A manufacturing company consists of layers that rarely follow the same clock. On the shop floor, it revolves around planning, quality control, maintenance and logistics within the premises. Above that sits engineering: design, specifications, test reporting. Above that again sits purchasing, order processing and the administrative handling of suppliers and customers. Each layer has its own ratio between routine work and work that requires judgment, and that ratio determines how sensitive a layer is to shifting once AI can take over part of the work.
What makes the sector different is that much of this work is tied to physical machines, certification requirements and safety regulations. This does not slow down the takeover of administrative or analytical tasks, but it limits where automation stops and where oversight remains mandatory. A planning model can optimize a schedule; whether that schedule is actually executed depends on machine availability, material supply and people on the floor who make the final decision.
In parts of the administrative and analytical chain, the shift is already visible. Order processing, inventory analysis, reporting on quality data and drafting first versions of technical documentation: these are tasks where a growing share is done entirely by AI or where an employee mainly still approves or rejects, with reason. That is not a future scenario. At companies that have their data in order and their processes documented, this is already the case today.
At companies where data is spread across separate systems, where process documentation is outdated or where quality control relies largely on the experience of individual employees, that same shift proceeds more slowly or does not occur at all. Not because the technology is less capable there, but because the preconditions for applying it are missing. That difference between companies is the first thing a management team should be able to place before drawing conclusions about what AI does or does not do in its sector.
A strategy rests on assumptions about where capacity sits, where costs weigh heavily and where competitors are faster or slower. If an assumption states that quality control costs a fixed number of FTEs, and part of that control is now being supported or taken over by a system that detects deviations, then that assumption is no longer true without anyone having established that. The strategy on paper still checks out. The reality underneath has shifted.
This is precisely why the shift rarely stands out at the top. A management team steers on margins, lead times and market position. The question of which tasks are now partly or fully done by AI sits one level deeper, in the hours of engineering, planning and administration. Only when the combined shift in those layers is large enough does it show up again in the figures the management team steers on, and then usually as a surprise.
This dynamic is not unique to the manufacturing industry. In the transport sector, the pressure shifts in a similar way through planning and route determination, as shown in how AI is changing strategic pressure in transport and logistics. In professional services, the center of gravity lies precisely with analysis and reporting, described in the shift of strategic pressure in professional services. What these sectors have in common with manufacturing is not the task itself, but the pattern: tasks that on paper belong to people are already partly shifting in practice to systems, at varying speeds per company.
The question of which work in a specific company can genuinely be taken over by AI is not a question that can be answered at sector level. It requires a look at the company's own FTE distribution, its own data quality and its own processes. Who within the management team should put that question on the agenda, and with whom, depends on how the organization is structured; an overview of who should be involved in a strategy review offers a starting point for that. For companies that do not want to wait for the annual report to see whether their strategy still matches reality, there is also a way to test that in the interim, as described in how to measure whether a strategy still works without waiting for the annual cycle.
This deliberately does not touch on the question of what a company does with freed-up capacity. Whether freed-up hours lead to redeployment, an expansion of job duties or something else is a decision for the employer, with its own legal requirements when it involves personnel consequences. What is at issue here is the question that precedes it: which work is already partly or fully done by AI today, and which assumption in the strategy has thereby lapsed without anyone having noticed it.
The work scan from FTE TO AI answers the question of which work in a company, task by task, can be taken over by AI, partly with oversight or not. For a management team that first wants to know whether its strategic assumptions still hold, there is the free assumption check: a short round in which you name your key assumptions and see, per assumption, when it was last confirmed. The full pressure test, with sector data, a self-plot of the management team and an ongoing assumptions list, is under construction.
Stel uw vraag. Vaak zit de echte vraag een laag dieper — daar mag ik naar vragen.
Answers come from this site’s knowledge base. Not tailored advice, and not a scan of your company.