AI is becoming increasingly easy to deploy. Redesigning the business around it is still much harder.

BMW’s latest strategy announcement in September 2026 contains a number that deserves to be taken seriously: the company plans to reduce the total number of divisions and associated management roles by 20% by mid-2027, as part of a broader programme aimed at making the business more agile, efficient and profitable. BMW is equally explicit about the role of technology in that effort, describing the systematic use of AI across development, purchasing, production and aftersales as one of the measures that will help create leaner organizational structures. Reuters has reported that the broader restructuring is expected to involve around 8,000 jobs in Germany.¹²³

For the people whose roles are affected, this is obviously more than an operating-model case study. It means careers changing, teams being merged or removed, responsibilities being redistributed and, in some cases, people losing their jobs. That human impact should not be obscured by the excitement around AI. At the same time, the development is important precisely because it shows that AI is moving beyond the realm of employee productivity tools and into decisions about how large organizations are actually structured.

BMW is not saying that AI will simply make individual employees faster. It’s using AI as one of the mechanisms through which it believes the organization can operate with fewer layers and less administrative work. That is a much bigger change.

The real shift begins when AI changes the work itself

BMW has already provided a useful example of what this looks like in practice.

In May, the company described several agentic-AI applications being introduced into everyday operations. In its fleet business, an internal AI agent takes unstructured requests from corporate customers, transfers the relevant information into internal systems and initiates the required process; BMW says this replaces around 90% of previously manual tasks in that workflow. In another case, involving the management of roughly 250,000 specialized tools worldwide, AI can draft orders, send them to suppliers, review responses and approve straightforward cases, while employees remain responsible for checking the results and intervening where specialist judgment is needed.⁴

That distinction matters. The interesting part is not that BMW has deployed an AI model. It’s that the company has changed the sequence of work around the technology. Tasks that once required people to collect information, transfer it between systems, perform routine checks and coordinate the next step can now be handled largely by software, leaving people with exceptions, judgment and customer interaction.

BMW describes that model itself as “Prepared by AI, decided by people.”⁴

That is a useful way to think about the organizational transition now under way in many companies. When technology starts handling more of the preparation and coordination, some roles inevitably change. Some disappear. Others become more valuable because they are concentrated around the decisions that machines are still not good at making.

The difficult part is not deploying AI. It’s deciding what should change around it.

This is where many companies are likely to struggle.

McKinsey’s 2026 research argues that most large organizations are still using AI to accelerate existing activities while leaving the underlying operating model — governance, teams, capabilities and decision processes — largely intact. In its research, only 21% of companies had fundamentally redesigned their operating models around AI. Among the companies generating at least 5% of EBIT from AI, broad operating-model redesign was three times as common.⁵

A separate McKinsey analysis published in September makes the same point more directly: AI, particularly agentic AI, changes how decisions are made, how work moves between functions, how capabilities are developed and where value is created. Capturing that value therefore requires intentional choices about the operating model rather than simply adopting the newest technology.⁶

Deloitte’s 2026 enterprise research shows a similar divide in the Middle East. 66% of organizations surveyed report improved efficiency from AI, but only 34% say they are using AI to fundamentally redesign products, processes or business models. Deloitte also reports that 84% of regional organizations have yet to redesign roles or workflows around AI capabilities.⁷

That is the gap between AI adoption and AI transformation.

The first can be done by giving people better tools. The second eventually forces management to confront a more uncomfortable question: if the technology changes the economics of the work, why should the organization remain structured around the old economics?

For employees, the answer is not simply “fewer people”

That is where the conversation around AI can become too simplistic.

There will clearly be roles that become smaller or disappear. BMW’s 20% reduction in divisions and associated management roles is a concrete example, and it would be misleading to pretend that organizational redesign somehow eliminates the disruption experienced by the people affected.

But headcount is only one dimension of what changes.

A manager whose time is largely consumed by consolidating reports, checking routine approvals or coordinating information between departments may find that much of that work disappears. That is a genuine productivity gain, but it does not necessarily mean that management itself has become less important. The value of the role may shift toward setting priorities, managing exceptions, developing people, resolving trade-offs and making decisions where context matters.

The same is true elsewhere in an organization. AI can remove work without removing the underlying business need that the work was serving.

The question becomes what happens to the capacity that has been released.

That is where the economic logic becomes more interesting

A company can use AI in at least two fundamentally different ways.

The first is to reduce the amount of labor required to perform today’s processes.

The second is to change the processes themselves so that the company can operate differently.

The first can produce a relatively quick saving. The second can create a structural advantage.

Neither is inherently better, and there is commercial sense with taking cost out of a process where the cost no longer creates value. But there is a danger in making the saving itself the strategy. If experienced employees leave before the organization has captured their knowledge, if customer relationships become weaker because interactions are over-automated, or if the company removes the people who used to handle difficult exceptions without redesigning how those exceptions are managed, the short-term improvement can create a longer-term problem.

The more interesting question is what the company does with the capacity it frees up.

Does it serve customers better? Does it develop products faster? Does it enter markets it previously could not afford to enter? Does it strengthen commercial execution? Does it put more expertise behind the decisions that matter?

Those are the outcomes that determine whether AI has actually improved the business rather than simply reduced its cost base.

BMW’s own examples point in that direction. The company says its fleet AI application frees capacity for direct customer engagement and support, while its Purchasing application automates routine work but keeps humans responsible for review and decision-making.⁴

The manager is not really the unit of analysis

The attention on management reductions is understandable because management headcount is visible and measurable. But the deeper unit of change is the workflow.

Take a process that currently requires several people because information sits in different systems, approvals are sequential and every exception gets handled manually. Introduce AI that can collect the information, prepare the recommendation, execute routine steps and escalate only the unusual cases, and the organizational requirement may change significantly.

That can mean fewer roles, fewer layers or smaller teams. It can also mean that the remaining roles become more specialized, more commercial or more consequential.

This is why the organizational implications of AI are so difficult to model from a simple “jobs lost” perspective. The technology is changing the composition of work, not simply subtracting a fixed number of people from a fixed set of tasks.

The strategic question comes after the efficiency question

For management teams, the starting point should therefore be broader than “where can we automate?”

A better sequence is: where is the economics of the current workflow under pressure? Which activities consume too much management attention? Which decisions are slowed down because people are moving information rather than exercising judgment? Which capabilities will become more important as AI takes on more routine work? And, crucially, if the business releases significant capacity, where should it reinvest it?

That last question is where I think many AI programmes will ultimately be judged.

A company that reduces 20% of a management layer has certainly changed its cost structure. Whether it has made itself a better business depends on what happens next.

AI changes the organization. Management still decides what it becomes.

BMW’s move matters because it shows that this transition is already reaching the organizational core of major companies. AI is no longer simply being introduced as a productivity tool for employees to use alongside the existing way of working. It’s becoming part of the rationale for changing processes, reducing layers and reallocating work.

That will have real consequences for people, and some of those consequences will be difficult. The responsible answer is not to pretend otherwise, nor is it to resist automation simply because the organizational consequences are uncomfortable. It’s to be much more deliberate about what the technology is actually being used to achieve.

The strongest AI transformations will not necessarily be the ones that remove the most work or produce the most impressive short-term efficiency number. They will be the ones that understand which parts of the organization should become smaller, which capabilities should become stronger, and where the capacity released by automation should be put to better use.

AI can change the economics of work very quickly. The harder task for management is deciding what kind of business to build with that new economics.

Sources

1. BMW Group — “Agility, efficiency, AI-based innovations: BMW Group sets course for greater profitability and resilience” (30 September 2026) — primary source for the 20% reduction in divisions and associated management roles, and for AI deployment across the value chain. press.bmwgroup.com (opens in a new tab)
2. Reuters — “Key points of BMW’s strategy update” (30 September 2026) — independent reporting on the restructuring, profitability targets, AI integration and management-role reduction. reuters.com (opens in a new tab)
3. Reuters — “BMW bets on cuts, new models and AI to revive its fortunes” (30 September 2026) — independent reporting on the broader restructuring and the reported employment impact in Germany. reuters.com (opens in a new tab)
4. BMW Group — “Agentic AI is part of our daily business” (May 2026) — BMW’s own examples of AI-enabled workflow redesign, including the reported 90% reduction in previously manual tasks in one fleet workflow. bmwgroup.com (opens in a new tab)
5. McKinsey — “The operating model advantage: Why AI winners are rewiring their organizations” (7 July 2026) — research on operating-model redesign and the limited enterprise-level impact of simply accelerating existing activities. mckinsey.com (opens in a new tab)
6. McKinsey — “What AI reinventors do differently to create value” (9 September 2026) — research on how AI changes decision-making, work design, capabilities and value creation. mckinsey.com (opens in a new tab)
7. Deloitte Middle East — “Deloitte unveils new era of enterprise AI” (4 June 2026) — Middle East findings on efficiency gains, business-model redesign and workforce / workflow readiness. deloitte.com (opens in a new tab)

This note deliberately separates BMW’s explicit use of AI as a lever for organizational streamlining from the wider restructuring of the company. It does not claim that every job affected by that restructuring is being removed because of AI. BMW’s own announcement links AI adoption to leaner organizational structures while setting a 20% reduction target for divisions and associated management roles.