Artificial Idea | AI careers · practical prompts · no hype Monday, September 29, 2025 · Issue #17 · Jobs

The invisible restructure

The last wave of automation went after the factory floor. This one is going after the floor above it.

In August 2025, Shopify's internal memo to all staff leaked onto Reddit and spent three days at the top of every major technology publication. The memo, authored by CEO Tobi Lütke, instructed managers across the organisation to justify every headcount request by first demonstrating that the work could not be done by AI. It went further. It told teams to expect that AI would be embedded into every workflow, every reporting structure, and every coordination function within twelve months, and that roles whose primary value was information coordination rather than judgment would be evaluated accordingly.

The memo was reported as a bold, controversial move by a visionary CEO. It was neither bold nor controversial to anyone paying close attention to what has been happening inside large organisations for the past eighteen months. Shopify said out loud what dozens of organisations have been doing quietly. The restructuring of middle management around AI capability is not a future event. It is a current one, and it is moving faster than the professionals most affected by it currently understand.

What middle management actually does

To understand why middle management is the current target, it helps to be precise about what middle management actually does, as opposed to what it is supposed to do.

In theory, middle managers translate organisational strategy into team execution, develop the professionals who report to them, make judgment calls on the operational decisions that do not require senior leadership attention, and maintain the human relationships and cultural coherence that hold teams together under pressure. These functions are genuinely valuable and genuinely difficult to automate.

In practice, a significant portion of what many middle managers spend their time on looks different. A 2024 analysis by Bain and Company of time allocation across 1,800 middle managers at large enterprises found that 54% of their working week was spent on activities that fall into three categories: gathering and synthesising information from direct reports, producing status updates and reports for senior leadership, and coordinating scheduling and resource allocation across teams.

All three of those categories are being automated, rapidly and at scale, by AI tools embedded directly into project management, communication, and business intelligence platforms. When the platform can automatically synthesise what happened last week, surface which projects are behind schedule and why, draft the status report for senior leadership, and flag resource conflicts before they become problems, the human layer that existed primarily to perform those functions becomes structurally redundant.

This is not a prediction. Microsoft's 2025 Work Trend Index, which analysed usage patterns across 31 million Microsoft 365 users, found that AI-assisted synthesis and reporting features reduced the time managers spent on information coordination tasks by an average of 38% within six months of adoption. That time did not automatically flow into the higher-value management activities that justify the role. In many cases it revealed that the higher-value activities represented a smaller portion of the role than either the manager or their organisation had previously acknowledged.

The two types of middle managers

The research on AI adoption in management contexts consistently produces a finding that is more useful than the aggregate statistics: the impact of AI on middle management roles is not uniform. It is bifurcating in a pattern that closely mirrors the broader labour market bifurcation described in Issue #1.

The middle managers whose roles are most at risk share a specific profile. Their value to the organisation is primarily informational and coordinative. They are skilled at gathering, synthesising, and communicating information about what their team is doing. They manage processes competently. They are reliable. They are also, when AI handles the information and coordination functions, difficult to justify at their cost relative to the organisational value they produce.

The middle managers whose roles are becoming more valuable share a different profile. Their value is primarily developmental and judgmental. They are skilled at identifying capability in the people they manage and building it deliberately. They make good decisions on problems that do not have obvious answers. They maintain relationships and trust under conditions of organisational stress and ambiguity. They understand the business deeply enough to translate strategic intent into team action without losing the meaning in the translation.

These two profiles have always existed within the same job title. The difference is that AI has made the distinction between them visible in a way it was not previously, by removing the coordinative and informational work that previously occupied enough of every manager's week that the underlying capability question was obscured.

A 2025 Gartner survey of 400 senior HR leaders found that 61% planned to reduce middle management headcount by at least 15% over the following two years, and that the primary selection criterion for which roles to eliminate was the proportion of the role that consisted of information coordination versus judgment and development. That criterion, applied systematically, produces a restructuring that looks random from the outside and is entirely legible from the inside.

The India context

For readers in India, where middle management has historically represented one of the most reliable pathways into upper-middle-class professional stability, the dynamics described above have specific and significant implications.

India's large technology services and business process outsourcing sectors employ a substantial proportion of the country's professional middle management workforce in roles that are disproportionately weighted toward the information coordination and process management functions most vulnerable to AI displacement. The scale of employment in these sectors means that even a modest restructuring of management layers produces a large absolute number of affected professionals.

A 2025 report by the National Association of Software and Service Companies found that Indian IT services companies collectively reduced middle management headcount by an estimated 8% in the first half of 2025, with the reductions concentrated in project coordination, delivery management, and client reporting functions. The report projected a further 12 to 15% reduction over the following eighteen months as AI-assisted project management and client reporting tools reach full deployment across major organisations.

These figures are not cause for panic. They are cause for the kind of honest self-assessment that this newsletter has been advocating since Issue #1. The question for every middle manager reading this is not whether their organisation will embed AI into management workflows. It is whether the value they produce is primarily in the functions AI is replacing or primarily in the functions it is not.

What the transition looks like in practice

The restructuring of middle management around AI is not, in most organisations, happening through large announced layoffs. It is happening through a quieter process that has three typical stages.

The first stage is tool adoption without role redesign. AI tools are embedded into workflows, reducing the time required for coordination and reporting tasks. Managers are expected to use the tools and benefit from the efficiency gains. The role description does not change. The implicit expectation about what managers should be doing with the recovered time does.

The second stage is performance evaluation against the new expectation. Managers who used the recovered time to go deeper on the judgment and development functions of their role become visibly more effective. Those who used it to do the same coordinative work more efficiently, without expanding into higher-value activity, become visibly less differentiated from the tool that now handles most of that work.

The third stage is restructuring, framed in the language of organisational efficiency rather than AI displacement. Roles are consolidated. Spans of control are widened. Layers are removed. The managers whose roles survive are those who had already made the transition to higher-value activity. The ones whose roles are eliminated are those who had not.

The entire process typically takes eighteen to thirty-six months from initial tool adoption to structural change. At many large organisations, that clock started in 2023 or 2024. The restructuring decisions it produces are being made now.

The action

If you are in a middle management role, conduct the task classification exercise from Issue #15 with specific reference to your management responsibilities. Which parts of your role consist of gathering, synthesising, and communicating information? Which parts consist of judgment, development, and relationship management?

The first category is where your role is being compressed. The second is where it is becoming more valuable. The professionals who explicitly and deliberately shift their time allocation toward the second category, before their organisation makes that shift mandatory, are the ones whose roles survive and grow through the restructuring. The ones who wait for clarity that comes too late are the ones for whom the restructuring is a surprise.

It should not be a surprise. The pattern is visible, the timeline is measurable, and the action required is clear. What it requires is honesty about which category your current work primarily falls into, and the willingness to change that allocation before someone else changes it for you.

Thursday we are giving you the research prompt stack that consultants and senior analysts are using to compress four hours of competitive and market research into twenty minutes, without sacrificing the analytical depth that makes research worth doing in the first place. It is one of the highest-leverage prompt applications we have covered and one of the most immediately applicable regardless of your industry or function.

The time it saves is real. What you do with that time is the question this newsletter keeps returning to.

— The Artificial Idea team