The automotive industry is celebrating two milestones at
once: 140 years since the dawn of the automobile and 30 years of the AUTOMOBIL-ELEKTRONIK Kongress. At this year’s AEK, Magnus Östberg, CSO at Mercedes-Benz,
used his keynote “140 Years of Innovation and 30 Years of AEK – The Automotive
Industry in the AI Era” to show how vehicle electronics and software have
accelerated innovation, changed collaboration models and opened the next chapter of software-defined vehicles.
As the executive responsible for MB.OS,
Mercedes-Benz’s software-centric vehicle architecture, Östberg is now
focused on what happens after the platform reaches customers in the field. In
the following interview, he explains how real-world data, localisation, open
source, software monetisation and AI agents will shape the next phase of
automotive software.
AEK: Mr Östberg, in our previous
conversations we spoke extensively about the development of MB.OS and how
Mercedes-Benz is turning software into a product. Would you say that the
build-up phase of the software-defined vehicle (SDV) is now complete? Has the
real test in the field begun?
Yes, absolutely. We now have our software-defined vehicles
on the road across global markets. This is an incredibly exciting phase for us
and, of course, a very rewarding one because we have spent many years preparing
for this transformation. Today, we are no longer talking about a single SDV,
but about multiple model series, vehicle architectures, powertrains and brands.
Ola Källenius once described it as the biggest rollout in Mercedes-Benz's
history. We started with the new CLA, naturally with our
battery-electric vehicles. Since then, we have expanded to hybrids, the
S-Class, premium combustion-engine models and our vans.
What makes this phase particularly interesting is the data
we are now receiving. For the first time, we are getting real-time feedback
directly from vehicles in the field. Which functions do customers actually use?
How well do they understand them? Which features do they appreciate? That is
precisely why this stage is so important.
The SDV also raises the question of how software can be monetised inside the vehicle.
While this may not be your primary responsibility, how important is this aspect
for your work and your collaboration with the product teams?
One of the greatest advantages of the software-defined
vehicle is that many discussions no longer need to be based on assumptions.
Today, we can use real-world data to understand what works and what does not.
In the past, many decisions relied largely on hypotheses. Naturally, you still
need a vision of which products and functions you want to develop. But whether
a feature is actually adopted by customers, whether it needs refinement and
whether those refinements improve its performance can now be measured much more
accurately. That fundamentally changes how investment decisions are
made. Ultimately, it becomes a very concrete question of return on investment.
Which function delivers real customer value? How widely is it used? And what
commercial impact does it generate?
For a premium manufacturer, this is a particularly
sensitive issue. Customers expect certain features to be included as part of
the premium experience. At the same time, new digital services can be offered
separately. Where does Mercedes-Benz draw the line?
Our customers have been familiar with the vehicle
configurator for many years, certainly in Europe. They choose the options they
want and can clearly see what each feature costs. With the software-defined vehicle, what changes is often
not the product itself, but the timing of the decision. Customers no longer
have to decide before delivery which digital functions they want to use. They
can activate them after placing the order, after taking delivery of the vehicle
or even one or two years later. A second owner can also choose to unlock
additional functions. The crucial point is transparency. Customers must
understand exactly what they are paying for – whether it is one month, one year
or the entire lifetime of the vehicle. That is one of the fundamental
differences compared with the traditional vehicle purchasing process. Software
makes the offering far more flexible and gives our customers significantly
greater choice.
In other words, customers are no longer under pressure to
decide at the point of purchase whether they need functions whose value they
may only fully appreciate in everyday use...
Exactly. Of course, there are still decisions that are
determined by the physical hardware. Certain components simply have to be
installed in the vehicle. If I need an additional camera for specific ADAS functions, that camera must already be
fitted. Likewise, you cannot change the number of cylinders in an engine
through software afterwards. However, many digital functions can be activated,
adapted or expanded throughout the vehicle's lifetime. That fundamentally
changes the relationship between the customer and the vehicle.
You have already mentioned data. How does Mercedes-Benz
actually measure which software functions create value for customers?
For that purpose, we have built our own data pipeline.
Today, around 17 million vehicles are connected to our backend. There is a
well-known saying that data is the new oil. Personally, I would rather say that
data is the new lubricant. It is not simply a raw material that you sell.
Instead, it enables the entire system to operate more effectively. For example,
we can analyse how navigation usage changes before and after a software update.
How often is navigation being used? How many successful route guidance sessions
are completed? Are customers switching to alternative navigation apps instead?
If they are, what is missing from our own solution?
We ask these kinds of questions across all relevant vehicle
domains. With voice control, for instance, we analyse how many voice commands
users issue and how many of them result in successful actions. We are not
interested purely in usage statistics; we are interested in successful usage. To
support this, we have established a comprehensive programme covering data
visualisation, dashboards and KPIs. Our data strategy provides the foundation
for systematically feeding these insights back into the development process.
How do you decide, based on this data, which functions
should be further developed, modified or perhaps even discontinued?
We work very closely with our Sales and Marketing teams, but
above all with our global market organisations. The relevance of individual
features depends heavily on the respective market and customer group. Take the
US market as an example. There, we see significantly higher demand for
functions such as Dog Mode, which allows pets to remain safely inside the
vehicle under specific conditions. Features such as quick ordering through
services like Starbucks are also considerably more relevant there than they are
in Europe.
These requirements vary greatly from region to region. That
is why close collaboration between development, product management and our
market organisations is essential. We work using agile methodologies based on
SAFe. Together, we create our backlogs, and priorities are driven by data.
Features or KPIs with the highest relevance move to the top of the backlog. In
this way, development becomes much more customer- and market-driven.
The automotive industry is currently placing great
emphasis on localisation. In the past, vehicle concepts were often developed
globally. Today, there is a much stronger focus on local production, local
engineering and local customer requirements. Software creates new
opportunities, but also adds complexity. How does Mercedes-Benz strike the
right balance?
For us, global standards are the key. We work within clearly
defined guardrails. Every market knows which standards must be followed and
where regional adaptation is possible. In our largest markets, we have local
teams responsible for specific applications and local partners. These partners
are often the local champions within their respective regions. South Korea is a
good example. We have local teams there that understand exactly what customers
expect in terms of features, interaction and digital services. Localisation
goes far beyond language or navigation. It also includes the overall look and
feel, as well as the way users interact with the vehicle. Language and culture
play a major role in shaping how people engage with digital vehicle functions.
Are successful local ideas also transferred to other
markets?
Absolutely. We actively encourage the exchange of ideas.
Internally, we call this approach "Copy with Pride". We have been
promoting this principle for years. The objective is to share successful ideas,
even if the source code itself cannot always be transferred directly. It can
still serve as valuable inspiration. Developers can evaluate how a particular
feature could be adapted for use in other regions. A good example is karaoke.
We found ways of making such services legally compliant in different countries
while preserving the overall customer experience.
Open source has become an important topic in the context
of software-defined vehicles. Mercedes-Benz is
actively involved in projects such as Eclipse S-CORE. Where do you draw
the line between differentiating and non-differentiating software?
You are absolutely right. Open source is a key strategic
element for us. Together with BMW and Cariad, we initiated the Eclipse S-CORE
project. Since signing that agreement, we have consistently pursued this
approach. A good example is diagnostics. We contributed our diagnostics
software upstream, making it openly available through the project. The reason is quite straightforward. Every manufacturer
requires high-quality diagnostic software. We had already developed a very
comprehensive and capable solution.
However, from the customer's perspective,
it is not important who originally created or defined the diagnostics software. If we can jointly develop or provide these must-have
capabilities, we spend far less time reinventing technology that everyone needs
anyway. Instead, we can invest those resources in features that genuinely
differentiate our products from a customer's perspective.
What is the primary driver behind your open-source
strategy: faster development, lower costs or creating more capacity for
differentiating innovation?
In reality, it is all of those factors. For me, however,
speed is the most important one. We want to accelerate development while
simultaneously reducing costs so that we can bring more compelling products to
market. The less time and fewer resources we have to spend on capabilities that
every manufacturer requires, the more we can focus on innovations that truly
matter to our customers.
In this context, one regulatory topic is particularly
important: the Software Bill of Materials. Is the SBOM at Mercedes-Benz already
an operational management tool, or is it primarily a compliance issue?
Our end-to-end software journey already incorporates the
SBOM. It is part of our digital supply chain. When we deliver software to our
end customers or into our vehicles, that supply chain is also digitally
connected. This helps us provide authorities and partners with a transparent
view of which software components are being used. So it is not merely a
compliance issue in the narrow sense, but an integral part of the software
supply chain.
At the beginning of March, Mercedes-Benz also published a
FOSS Disclosure Portal and the Disclosure CLI on GitHub to provide further
transparency. What is the idea behind this?
The idea is to follow words with action. The Disclosure
Portal makes transparent which open-source components and licences are included
in our products. When we contribute to open source, we do not do so merely
symbolically. We also maintain the code we provide and automatically integrate
it into our build system, meaning our CI/CD chain. This is not only about code
we have made public ourselves. Code published by others can also be integrated
into our systems. Proprietary code that is not competitively differentiating
for us and performs similar functions is reworked or replaced accordingly. We
set that proprietary code aside and use the open-source repository instead.
That, in turn, creates the opportunity to contribute
improvements we discover elsewhere back upstream. That is how open source
works: all participants collaborate in a shared repository. This repository is
integrated into the company’s own development environment, an end product is
created from it, and the entire chain of continuous integration, continuous
delivery or deployment, and continuous testing has to function automatically. Only
then can we be legally compliant in the end. We must document open-source
licences properly, provide the relevant information in the portal and be able
to show authorities or partners which licences and which code have been used.
All of this is integrated into our build system.
With the software-defined vehicle, traditional vehicle
development, IT and software engineering are moving closer together. What does
your collaboration with Katrin Lehmann and her IT team look like?
It is a very strong partnership. The IT organisation under
Katrin Lehmann is responsible for our base systems and for many systems used
across the organisation outside R&D. Collaboration in the cloud environment
is particularly important. There, we have a mixed situation: we use services
from the general IT landscape, but we also have specific requirements from
research and development. AI makes this even more important. In a world where agents,
resources and capabilities are located in different places, very close
communication with IT is essential. This is not only about technical
efficiency, but also about cost efficiency. Token and cloud costs have become
relevant to competitiveness.
Finally, if we speak again in two years’ time, what would
indicate to you that Mercedes-Benz has taken the next step with the SDV?
We will certainly be talking about the future S-Class
by then. That will be the next major step in our SDV development and in vehicle
intelligence. Customers will be able to look forward to more autonomous
functions. AI agents will make their lives
significantly easier. And then we will probably be talking about how slow we
still were back in 2026 (laughs). Seriously, the pace will continue to
increase.
