NO TIME TO READ THIS WEBSITE?
urope's and North America's automotive sectors are under increasing pressure as Chinese OEMs rapidly produce high-quality vehicles at significantly lower costs, especially in markets where price sensitivity is rising.
In manufacturing alone, cost gaps can reach 60–75%, with direct labor being a major driver. Labor costs in Europe are eight to ten times higher than in Asia, yet the output per worker is falling behind. While China recorded a 4% increase in labor productivity between 2023 and 2024, Germany saw a 4.3% decline from 2024 to 2025. However, a significant portion of the cost gap is controllable through automation, tooling, facilities, and equipment choices.
Figure 1 – Labor productivity China (2023-2024) and Germany (2024-2025)
Source: Berylls by AlixPartners
Furthermore, the shift to EVs has eroded traditional advantages of established OEMs. EV designs are typically less complex than ICE designs, enabling more streamlined assembly and higher levels of automation. However, many legacy OEMs operate out of brownfield plants and rely on production systems originally designed for ICE vehicles. While these legacy setups once represented operational excellence, they now often come with rigid layouts, outdated tooling and equipment, and limited digital integration. These structural constraints limit attempts to modernize brownfield environments because they typically cause high costs and operational disruption. Implementing automation or reconfiguring lines for EV-specific purposes is not only technically challenging but also financially demanding. Consequently, legacy OEMs are less efficient than purpose-built greenfield plants, which further widens the productivity and cost gap. These dynamics also extend to automotive suppliers, who face similar challenges.
Advanced manufacturing, including the use of AI, can help close the gap. However, these solutions are not plug-and-play. Value depends heavily on solid foundations, such as clean data, redesigned processes, and personnel who have been upskilled. A useful rule of thumb is that in AI programs, roughly 10% of the benefit stems from the algorithm, 20% from technology and data, and 70% from people applying insights to day-to-day decisions. This ratio underscores a critical truth: the applicability and success of AI depend heavily on organizational capabilities and employee AI literacy. Although AI has evolved from a buzzword to a necessity due to its increasing application, upskilling the workforce remains limited. The question is no longer whether AI is helpful, but rather, how to quickly deliver pragmatic, measurable impact.
Figure 2 – Where AI value really comes from
Algorithm
Technology & data
People
Source: Berylls by AlixPartners
In automotive plants, the largest portion of controllable losses occurs during the “make” phase of the supply chain due to process instability, scrap, micro-stops, and cycle time variance. These issues directly impact profit and loss. AI improvements directly translate into cost savings and increased throughput per vehicle without requiring new platforms or supplier renegotiations. The impact is tangible: reductions in scrap of 10-30%, improvements in first-pass yield of 5-10%, and reductions in unplanned downtime of up to 50% – often within a single quarter. By reducing breakdowns and rework, AI maximizes output from existing assets, defers capital expenditures, and unlocks capacity before new equipment is needed. These improvements have a fast time to value because the necessary data – from MES, PLC, vision systems, and quality records – is already available in most plants. AI solutions are designed to support existing operations rather than disrupt them.
Instead of invasive line rebuilds, manufacturers benefit from tools like AI-powered set-point guidance, which are systems that continuously analyze production data to recommend the optimal machine settings for quality, speed, and efficiency. These tools support human operators and automated systems, helping them make better, faster decisions and improve performance without interrupting workflows. Predictive maintenance builds on this foundation to take operational efficiency a step further. Leveraging AI and Internet of Things (IoT) technologies, manufacturers can anticipate equipment failures before they occur. Machine learning algorithms detect patterns in historical and real-time data that indicate early signs of wear or malfunction, and anomaly detection models flag unusual behavior. This approach is particularly valuable for rate-limiting assets, helping manufacturers avoid unplanned downtime, maintain consistent throughput, and extend equipment life. Another predictive use case is AI-driven material ordering, which forecasts subcomponent shortages by analyzing production schedules and supplier lead times. This helps minimize bottlenecks and maintain a steady flow of materials for uninterrupted assembly.
These manufacturing advancements improve not just plant-level performance but also have a ripple effect across the entire supply chain. More stable production enables better forecasting, more reliable delivery schedules, and fewer returns. Thus, AI-driven optimization of manufacturing processes becomes a strategic lever for driving systemic improvements, transforming not just operations, but also the broader ecosystem they support.
The transition from concept to impact is difficult, not because AI models are inadequate, but because most automotive plants are designed for launches, firefighting, and cost reductions, rather than data-driven operations. Systems, processes, and structures were never designed with AI in mind. To identify these gaps, we apply two sequential lenses:
The Operations Maturity Assessment identifies structural bottlenecks in current performance, such as chronic bottlenecks, scrap hotspots, and launch instability, and prioritizes areas where AI can have the greatest impact. The second is the AI Readiness Assessment, which evaluates seven capabilities across more than 100 criteria to determine whether a plant can reliably scale those use cases beyond a single line or program.
Figure 3 – AI Readiness Assessment
Source: Berylls by AlixPartners
AI Readiness evaluated across seven capabilities
1. Data readiness is limited by fragmented legacy systems. Shop-floor data is static and is only collected at the start and end of processes. Resource analysis focuses on past states and requires significant manual effort. Poor quality, limited accessibility, and minimal external integration further hinder data readiness, creating a major barrier in AI projects.
2. Automotive plants typically run on a patchwork of technical infrastructure, including outdated PLCs and aging MES releases locked into vendor contracts. Any change to a line recipe, interface, or dashboard requires weekend downtime and long validation cycles. Cloud adoption is often blocked by corporate security or cost concerns. Consequently, promising AI pilots reside on “shadow IT” laptops and local servers, while the official production stack cannot absorb or operate them at scale.
3. Process integration is rigid and fragmented. Decision-making is rarely digitized, interoperability is weak, and KPI tracking inconsistent. Scaling AI from pilot to production is slow due to the lack of templates and limited flexibility.
4. Operational technology and machine connectivity in manufacturing are often fragmented, especially in brownfield environments. Production equipment is rarely fully integrated, and inconsistent machine data limits visibility into cycle times or process parameters. Quality systems are siloed, making end-to-end traceability and automated documentation difficult. Sensor and IoT coverage is often incomplete, and real-time data transmission rarely in place. Furthermore, logistics data integration is minimal, leaving gaps in material movement tracking and predictive analytics for inventory optimization. These shortcomings create a serious bottleneck to achieving seamless connectivity and AI-driven efficiency.
5. Core IT capabilities are often geared toward transactional stability, not analytics and AI. Even when business cases are strong, IT lacks the tools and skills to operate AI models like any other critical service – with monitoring, versioning, rollback, and standardized deployment patterns.
6. AI in manufacturing is not just a technical line upgrade but a cultural shift within the organization. Many firms lack a clear AI strategy, defined roles, and sufficient budgets. Workforce upskilling and change management remain major gaps.
7. Cost optimization in brownfield plants faces structural challenges. Retrofitting legacy systems often drives high upfront costs and consumes scarce engineering capacity. These investments compete for capital with pressing operational priorities – such as new tooling for OEM launches, mandatory safety upgrades, and overdue equipment replacements. Adding sensors and connectivity to older equipment is expensive and requires significant capital investment. The expense is hard to justify without a clear business case that demonstrates measurable production benefits – fewer line stops, lower scrap rates on critical parts, faster ramp-up for new launches, or reduced warranty risks. Without these outcomes, projects are often viewed as discretionary “IT spend” and become easy targets for budget cuts.
The challenges are real – fragmented data, legacy systems, unclear strategy – but so are the opportunities. Many plants already have valuable data sources, experienced teams and proven technologies in place. Often, it is a matter of connecting the dots and aligning efforts. Our structured approach combines the Operations Maturity Assessment and AI Readiness Assessment. While the Operations Maturity Assessment identifies gaps and prioritizes high-value areas, the AI Readiness Assessment evaluates each plant and its processes across seven key capabilities using a 1-5 scale. Together, these frameworks provide transparency on both current maturity and future opportunities. The result is a clear path forward:
If you want to understand where you stand today, which levers will improve performance the fastest, and how AI can make that happen, let’s get in touch.
Copyright Berylls by AlixPartners