Much of what employees experience as AI today is initially a new way of working with information.
For many people in companies, their first contact with generative AI consists of asking questions, bringing information from different sources together, summarising texts, formulating content or making large volumes of information accessible more quickly.
That is already valuable. And modern AI can do considerably more: it can automate individual activities, analyse data, generate content and increasingly initiate actions within processes.
Even so, I believe we should distinguish between a powerful technology and a genuinely more intelligent company.
A company does not become more intelligent simply because an AI system is now running alongside ten existing systems.
What matters is whether information becomes more accessible, decisions can be made faster and on a more reliable basis, errors decrease, processes become simpler or employees gain time for more valuable activities.
The increasingly scarce resource will be available human capability.
Germany does not face a blanket shortage of labour across every occupation. At the same time, the German Federal Employment Agency continued to identify 157 occupational categories with skilled-worker shortages for 2025. Demographic change will further shape this discussion over the long term. 1
For companies, this means we need to think more carefully about which activities we use the people available to us for.
Where technology can reliably take over routine work, I believe this opportunity should be examined consistently. Not with the objective of removing people from companies as quickly as possible, but of deploying human capacity where it creates greater value.
The International Labour Organization reaches a similarly differentiated conclusion regarding generative AI: around one in four workers worldwide is employed in an occupation with some degree of GenAI exposure. Because human input remains necessary, the ILO expects transformation rather than complete replacement for most occupations. 2
Automate recurring and clearly defined activities where quality and control can be ensured.
Condense data, documents and knowledge more quickly so that people do not spend their time searching for information.
Make patterns, deviations and alternatives visible without confusing the system with responsibility.
Use the capacity gained more strongly for customers, improvement, innovation, leadership and difficult exceptions.
Not every technically possible form of automation improves the customer experience.
Customer interaction is therefore one area where I see a boundary that every company should consciously define for itself.
AI can of course create significant advantages in customer service: finding information faster, summarising cases, classifying enquiries, dealing directly with simple issues or providing an employee with the full context of a customer enquiry before the conversation begins.
But in my view, the customer should never feel that technology is standing between them and the company.
A Gartner survey of 3,566 B2B and B2C customers conducted in spring 2026 illustrates precisely this ambivalence: 50 percent said GenAI could make their service interaction easier. At the same time, 87 percent considered the ability to reach a human representative essential. 3
Good customer service uses AI to make people faster and better – not to make people inaccessible to the customer.
Perhaps this expectation will change over the coming years. Today, however, I would examine very carefully where automation creates convenience – and where it costs trust.
Every gain in efficiency also creates a new dependency.
An automated warehouse process is a good example. Automation can increase speed, standardise processes and significantly reduce the amount of manual work required.
At the same time, the risk profile changes.
Dependency on available labour decreases. In return, dependency on power supply, control technology, networks, software, interfaces, data, spare parts, technical service and increasingly external cloud and platform services increases.
That is not an argument against automation. Quite the opposite: in many areas, automation is economically and operationally absolutely sensible.
But a business case that considers only labour savings, speed and productivity is incomplete in my view. It must also answer what new dependency the company is taking on in return.
An IT problem rarely remains inside IT.
One experience from the early years of my operational career has stayed with me to this day.
At a large industrial site, a system migration temporarily prevented goods from being released from the warehouse as planned. A vehicle I had scheduled was consequently tied up for almost a week.
That was more than two decades ago, and the technological capabilities of that time are hardly comparable with today. The lesson behind the experience is still relevant.
The IT problem did not remain on a screen. It stopped a physical process.
The more tightly operational processes are digitally connected, the more directly errors in one system can affect the real performance of the company.
An incorrect data record, a faulty interface, a cyberattack, an unavailable cloud application or a prolonged failure of critical infrastructure may then do more than delay administration. In the worst case, the operational process stops.
Mid-sized companies in particular should ensure that AI does not simply become another layer of complexity.
Many companies already operate historically grown IT landscapes. ERP, CRM, production systems, warehouse or supply-chain systems, Excel solutions, individual applications and numerous interfaces exist alongside one another.
If additional AI applications, agents, data platforms and further interfaces are now placed on top without a clear architecture, the result is not automatically a more intelligent company.
It may simply become a more complex company.
The NIST AI Risk Management Framework and its accompanying Generative AI Profile therefore deliberately take a risk-based approach. Organisations are encouraged to identify, assess and manage the risks of AI systems systematically throughout the technology lifecycle. 4
New technology should make complexity more manageable. It should not simply add new complexity.
The label “AI” is not yet a business case.
AI has naturally also become a sales argument for enterprise software. There is nothing inherently problematic about that. Many new capabilities can create genuine value.
For me, the problem begins when the technical label becomes more important than the question of which concrete operational problem is actually being solved.
In 2026, for example, Gartner explicitly warned of “agent washing” in the supply-chain-planning technology market – the practice of presenting existing or limited capabilities as more advanced through the use of the new agent terminology. Gartner recommends a step-by-step approach based on robust use cases, data and governance. 5
For me, that is not criticism of software vendors. It is a reminder of the buyer's responsibility.
A company should not buy AI because a vendor sells AI. It should use technology because a relevant problem can demonstrably be solved better with it.
Six questions interest me more than the next product presentation.
Before another technological layer is introduced into a company, I would first want answers to several very simple questions.
Not: What can the tool do? But: Which existing process will measurably improve?
Time, quality, speed, cost, customer experience or better decisions?
What happens when results are wrong, and how are errors detected?
Providers, data, interfaces, cloud, infrastructure, know-how or individual key people?
Can we continue operating? For how long? And what minimum level of operation remains possible?
Where does technology decide – and where must a human deliberately decide or be able to intervene?
Efficiency is not the same as resilience.
A company can be extremely efficient during normal operations and at the same time highly vulnerable during a major disruption.
The more strongly processes are automated, digitalised and interconnected, the more important the question becomes of how the company responds to a failure.
This does not mean that every modern solution should be backed up by a fully manual parallel process. That would often make little economic sense.
It means understanding which processes are critical, which dependencies exist, which systems represent single points of failure and what minimum level of operation must be maintained during a crisis.
Use AI. But with judgement, ambition and responsibility.
I consider artificial intelligence to be one of the technologies that will significantly change the way companies work over the coming years.
Precisely for that reason, we should neither underestimate what it is – nor exaggerate it.
It can relieve people of routine work, make knowledge available, automate processes, reveal patterns and improve the preparation of decisions.
At the same time, it remains a technology that can make mistakes, fail and create new dependencies. Responsibility, judgement and the design of a resilient company remain leadership responsibilities.
For me, the real opportunity therefore does not lie in building the most “AI-heavy” company possible.
This article reflects my personal operational perspective. The following sources provide context for selected statements and current developments.
157 occupational categories with skilled-worker shortages.
02 · INTERNATIONAL LABOUR ORGANIZATION Generative AI and Jobs: A 2025 UpdateGenAI is expected to transform more jobs than it completely replaces.
03 · GARTNER GenAI in Customer Service87% of respondents consider access to a human representative essential.
04 · NIST AI Risk Management Framework · GenAI ProfileRisk-based framework for the responsible use of generative AI.
05 · GARTNER Agent Washing in Supply Chain Planning TechnologyWarning against hype and recommendation to distinguish real capabilities, robust use cases, data and governance.
When new technology creates more questions than problems it solves.
Brueggemann Interim assumes operational leadership responsibility on an interim basis when processes, systems, information and decision-making paths need to be brought back together into one manageable operating picture.