The Munich car manufacturer wants to be an “AI-orchestrated company” from 2028. What this means becomes apparent in purchasing. Autonomous agents will work there in the future.
BMW board member Nicolai Martin: AI transformation in purchasing. Photo: BMW, Getty Images [M]
BMW aims to become an “AI-orchestrated company” by 2028. Purchasing is considered a pioneer for the other areas.
BMW is using four AI agents in purchasing, which are expected to save billions in the long term.
The background: pressure on the Chinese market, trade policy distortions and stronger competition in Europe, which is forcing car companies to save money.
The summary was created and editorially checked using AI.
Munich. At BMW they like to speak in abbreviations. Best in a triad, like BMW. Purchasing director Nicolai Martin (internally “the M”) presents at the FIZ (Research and Innovation Center) how BMW wants to achieve OM 3.0. It is probably the most important project in the coming years.
The abbreviation stands for “Operating Model 3.0”. This represents the final stage of the major restructuring with which the car manufacturer is repositioning itself thanks to artificial intelligence (AI).
The goal: BMW wants to transform itself into an “AI-orchestrated company” in the next two years. But what does that actually mean exactly?
Pressure on the Chinese market, trade policy upheavals, stronger competition in Europe – given the many challenges, margins and sales figures in the auto industry have shrunk massively in recent years.
CEO Milan Nedeljkovic has therefore announced a “transformation program” for the entire group. Integral component: more intensive use of AI. According to the BMW boss, this program is intended to “link the entire structure of our company very strongly to value creation”. And save billions.
AI: From assistant to agent
The company wants to implement this in three steps. In the first stage, AI supports employees with individual tasks, such as analyzes or tenders. In the second, AI agents take over entire processes: they access various data and systems and prepare complex decisions.
In the third stage, several agents work together across departmental boundaries. For example, they link information about supply chains, capacities and costs and derive concrete options for action from this.
BMW is currently on the way from “OM 1.0” to “OM 2.0”. Assistants should become agents who in the future will not only provide answers but will increasingly be able to carry out tasks themselves. The third stage is scheduled to follow in 2028.
Vanessa Cann, Managing Director at Accenture: The automotive industry is reacting to high cost pressure. Photo: Photo Vogt
Vanessa Cann, Managing Director at Accenture, supports companies in similar transformations. She says: “Companies can’t afford to spend two years developing and only rolling out agents once everything is ready.” We should think “top down”, i.e. from above. This creates the conditions to proceed faster and more strategically and to “find the big levers,” says the consultant.
The automotive industry in particular needs to look for “radical measures” given the high cost pressure, says Cann. AI is a technology that many companies rely on.
Board member Martin sees purchasing as particularly suitable for this transformation. He is convinced that German companies can not only keep up with China and the USA in the industrial use of AI, but can also take a leading role.
Martin is not alone in his confidence: According to the “BCG AI Radar 2026”, German business leaders would expand their AI investments more often than Chinese and US ones – even if the hoped-for financial goals are not achieved in the next twelve months.
With his department, he wants to prove that AI can not only accelerate processes, but also reduce costs and strengthen the resilience of supply chains. To this end, the purchasing manager has adapted the most important processes in his area to the use of AI. "We have set up a kind of subway network for our purchasing processes - from supply and tendering to negotiations and awarding. Here we can always apply the AI applications precisely," says Martin.
Purchasing is particularly suitable for this: BMW supplies 31 factory locations, works with around 2,600 main suppliers and manages the logistics for more than 36 million parts every day. According to Martin, the group covers around 70 percent of its value creation through its supplier network. The annual purchasing volume is 80 billion euros. Decisions in this area have a direct impact on costs, speed and security of supply.
He initially wants to save at least a mid-three-digit million amount with AI agents. But that should only be the beginning. “If I have a reference base of 80 billion, of course I want to think quickly about billions and not about millions,” he says.
Nicolai Martin: The BMW purchasing director has gained four new team members with the agents. Photo: BMW
Martin is less concerned with his own personnel costs than with savings in business with suppliers. “The fixed costs for my buyers are in the mid three-digit million range,” says Martin. Making this area more efficient brings comparatively little compared to the potential that it could tap into given the purchasing volume.
Consultant Cann reports on a software customer who achieved savings of 17 to 35 percent, depending on the area. The “agency” of purchasing is therefore a major project for BMW.
Four new colleagues for purchasing
For this purpose, BMW first built the “Logic Hub”, a common database, which connects world, company and purchasing knowledge and uses this to create the business context for the queries of the purchasing employees.
At the start of his term of office 16 months ago, Martin made AI and digitalization a priority. The manager first announced the new agents in an interview with Handelsblatt:
Mr. Flexibility: The first agent already in action. The AI-supported tool is intended to support procurement and thus reduce flexibility costs.
Ms. Resilience & Risk Management: The AI colleague is supposed to support the group in keeping geopolitical risks under control. According to Martin, this tool is already well advanced.
Ms. Product Quality: The assistant is intended to help ensure the quality of components and suppliers at more than 30 locations worldwide.
Mr. MatKo: The agent should use the large purchasing database to reduce material costs across different areas.
AI agents in the background: BMW purchasing manager Martin wants to “expand the management circle to include agents who have their own persona”. Photo: BMW
The agents' names are deliberately humanized. BMW does not want the systems to be seen as more software that buyers have to use. In the future, the agents should take on tasks independently and interact with each other – like real employees. They should be able to interact with each other by the end of the year.
Martin will present an example of how an AI assistant can become an independently working AI agent in the long term in the Research and Innovation Center together with his division manager for battery cell purchasing, Hendrik Lang, and the product manager, Maik Leuthold: Mr. Flexibility.
This is how Mr. Flexibility works
Especially in view of geopolitical risks such as tariffs and trade barriers, it is crucial in purchasing to solve problems quickly and check early on whether additional vehicles are even profitable. In the past, purchasing would sometimes already agree to production without knowing exactly the costs, explains Martin. “Then we built more cars that customers were happy about, but that didn’t do much for us when the volume was higher.”
The BMW managers show how Mr. Flexibility should now prevent this using the example of the battery cells for the electric SUV iX3. Leuthold opens up the agent and asks his question: “Do we have enough capacity to build 10,000 more iX3s in 2028?”
BMW board member Martin: “Agentization” of purchasing. Photo: BMW
“For the simple question about 10,000 additional iX3s, I have to connect the data from three different departments and then answer the question manually from this data,” explains cell buyer Lang. The batteries and their cells are among the most important, yet most expensive and most strictly regulated parts in an electric car.
Mr. Flexibility first checks whether he has understood the request correctly and what data is required. The assistant then accesses the available planning and capacity data and merges the information. In response, the AI generates an analysis graphic. It shows either green or red tiles per supplier and month - green stands for sufficient capacity, red for a bottleneck.
Product manager Maik Leuthold: Green fields stand for availability, red ones for bottlenecks. Photo: BMW
Within a few minutes it becomes clear: In order to build and sell more than the initially planned units, BMW needs additional cells. Not all of them are easy to obtain.
Mr. Flexibility suggests various measures, such as pre-production in a certain time window with free capacity, delivery from China or swapping capacity between suppliers' locations. At the same time he lists the costs for this, a six-figure sum.
By the end of the year, Mr. Flexibility should be able to check capacities for the most important vehicle components of the new “new class” models and, if necessary, also suggest measures for purchasing. In the future, the agent could even independently write the confirmation of the construction program into the colleagues' production system overnight, says Martin. But BMW is not there yet.
41 percent of working hours in the automotive industry can be supported or automated with AI. Accenture comes to this conclusion in the study “Europe's AI Reckoning”.
Related topics
BMWAccentureChinaEuropeArtificial IntelligenceSoftware
To ensure that the transformation into an AI-orchestrated group is successful across departmental boundaries, Cann calls for group-wide standardization. Otherwise there is a risk of delays. Agents are still largely thought of in silos, says the consultant. If you initially concentrate on the easier-to-use potential within a silo, you run the risk of each department building on its own platforms and with its own standards. “Then you have to do a lot of cleaning up,” says Cann. A group-wide platform strategy is therefore needed right from the start.
More: How CEO Höttges wants to turn Telekom into an AI group
Published according to the editorial standards of the Handelsblatt. You can find more information in our guidelines.
More on the topic of our partners display
remind.me Take advantage of current low electricity/gas prices before prices rise again
