgment must stay, that is exactly what we advise on, quietly, with one accountable owner.

Request a private introduction About Servola →

Servola is technology counsel for a small number of families and offices. When a decision cannot be delegated, we sit on your side of the table.

Servola Systems GmbH · Ludwigshafen, Germany · [email protected]

← All articles