As I prepare for IAA Mobility 2025, I’m reflecting on a growing tension in the automotive world: With the move to SDM, OEMs are expecting a big share of their revenues to come from services in the near future – but realizing their predictions looks challenging right now.

Automakers haven’t been reticent about announcing their multi-billion-dollar revenue ambitions for subscription services. According to one projection, per-vehicle revenue could amount to $1,600 annually by 2035. Another projection suggests that the global market for connected car solutions will total almost $150bn as soon as 2030.

Some of the individual company estimates that go to make up these projections look a tad optimistic to me. Don’t get me wrong – I agree that software-enabled subscription services are going to represent a major slice of OEMs’ future business. But I also believe that the industry urgently needs to adjust its approach if it is to realize the full potential here.

The mounting cost of vehicle ownership

The fact is that the cost of car ownership has become a nightmare for many drivers, who face monthly payments, EV charging surcharges, subscription-based features, maintenance costs, and more.

The rise of “subscription fatigue”

With all these costs stacking up, the vehicle that was once a symbol of freedom can now feel more like a bundle of recurring fees. As a result, customers are likely to be very selective about the in-car services they will pay for. A recent S&P study suggests that willingness to pay for connected car services has declined from 86% to 68% since 2024. Here’s some anecdotal evidence of this trend from me. I’ve recently unsubscribed from a self-driving service for one of my cars. The cost of the subscription just doesn’t seem worthwhile for the amount I use the service. On the other hand, I might have been willing to buy this service for individual journeys, or to subscribe if I could take it with me from one vehicle to another – but the vendor just doesn’t offer that flexibility.

So “subscription fatigue” is a thing – and not only in the context of driving, incidentally. Someone has even come up with a service for getting rid of unneeded (and, in my experience, often forgotten) subscription services.

What this means is that revenues from in-vehicle services can’t be taken for granted. True, some digital natives are already making money in this area, but probably not as much as they expected. That’s true even of those companies that prohibit the use of third-party apps in their cars, something which most OEMs don’t intend to do.

Let’s rethink the proposition from the customer’s perspective

For OEMs hoping to make money from subscription services, the message here is that customers are only likely to be willing to pay for services that they think are going to save them money, or that they will value for some other reason.

So what do OEMs need to do to unlock the service revenues they are relying on? In my view, SDM has to be approached from a customer-centric perspective, not a vehicle- or product-centric one.

Specifically, carmakers need to offer services that genuinely alleviate customer pain points, coupled with service delivery models that are flexible enough to suit every customer. Fortunately, software-defined vehicle (SDV) architecture is ideal for providing that type of flexibility.

OEMs just need to have the will to do it, which means taking on board that services must be delivered to benefit the customer as well as to generate revenue. Who knows – OEMs could even help customers with their subscription fatigue, instead of contributing to it.

Let’s look at what’s involved in practice. It’s helpful to think about this in terms of three strategic initiatives: Reimagine the value proposition around the customer, work with the delivery ecosystem, and optimize the quality of software and the human-machine interface (HMI).

1.    Reimagine the value proposition around the customer

Too often, SDV initiatives focus narrowly on pushing over-the-air (OTA) updates to cars. While OTA is a powerful enabler, it’s not the purpose of SDVs – it’s a tool, and we need to think carefully about the real reasons for using it.

In my opinion, SDVs should:

• Serve the customer’s immediate needs with flexible, personalized experiences

• Future-proof the vehicle, enabling it to evolve in line with technology and lifestyle
• Reduce total cost of ownership, not inflate it through endless monetization

The current model – hefty upfront cost plus recurring subscriptions – should be replaced by new models. For example, customers could pay a lower base price and a subscription for premium features. Or, even better, there could be a standard base price and then pay-as-you-use microservices.

With those models, customers could pay for vehicle features only when they’re being used, rather than for having them available all the time.

This approach shifts the focus from monetization to value creation, making mobility more customer-centric and affordable.

2.    Work with the delivery ecosystem

OEMs can’t achieve this shift of perspective on their own. They’ll need to share the task of delivering customer-centric and affordable services with their ecosystem of suppliers and other partners.

That could happen through smarter partnerships across tech, insurance, and infrastructure, balanced risk-sharing models, and the use of open platforms and APIs to enable innovation at scale.

3.    Optimize software and HMI quality

Trust is essential to securing customer buy-in for SDM, and that kind of trust is heavily dependent on the quality of software and of the HMI. People may forgive a glitchy mobile phone app, but they’re not going to accept glitchy in-vehicle software.

So the quality of software and SMI is now a customer experience imperative. It demands:

• Intuitive, responsive HMIs that adapt to user preferences
• Consistent performance across updates and environments
• Rigorous testing and validation to ensure reliability and safety
• Security and privacy baked into every layer of the software stack

Summing up

I hope I’ve convinced you that SDM needs to be reframed in terms of serving customers, not just increasing company revenues. Services need to be so valuable to customers that they’re seen as a welcome necessity, rather than a luxury. And the delivery model needs a rethink so that customers feel they’re getting value for money.

For those that get this trick right, the rewards will be substantial. The connected car market is predicted to be worth more than $500bn by 2033. If it reaches even a fraction of that, OEMs will be pleased they made the effort to serve, not just monetize, customers.

Join me at IAA Mobility to explore in-vehicle service quality optimization

Are you headed for IAA Mobility 2025? Along with our collaborator Profilence, I’ll be hosting a lunchtime session that’s very relevant to the quality angle on customer-centric SDM. We’ll be demonstrating how AI-powered data analysis can add stability and responsiveness to in-car services – an approach we’ve implemented across 17 infotainment programs.

The automotive industry is undergoing a transformation unlike any seen in its 130-year history. Software-defined vehicles (SDVs), autonomous driving, connected services, intelligent automotive customer experiences – trends like these mean that software is at the heart of the industry’s value proposition.

What’s more, frequent software updates are now mandatory, because today’s vehicles are expected to evolve continuously. Consumers want new features and updates to roll out in cars the way they do on smartphones, so capabilities, performance improvements, and safety patches must all be delivered seamlessly OTA. Digital natives have already shown that OTA updates are a significant competitive differentiator.

To succeed in this new world, OEMs must keep OTA updates in mind from day one of the software development process. That implies:

Development models have yet to catch up with SDV requirements

A major obstacle to meeting these requirements today is that many automakers and tier 1 suppliers continue to rely on legacy development models that are rigid, sequential, and slow.

Their SLM approaches usually embody the old V-model, with its rigid phases and sequential testing, which means that these approaches simply can’t keep pace with the demands of modern automotive software. Trying to manage OTA updates with a legacy SLM approach is likely to result in unacceptable time to market and inconsistent software quality, so that OTA becomes a tactical patchwork rather than a strategic advantage.

What’s more, with these approaches, integrating feedback or fixing bugs can take months once physical vehicle production is complete. That might have worked in the old world of “build and forget,” but it can’t work in the new world of continuous innovation.

The solution: advanced, unified SLM

To achieve their goals, automakers need to introduce a unified SLM approach that embeds OTA and real-time update capabilities into architecture and processes, and that fosters collaboration between functional teams (software, hardware, safety, product, cloud).

This new style of SLM needs to have the following characteristics. It must be based on a hybrid Agile/V-model framework providing speed, safety, and traceability. It must support left-shifting of testing and validation using virtualization and simulation. And it must be part of a wider shift in the company’s architectural focus from hardware to software.

Let’s look at each of these points in more detail.

Left-shifting and virtualization

Left-shifting is about carrying out testing, validation, and verification as early as possible in the development lifecycle, so bugs are caught sooner.

Virtualization makes this possible. Engineers can simulate entire vehicle environments – ECUs, sensors, networks, even real-world driving conditions – before physical prototypes exist. This enables early integration testing of software components, automated regression testing across multiple configurations, and faster feedback between development and testing teams.

Virtualization also allows teams to run thousands of test scenarios in parallel, drastically reducing the time spent on late-stage debugging and physical validation.

Hybrid Agile / V-model framework

Rather than choose between Agile and V-model, the aim should be to integrate the best of both. Agile brings speed, flexibility, and iterative delivery, and enables teams to deliver value in small, manageable increments – perfect for feature-driven OTA updates. V-model, on the other hand, ensures safety, compliance, and traceability – critical in regulated industries like automotive. A hybrid SLM framework combines these strengths. Agile can be used for feature development and iterative refinement (e.g. infotainment, ADAS features). V-model principles can be applied to safety-critical systems (e.g. braking, steering, powertrain control).

In addition, the framework should maintain end-to-end traceability from requirements to deployment, with automated tools tracking changes across all phases. It should include continuous integration and continuous delivery (CI/CD) pipelines with safety gates and compliance checks.

This hybrid model enables rapid innovation without compromising safety or regulatory standards.

Shift of architectural focus from hardware to software

The SLM should be introduced in the context of a more general organizational shift toward a software-centric automotive architecture featuring:

• consolidation of controllers/in-car-compute engines with clear software boundaries
• service-oriented architecture (SOA) enabling modular, reusable components
• open, standardized APIs and communication protocols
• cloud-native development practices for OTA orchestration and analytics

SLM adoption paths

The journey to SLM maturity is about building a hybrid, adaptive approach that can be adopted in phases and then evolve through iteration. To succeed, SLM must be owned by the entire organization, not just software teams. That requires leadership buy-in, cross-function collaboration, and investment in tools and culture.

On this journey, traditional OEMs will encounter different challenges and opportunities from those faced by digital-native disruptors.

With their hardware-centric engineering culture, entrenched V-model processes, and complex, siloed IT and development environments, traditional OEMs need to instigate cultural and organizational changes, as well as technical ones. Indeed, many are already doing so, modernizing software architecture, building OSs, and investing in cloud-native development. Some are creating internal SLM platforms.

Digital natives already have the necessary agility, but may develop technical debt in their systems architecture and, eventually, legacy problems, because of a lack of SLM  rigor. As fleets grow to include millions of vehicles, these companies too need robust SLM backbones to manage software updates across diverse hardware variants, ensuring safety and compliance and maintaining backward compatibility.

SLM as an enabler of industry change

SLM and the automotive supply chain

The transformation of the automotive industry into a software-defined, cloud-connected mobility ecosystem is reshaping the supply chain.

Tier 1 suppliers and semiconductor manufacturers are becoming strategic partners in software lifecycle orchestration, and are evolving into system integrators and platform enablers. As well as components, suppliers may provide reference architectures, cloud-based development and testing environments, OTA management platforms, and fleet-wide diagnostics and analytics via cloud integration.

SLM has to be at the center of this transformation, furnishing the common language and governance layer that binds hardware, software, and cloud services together. It follows that SLM can’t be the sole responsibility of the OEM; it needs to evolve into a shared framework, with responsibility shared across the supply chain.

SLM and AI-first development

To take full advantage of AI, automakers will require a hybrid SLM approach that is optimized for AI. Such an approach must combine not only V-model and Agile but also DevOps (to automate software deployment) and MLOps (to automate AI models’ lifecycles through model training, versioning, validation, and so on).

Left-shifting validation and virtualization will be particularly important in the context of AI, because of the need to start training and validating AI models as early as possible in the lifecycle. (AI itself will help with this requirement since genAI can rapidly produce virtual sensor data, traffic patterns, and environmental conditions to be used in model training and validation.)

Modern SLM approaches need to cater for these needs. They must also support AI model certification, version control, and audit trails to address emerging regulatory requirements.

The road ahead

The next revolution won’t be in the car, but in how we create and manage the software that powers it. As the automotive industry evolves into a global mobility ecosystem, the value of software and connectivity will surpass that of hardware. The winners will be the companies that can deliver secure, reliable, and continuously evolving software at scale, rather than those with the fastest chips or the most powerful engines.

It all starts with efficient, effective, SLM, functioning not just in the OEM’s development lab but at every level of the supply chain, and serving as a catalyst for transformation. An advanced SLM approach can bring genuine competitive advantage.

If you’ll be at IAA Mobility 2025, please come and discuss these ideas with us at Capgemini’s booth, B22 in hall B1. Software is one of Capgemini’s major themes for this event, and we’re lining up software-related speaker sessions and demos to help our clients drive their businesses forward into a future of software-driven mobility.

The automotive industry predominantly operates on single-digit margins, making manufacturing efficiency a top strategic imperative for companies.

However, today manufacturing itself faces significant challenges. New and disruptive products are introducing complexity as well as diversity; in addition, they often necessitate accelerated time to market. This challenge is compounded by others, including talent shortages and uncertainty around tariffs.

Challenges like these could transform the manufacturing footprint, along with the fundamentals of manufacturing efficiency and the management of product diversity.

To overcome the challenges, companies need to take manufacturing efficiency and agility to the next level, so that managing product diversity is no longer a challenge but a source of competitive advantage.

AI-powered automation is already underway

Automation and AI, including agentic AI, can enhance manufacturing efficiency and productivity in a wide range of ways, improving every aspect of production and helping companies hit their “right first time” quality targets, among other benefits. Adding AI plus sensors to robots can elevate them to “cobots” that are able to respond intelligently to their environment.

Already, AI-enabled smart factories are becoming the norm, with predictive maintenance and robotics playing a central role. For example, at Capgemini we have leveraged AI to help a major European OEM design and implement a data-driven manufacturing organization. This work established the operating model and capabilities needed to realize value from data and analytics at scale.

For the future, AI agents show special promise because of their ability to initiate actions autonomously, without prompting by humans.

Fully autonomous agents will take time to arrive

Introducing collaborative and trusted agents to a business requires a transformational journey. This journey begins with the identification of tasks and activities – both repetitive and non-repetitive – that could be undertaken by AI agents. The company must also identify appropriate types of agents (autonomous or not) for each of these activities and tasks.

The most important part of the journey, however, is the alignment of AI agents with human ones, so that their interactions add up to a flawless agentic approach, flawlessly executed.

This journey is likely to proceed stepwise. Initially, AI agents will just take over repetitive and lower-value tasks.

As the model matures, however, agents will be used to perform more advanced tasks, but still with human monitoring.

Eventually, AI agents will be able to anticipate human needs, providing companies with the ability to tackle a wide range of automotive manufacturing pain points.

For example, AI agents will be able to accelerate time to market in response to changing consumer demands. They will enhance automotive sustainability by supporting better traceability, smarter energy use, and more efficient waste management. Everything from productivity to material flows can be improved using these techniques.

Leveraging the collaborative power of AI agents will be essential

Much of AI agents’ strength lies in their ability to collaborate. Adding AI agents to a team can mitigate skills shortages and provide new ways to get the best out of the existing workforce. That is because working with agents empowers people, enhancing their efficiency and effectiveness.

As well as working in teams alongside humans, AI agents could in the future collaborate extensively with one another, each agent being responsible for a specific task. This approach could create immensely flexible and scalable solutions. Ultimately, a team of agents could take over the day-to-day running of a major manufacturing process or even an entire plant, freeing human experts to deal with more demanding work.

Capgemini is actively exploring the potential of agentic AI for automotive manufacturing

Applications of agentic technology that Capgemini has pioneered include a compliance assistant for lighting and signaling devices. This makes processes around 200 times faster and 1,000 times cheaper, as well as facilitating regulatory compliance.

We are currently working on a range of copilots (AI-powered conversational assistants advising and supporting employees) for the entire manufacturing lifecycle. Our line design copilot, for example, can de-risk shopfloor layout configuration and accelerate decision-making, improving areas such as crash detection, ergonomics, and maintainability, and generally enhancing manufacturing efficiency.

Another example from this range is the manufacturing execution system (MES) copilot, which accelerates requirements definition, configuration, and setup, and addresses regulatory topics and documentation creation. There are benefits in terms of both lead times and costs.

As a final example, we are working with agentic AI in the context of Identity & Access Management (IAM). AI can make “role recommendations” that optimize organizational structures from the point of view of both security and efficiency. In addition, it can also be an integral part of identity as a service (IDaaS) solutions.

Implementation of agentic AI in manufacturing is not straightforward

AI-based automation can be a challenge in its own right for several reasons. Some areas of operation may lack the real-time data needed to support AI-enabled decision-making. The diversity of hardware in the ecosystem can cause complexity, as can the need to rethink the operating model.

Before implementation starts, the lack of live use cases may make it difficult to obtain the return on investment stats needed to build a business case. And once a pilot exists, it may prove hard to scale and onerous to deploy enterprise-wide, or even factory-wide.

So how can automotive companies overcome the barriers and succeed with AI in manufacturing?

Four elements provide a foundation for implementing AI in your automotive manufacturing operation

To overcome these barriers, we suggest putting four basic elements in place.

1. Strong data & infrastructure foundation. Connect machines, assets, and control systems securely on the shop floor, and use industrial edge computing to unlock real-time data.
2. Integrated operational systems. Integrate and orchestrate MES, SCADA, and so on into a unified edge-to-cloud ecosystem, increasing data consistency, transparency, and control.
3. Support for intelligent & autonomous systems. This can include industrial AI, machine vision, predictive maintenance, and low-code platforms.
4. Advanced industrial simulation. Use digital twins to enable virtual commissioning, validation, and optimization of industrial systems, applying simulation to everything from IT-OT validation to warehousing.

As well as putting all these elements in place, companies must adopt a strategic approach to this major transformation of their businesses. This includes establishing the governance structures and mechanisms needed to keep a check on autonomous AI-powered systems.

Successful AI adoption in automotive manufacturing depends on specific capabilities

To develop and implement their AI strategy, automakers will require a range of technical and management expertise. As well as hands-on experience with all branches of AI, they need excellent abilities in organizational transformation management, since AI will not work unless it is fully embedded in the manufacturing operation and accepted by the workforce.

In addition, implementers should be conversant with Lean manufacturing principles. That is because to realize its full potential in manufacturing, AI, particularly agentic AI, must interact with Lean. For example, the elimination of waste required by Lean could be achieved more effectively with the help of an AI engineer.

Finally, companies must make workforce empowerment the primary focus of their AI initiatives. When AI enables people to realize their full potential, value can be unlocked from one end of the manufacturing operation to the other.

Of course, Capgemini would love to help you with this journey. Please get in touch to find out more. And, if you’ll be at IAA Mobility 2025, please come and meet our team at booth B22 in hall B1. They’ll be happy to discuss how advanced AI could enhance the efficiency of your own manufacturing operation.

Any inefficiencies, delays, or risks affect a business in real time, and therefore real-time responses to changing business conditions are essential.

Agentic AI makes these real-time responses achievable – one reason why its adoption within supply chains is accelerating rapidly. For automotive companies facing today’s disruptions and uncertainties, agentic AI could therefore be a gamechanger.

Current uncertainty around tariffs is just the latest contributor to today’s high levels of supply chain disruption, described in earlier blog posts like this one. As we reasoned then, AI has a major part to play in helping automotive companies deal with successive waves of disruption. The example given was that by reinforcing forecasting and planning capabilities with AI, companies can respond much more effectively to disruption.

This blog post will look specifically at how agentic AI can enable real-time responsiveness in the automotive supply chain, and thereby build resilience. It follows on from a recent Capgemini article about the value of agentic AI for supply chains in general.

How AI agents can help companies build automotive supply chain resilience

An important application of agentic AI within the supply chain is the use of agents to make persona-based recommendations. The “persona” in this instance could be the COO, CFO, CSCO, or CPO, or a member of their teams.

An AI agent can monitor conditions in a given area against preset business goals. When exceptions arise or thresholds are crossed, the agent can automatically recommend a course of action suited to the persona in question. Areas to be monitored might include:

• The accuracy of supply chain forecasts. The AI agent can assess the risk of forecasting errors occurring and predict their impact on inventory and service levels. Then it can recommend corrective action to the COO’s team, and even, if required, execute the selected actions itself.
• Quality risks and delivery risks associated with incoming parts. “Right first time” is key to automotive quality. The quality of parts received from suppliers, and the reliability of delivery, directly affect the quality of finished vehicles. Agentic AI can assess a part’s quality and delivery risks by analyzing vast sets of historic and real-time data. After comparing the results with the company’s risk thresholds, the AI agent can recommend ways for the CSCO’s team to mitigate unacceptable risks.
• The financial impact of specific supply chain disruptions. The AI agent can compare the impact of different disruptions to help the CPO’s team prioritize possible investments in resilience.
• The drivers of supply chain costs. The AI agent can recommend which drivers could be targeted for cost reduction, freeing up more funds for the CSCO’s team to invest in resilience-boosting measures.

In addition to the persona-based recommendations, agentic AI can further increase resilience if it is used in a coordinating role that transcends all these functions. Drawing data from internal functions and directly from the supplier ecosystem, an AI agent can monitor the entire supply chain, flag up risks, and coordinate the responses of every part of the organization. It can then proactively suggest adjustments to planned actions, internal and external, based on the results.

The potential impact of agentic AI on the automotive supply chain is sizeable and diverse, with research indicating that significant reductions in logistics spend, for example, are achievable. As adoption proceeds, even more business benefits are likely to emerge, and they will become easier to quantify.

AI agents are already in use

Although the concept of agentic AI is relatively new, companies are already embracing it. A recent Capgemini Research Institute report on agentic AI shows that 14% of organizations have implemented AI agents at partial (12%) or full scale (2%). Nearly a quarter (23%) have launched pilots, while another 61% are preparing for or exploring deployment.

Looking now at some functional areas closely linked to the supply chain, 39% of respondents from Operations expect AI to manage at least one process or sub-process daily within the next 12 months, and a total of 75% believe this will be the case within three years. For Finance, the corresponding figures are 30% and 63%.

Despite significant misgivings, including ethical concerns around AI, a sizable 38% of respondents expect AI agents to be functioning as members of human-supervised teams within the next three years.

How automakers can integrate AI into supply chain solutions

The CRI report on agentic AI makes a number of practical recommendations to help companies harness the full power of AI agents. Ensuring smooth collaboration between humans and AI agents is of fundamental importance, and depends on trust; the recommendations therefore center on the need to build trust in AI. Specific recommendations range from addressing ethical issues to overhauling processes, business models, and organizational structures to accommodate joint human-AI teams.

When agentic AI is deployed in the automotive supply chain, an additional dimension of trust needs to be considered. The all-important trust between supplier and OEM could be jeopardized by the introduction of AI agents without supplier acceptance. Therefore, companies should communicate with their suppliers up front about the intended use of AI, addressing any concerns raised. This openness is especially important if AI agents will interact directly with suppliers – for example to reroute transportation in response to a disruption.

Prewave and our commitment to revolutionizing supply chain risk management

Capgemini believes that agentic AI, and AI in general, is pivotal to making the automotive supply chain into a source of competitive advantage.

We’re helping our clients realize that advantage in a variety of ways. One is that we’ve joined forces with Prewave, a leading AI-driven supply chain risk intelligence platform that leverages cutting-edge AI technology to monitor and predict supply chain disruptions.

By combining Capgemini’s expertise in digital transformation with Prewave’s innovative solutions, we deliver end-to-end enhancements in supply chain transparency, compliance, and resilience.

Let’s talk agentic AI and supply chain at IAA Mobility 2025

Capgemini will be at IAA Mobility 2025. Our team will be ready to discuss any of the issues raised in this article, as well as other aspects of the automotive supply chain.

In addition, Capgemini and Prewave will be demonstrating an agentic AI solution that uses data to build a resilient supply chain – one that responds to change in real time, and even proactively.

Please find details of Capgemini’s IAA Mobility activities here. And if you’re coming, do remember to register.