AI and Middle Managers: Who Checks the Work When the Middle Thins?

The person in the middle

Somewhere in most organisations this week, a manager will spend an evening reading through reports, summaries and analyses that an AI tool drafted in seconds. Each one looks plausible. Each one needs checking before it goes anywhere. And the same manager may already suspect that their layer of the organisation is the one most likely to shrink.

New research from The Conference Board gives that experience some numbers. Its report, Leadership Disrupted: Preparing for an AI-Enabled Workplace, draws on a survey of 775 leaders internationally and 20 interviews. The report page is dated 12 October 2026 and the findings were released to the press on 20 October.

What the research found

AI is already part of everyday work for most respondents. 64% say it is used across several functions or is highly integrated into daily work.

Accountability has not kept pace. 37% say ownership of AI-related leadership decisions is inconsistent or unclear. The gap closes as organisations mature: 85% of organisations where AI is highly integrated report clear ownership or formal governance, compared with 26% of those still exploring AI.

The expected structural change falls unevenly. Just 4% of leaders expect a reduction in C-suite roles, while 38% anticipate fewer middle managers and 36% expect reductions in both frontline managers and project leaders.

The report also describes what AI speed is doing to leaders themselves: new managerial bottlenecks, fatigue and the risk of what it calls "work slop", as leaders evaluate an influx of AI-generated work without sacrificing quality, sound judgement or employee development. It calls for deliberate human checkpoints at critical stages, so leaders can review inputs, validate outputs, correct errors and confirm that decisions remain fit for purpose.

On skills, 52% rank critical thinking and decision-making as the most important enduring leadership skill. Continuous AI literacy and knowing when human judgement should override AI tie at 40% as the top AI-expanded competencies, and 33% identify redesigning work for human and AI collaboration. 72% of leaders favour hands-on practice with AI tools over more traditional development approaches.

A separate study released on 21 October points the same way. Research by Nielsen, commissioned by monday.com (vendor research, so read it with that in mind), surveyed 1,701 directors and senior managers in seven countries, including the UK, Germany and France. 53% say it is now harder to evaluate employees' original contributions, and 35% worry about being held accountable for decisions made by an AI agent.

What it means for leaders

Put those findings side by side and a tension appears. The managers being asked to act as the quality check on AI output are the group many organisations expect to need fewer of. If the review load rises while the number of reviewers falls, something gives. Usually it is the checking, the coaching, or the manager.

Ian Hughes reaches a similar reading in What Stays the Same. He expects the middle of the organisation to thin as machines take on more of the doing, but he is clear about what remains: "You still need the judgement above them. Arguably you need more of it."

John Sansbury's The Journey to Service Excellence offers a useful test for what that judgement is for. He argues that "the primary role of the manager or leader is to use their authority to remove the obstacles to their teams’ high performance." Applying his principle to AI, the question for a leadership team is not simply how many managers it needs. It is whether AI is freeing managers to remove obstacles, or burying them in a new kind of checking.

Where the AI Enablement Toolkit for Leaders helps

The toolkit is a leadership method rather than software, and several of its tools speak directly to this problem.

- Tool 10, Human and AI role design, makes delegation explicit. Draft mode, for example, means the AI creates a proposed output and a human must review, edit and approve it before use. Naming the mode tells you where review time is genuinely needed and where it is not. - Tool 11, the ownership and boundaries canvas, records the named human owner, approval points, quality checks and escalation route for an AI-enabled workflow. That answers the 37% who say ownership is unclear. - Tool 14, Protect the apprenticeship, helps leaders keep the experiences that build judgement in less experienced people, so that removing grind does not remove the route by which people learn what good looks like. - Tool 24, the ten-minute leader check-in, gives a light rhythm for asking where a workflow helped, stalled or created a new exception, which is where review bottlenecks first show up.

Training matters here too, and the research is specific about the kind. Leaders want hands-on, role-specific practice rather than general AI fluency courses. Esm's training and enablement is built around leadership teams working on their own real workflows together, practising how to set review points, judge AI output and redesign roles before any structural decisions are made.

Honest limits

The toolkit is a management control, not a technical fix. It will not tell you how many managers your organisation needs, and it does not replace HR, legal or employee relations advice on any restructuring. The Conference Board figures are self-reported by leaders, not measured outcomes, and the monday.com study is commissioned vendor research. Treat both as a prompt for your own evidence, gathered from the people doing the work.

A first step

Before any decision about management layers, ask your managers one question: how much of your week now goes on checking AI output, and what would you do with that time if the review work were properly designed? The answer is a baseline worth having.

Tags: AI leadership, middle management, AI governance, work redesign, AI Enablement Toolkit

Sources

- https://www.prnewswire.com/news-releases/report-ai-is-changing-leadership-faster-than-organizations-are-adapting-302914361.html

- https://www.conference-board.org/publications/leadership-disrupted-preparing-for-an-AI-enabled-workplace

- https://www.businesswire.com/news/home/20261021280713/en/83-of-Leaders-Are-Satisfied-With-Their-AI-Setup-but-7-in-10-Say-Their-Company-Hasnt-Fully-Integrated-It-monday.com-Research-Reveals