SI Integration in the Automotive Industry and Preparing for 2027

Competition in the automotive industry does not end on the day a vehicle leaves the production line. Software that is updated throughout the vehicle’s service life, the data a battery generates over its life cycle, more accurate maintenance decisions and the information customers encounter before they buy together shape the value a brand delivers.

SI works at different points in this transformation. It detects a welding defect in a factory, helps the engineering team evaluate design options, supports the interpretation of the surroundings in driving systems or makes it easier for customers to find the vehicle that suits their needs. The data requirements, error tolerance and commercial value of these applications differ from one another.

As they prepare for 2027, the fundamental question facing automotive companies is broader than which SI tool to buy: Which decisions will they make better, which data will support those decisions and how will they prove the improvement?

From the Webtures perspective, sustainable advantage comes from the product, the operation and the customer experience working with the same verified information. That requires designing SI investments not as a single technology project but as interconnected business processes.

Executive summary

What is SI integration in the automotive industry?

SI integration in the automotive industry is the connection of systems that generate predictions from data, interpret image and sensor information, create content or carry out tasks with defined permissions to vehicles and automotive business processes.

Integration is broader work than deploying a chat interface. The model needs to access the right information, connect with existing systems, have its output evaluated and hand over to human intervention when necessary. A service assistant preparing a maintenance recommendation and that same recommendation becoming a work order are different levels of authority.

SI in the automotive industry can be assessed through five main forms of use:

A digital twin, an industrial robot or an over-the-air update system does not on its own mean that SI is being used. SI can be added to these structures for a specific function; the value each function provides must be measured separately.

What has changed in the industry as of 2026?

To assess change in the industry, it is more revealing to look at data with a clearly defined scope before broad market forecasts. Electric vehicle sales, factory robot installations and the number of driverless rides point to different developments; none of them on its own is “the size of the automotive SI market”.

The sources for these indicators are IEA Global EV Outlook 2026, Waymo’s September 2026 announcement, BMW’s pilot assessment and Volkswagen Group’s AI applications page. The consumer research row is based on a 2026 global automotive consumer study.

The development the data points to is that SI has begun to be used at different points in the automotive value chain. Companies’ priority should go beyond increasing the number of applications, towards reusable data infrastructure, measurable results and sustainable operation.

Why do software-defined vehicles matter?

A software-defined vehicle is a vehicle approach in which a significant share of functions is managed by software and suitable features can be updated throughout the vehicle’s service life. Also referred to as an SDV, this structure requires hardware, electronic architecture, software development and service delivery to be designed together.

Functions that were distributed across many separate control units in traditional architectures can be organised in new platforms around more centralised computing resources and zonal control units. The architecture BMW announced for the Neue Klasse, in which four high-performance computers take on different groups of functions, is one concrete example of this approach. This is an architecture description; the hardware of every model in every market should not be assumed to be the same. BMW’s architecture announcement makes the role of software in product development visible.

OTA updates are changing product management

An over-the-air software update is a means not only of delivering new features but also of fixing bugs and maintaining security. However, the ability to update does not mean that every update is automatically appropriate.

Companies must know which vehicles have which combination of hardware and software; they must verify an update’s compatibility, monitor staged rollout and apply a safe rollback plan when necessary. A change that affects driving safety cannot be managed like a new feature in the customer interface.

When software generates value after the sale, it also generates costs across the life cycle. Cloud services, connectivity, security maintenance, model evaluation and customer support must be weighed against subscription revenue. A customer using a feature does not mean that they will pay separately for it.

Where are we with autonomous driving?

The most important distinction in discussions of autonomous driving is which task the system takes on under which conditions. “Hands-off”, “eyes-off” and “driverless” do not describe the same capability.

Automation levels are a technical classification. Terms used in marketing language, such as “L2+”, are not official new levels added to this classification.

At the root of these distinctions is the operational design domain, or ODD. The ODD is the combination of geography, road type, weather, speed and other operating conditions. A driverless service operating in a particular city should not be treated as a product that can work in the same way in every country and in every weather condition. The technical level and legal liability in a crash are also separate matters. NHTSA’s explanation of automation levels supports this conceptual distinction.

Three examples, three different limits of use

Waymo shows the scale of driverless commercial transport in specific service areas. Weekly paid ride data is a strong indicator of use; however, the launch plans the company has announced for Europe do not mean that commercial operations began across Europe on the same date.

Mercedes-Benz DRIVE PILOT offers a different example, with the approval obtained in December 2024 for Level 3 use at up to 95 km/h under certain conditions in Germany. This example shows that conditional automation is offered within defined limits. The approval cannot be generalised to mean that the same feature is available in every country and every model. Mercedes-Benz’s announcement requires the driver to take over when necessary.

Tesla’s consumer FSD (Supervised), according to the manufacturer’s current support page, requires active driver supervision and does not make the vehicle autonomous. Feature availability depends on hardware, software, vehicle and region. This product should be treated separately from robotaxi services that are operated separately. Tesla’s product description shows why inferring capability from a name can be misleading.

In preparing for 2027, marketing and sales teams must understand these distinctions as well as product teams. A driving feature that is described incorrectly on a website or in an SI answer directly affects customer expectations and trust.

How is generative SI changing vehicle design and R&D?

Generative SI can be used to generate design options, review technical requirements, develop software and diversify test scenarios. Helping engineers evaluate more options does not remove the need for physical validation.

Proposing a part’s geometry and verifying its manufacturability, strength and safety requirements are different stages. Similarly, a driving behaviour that succeeds in simulation is not considered sufficiently validated in the real world.

The value of digital twins depends on data connectivity

A three-dimensional model of a factory or vehicle is not, on its own, a functioning digital twin. The model must be linked to the relevant physical process, measurements and current version information. If changes to equipment on the production line are not reflected in the model, optimisation recommendations may rest on outdated conditions.

Digital twins are valuable for evaluating line layouts, examining maintenance scenarios and testing the impact of changes in advance. The validity of the results depends on knowing which conditions the simulation represents.

World models and VLA approaches

VLA approaches, which combine vision, language and action, and world models, which aim to model how the environment evolves, are important topics in physical SI research. They offer new possibilities for evaluating rarely encountered scenarios and enriching development data.

The Alpamayo family that NVIDIA announced in January 2026 is an example. What matters in the announcement is that these models are positioned not as a ready-made driving system run directly in the vehicle but as teacher models that developers can adapt and use to transfer knowledge to smaller models. NVIDIA’s announcement clearly distinguishes between a development tool and a product approved for production.

What companies need to assess in this area is not only the complexity of the model. Test coverage, data provenance, failure scenarios, compute cost and compatibility with the existing safety architecture must be examined together.

Quality, maintenance and robotics in smart manufacturing

Areas where SI can be applied in automotive factories include visual quality control, detection of process deviations, predictive maintenance, production planning and material flow management. It cannot be assumed that the same method will deliver the same result in different plants. Lighting, product variety, sensor setup and workflow can change performance.

The World Robotics results that the IFR published on 24 September 2026 show that more than 600,000 new industrial robots were installed worldwide in 2025 and that the total number of operating robots reached around 5 million. These figures cover industrial robots across all sectors; they are not the number of robots using SI or of humanoid robots. The IFR’s announcement can be used to understand the scale of automation.

A single accuracy rate is not enough in visual quality control

Even if a model’s overall accuracy is high, it may still miss critical defects. The detection rate by defect type, false rejects, defective products passing to the next stage and the manual inspection workload should therefore be measured together.

For example, a cosmetic defect on a painted surface and a safety-critical joining defect should not have the same acceptance threshold. In datasets where defective samples are rare, the overall accuracy rate can be particularly misleading.

Predictive maintenance needs a decision chain

Calculating failure risk from sensor data does not show that maintenance activity has automatically improved. The alert has to become a work order, the required part has to be available and the intervention has to take place at the right time.

A successful setup links signals such as vibration, temperature or electricity consumption to historical maintenance records. The alert generated by the model is recorded together with the maintenance specialist’s assessment. When a failure does not occur, whether this was correct prevention or unnecessary maintenance is examined separately.

Assessing humanoid robots task by task

According to BMW’s February 2026 announcement, Figure 02 moved more than 90,000 parts during an approximately ten-month pilot in 2025 and supported the production of more than 30,000 BMW X3s over roughly 1,250 operating hours. This result relates to a specific task; it does not show that the entire factory runs on humanoid robots.

In humanoid robot investments, task flexibility, cycle time, the need for intervention, safe operation and total cost should be assessed together. For some jobs, traditional industrial robots, fixed automation or solutions that work alongside people may be more suitable.

The number of applications across the Volkswagen group carries a similar lesson: Seeing how widespread use has become matters; for economic value, you need to know which process each application improves.

SI in the supply chain and logistics

The automotive supply chain is complex because of multi-tier supplier relationships, long parts lists, changing demand and differing lead times. SI can support decisions by combining signals on orders, inventory, capacity, quality and external risks.

However, when forecasts are fed with incorrect part codes or outdated lead times, they can reduce decision quality. The first step is as much about clarifying what the same data means at different companies as it is about bringing all data into one place.

The priority use cases are:

• Forecasting demand by model and trim.
• Assessing single-source dependency and the impact of delays for critical parts.
• Linking spare parts to vehicle age, usage and regional needs.
• Connecting supplier quality signals to product and batch records.
• Comparing the cost of alternative production and shipping scenarios.

Fleet economics in autonomous freight

In its announcement dated 23 September 2026, Aurora reported that it has covered more than 500,000 driverless miles since its commercial launch. The fleet sizes for the end of 2026 and for 2030 in the same announcement, however, are growth targets, not realised results. Aurora’s announcement presents current operations and future plans together, but with separate measures.

For transport companies, technology assessment should also include vehicle utilisation, empty kilometres, terminal operations, remote support, maintenance, insurance and route constraints. Driving that requires no human intervention does not mean that the entire logistics operation runs without people.

Electric vehicles, battery health and energy management

The IEA’s 2026 report expects electric car sales to reach around 23 million by the end of the year, a 28% share of total sales. These values are 2026 estimates. They should not be presented with the same status as the more than 20 million sales realised in 2025. The IEA’s executive summary makes this distinction.

The growing user base of electric vehicles is making battery health, charge planning, energy cost and used-vehicle valuation more important. SI can provide decision support in these areas; however, the conditions under which the model output was generated must be known.

State of charge relates to SoC, which expresses the energy currently available in the battery. Battery health is addressed through the SoH approach, which assesses the state of capacity and performance over time. These two indicators should not be used interchangeably.

Battery health estimates are sensitive to chemistry, temperature, charging behaviour, usage history and measurement method. Giving a definitive remaining life or used-vehicle value for a battery with an incomplete history can be misleading. Data coverage and uncertainty should be presented along with the result.

In fleet management, charging optimisation can consider vehicles’ duty hours, station capacity, energy tariffs and battery constraints together. Simply choosing the cheapest hour can increase total cost if it leaves the vehicle not ready for operation.

The sustainability impact must also be measured separately. Gains such as lower energy consumption, a lower scrap rate or longer battery life should be separated from changes in production volume and product mix. The energy needs of the computing infrastructure that runs the model must also be taken into account; SI use on its own should not be turned into a claim of a lower carbon footprint.

Battery passport readiness for 18 February 2027

Under the EU Battery Regulation, an electronic battery passport is provided for from 18 February 2027 for electric vehicle batteries, light means of transport batteries and industrial batteries above 2 kWh that are placed on the EU market or put into service. This scope should not be interpreted as a single obligation applied retroactively to all existing vehicles worldwide. Articles 77 and 78 of the Regulation should be taken as the basis.

For global automotive companies, readiness requires the battery identity, the relevant composition and sustainability information and life cycle records to be fed from reliable sources. The fact that not all data is public and that different access rights exist must also be reflected in the design.

The battery passport does not make the use of SI mandatory. SI can help detect inconsistent records or missing fields; it should not fill gaps by generating an unsourced carbon footprint, composition information or health measurement.

In-vehicle SI assistants

In-vehicle assistants are evolving from fixed voice commands towards conversations that follow context. Navigation, the owner’s manual, vehicle settings and trip planning can come together in the same interface.

On 30 April 2026, Google announced the rollout of Gemini to eligible vehicles with Google built-in. The initial scope was announced as English-language use in the US. Answers based on the vehicle manufacturer’s manual and vehicle-connected functions vary by brand and model. Google’s announcement and the current support conditions show that availability must be assessed together with the vehicle, language, region and connectivity plan.

The critical value here is not just fluent conversation. The assistant needs to access information for the correct version of the user’s own vehicle. Explaining the warning light of another model or a feature of a different trim can produce an answer that is persuasive but wrong.

In-vehicle design must answer the following questions: Which functions continue when there is no internet connection? How does the assistant behave when it is not sure? What authority does it have to change vehicle settings? When does it interrupt the driver’s attention unnecessarily? How is personal information managed after a rental or the sale of the vehicle?

Conversational ability is not driving control ability. Clear task and access boundaries must be set between the cockpit assistant and safety-critical driving systems.

Agentic SI in sales, service and customer experience

Customer experience in the automotive industry consists of vehicle discovery, comparison, quotes, test drives, financing, delivery, maintenance and used-vehicle processes. SI agents can take on information-gathering and coordination tasks along this journey.

When a customer says, “I will drive 400 kilometres a week, I have no home charging and I need three child seats”, a good system should not simply return a list of vehicles. It should understand the conditions of use, make a comparison based on verified features and clarify any points that are unclear.

Moving from providing information to taking action

An agent showing a price and approving that price, or suggesting a service slot and creating the appointment, are different stages. Transaction permissions should therefore be set up in stages.

An example service flow can be designed as follows: the customer request is received, vehicle and service information is queried from the authorised system, suitable times are offered, the scope and any fee are explained, and the appointment is created after the customer approves. If the transaction fails, an alternative channel is shown; repeating the same request does not lead to a double booking.

Transactions such as finance applications, payments, contracts and sharing personal data require separate authorisation and approval. The fact that “the agent can do it” does not mean that every transaction should be executed automatically.

The economic value of customer trust

A 2026 global automotive consumer study shows the importance of safety and security expectations for connected features, and that reservations about data sharing persist. The study’s fieldwork was carried out in October and November 2025. The results help to understand consumer expectations; they do not show that the same behaviour will emerge in every market.

In the service experience, accurate information and explainable costs can be more decisive than the visibility of the technology. An assistant that answers quickly but incorrectly can lengthen resolution time. Alongside the automation rate, first contact resolution, repeat contacts and customer complaints should therefore also be monitored.

SI visibility for automotive brands

Customers can also research vehicle comparisons through SI assistants. In this environment, it matters as much that a brand’s vehicles are described with the correct features and in the correct context of use as that the brand’s name is mentioned.

A model may come with different engine, battery, trim and warranty options in different countries. When SI systems mix up this information, consumers may believe that a feature not offered in their own market is available. Visibility work in the automotive industry must therefore be carried out together with work on the accuracy of product information.

Machine-readable product information

Model pages and connected data sources should consistently carry at least the following distinctions:

• Brand, model, model year, trim and sales market.
• Engine or battery option, the relevant measurement standard and feature scope.
• The currency of the current price, its validity date and the conditions it includes.
• Stock, delivery, warranty, maintenance and return or cancellation terms.
• The software, hardware and usage limits of driver assistance features.
• The source of the information, its last update and the team responsible for the content.

This structure does not concern the website alone. Consistency is needed between the product information system, the dealer system, customer communications and the information on external platforms. Using structured data can be useful; however, it does not guarantee inclusion or recommendation in a particular SI answer.

How should visibility be measured?

Measurement should start with a repeatable set of questions. Each question should clearly define the market, language, budget, usage need and product category. The platform, model or access method used, the measurement date and the response conditions should be recorded.

The brand being mentioned, being evaluated positively, being shown as a linked source and being recommended with accurate information are separate indicators. The result of a single test should not be turned into a general market ranking. Because of model updates and response variability, measurements should be repeated at regular intervals.

From the Webtures perspective, this work should connect visibility data to business decisions: Which model is being described incorrectly? For which usage need is the brand not being considered? Which information gap prevents customers from reaching the right vehicle or service?

The technical foundation for SI integration

In scalable automotive SI applications, the link between data, model and transaction must be traceable. Not every use case needs to share the same architecture; but common data definitions and control principles are required.

Data layer: Vehicle, part, battery, customer and transaction records are linked with the correct identifiers. Model year, software version and validity dates must not be lost. Unnecessary personal data should not be sent to the model.

Knowledge and model layer: A system that generates answers from technical documents must find the relevant source and use the correct version. A RAG approach can establish this link; however, on its own it does not prevent the wrong document from being retrieved or the model from misinterpreting it.

Transaction layer: Connections to CRM, service, inventory or order systems are set up with explicit permissions. Read access is separated from access to change records. Repeated requests must not produce duplicate transactions, and safe recovery is ensured in the event of a timeout.

Evaluation layer: The system is tested with representative examples before real-world use. When the model, prompt, data source or integration changes, critical scenarios are re-evaluated.

Operations layer: Errors, latency, cost, user corrections and human intervention are monitored. Who is responsible for the system, in which situation it will be stopped and how to return to the previous version are defined.

For models running on the vehicle, latency, connectivity and hardware limits come to the fore. Cloud-based applications require a different balance of capacity, data transfer and service cost. Processing data in the vehicle is not, on its own, proof of security or regulatory compliance.

Safety, cyber resilience and compliance

SI governance in the automotive industry should be set up according to the use case. A tool that generates marketing content and a system that influences a braking decision do not have the same error tolerance.

In safety-critical applications, data and software versions, test scenarios, operating limits and update processes must be managed together. In commercial assistants, accurate product information, personal data protection, transaction permissions and customer consent take priority.

Cybersecurity and software updates

The UNECE R155 and R156 frameworks are important for cybersecurity and software update management in vehicle approval processes in the markets that apply them. R155 focuses on cyber risk management and R156 on the management of software updates. They are not a single global regulation that will start for the first time in 2027. UNECE’s announcement sets out the relationship between the two areas.

A change to an SI model must also be assessed according to the function it affects. When deploying a version that changes a safety function, testing, traceability and the relevant approval impact must be examined separately.

Distinguishing use cases under the European AI Act

Taking into account the regulatory changes in 2026, it is not correct to tie all SI systems in the automotive industry to the same high-risk timeline. For vehicle safety components, the relationship between the AI Act and vehicle type approval and sectoral regulations must be addressed together. A dealer assistant, a system that selects employees and a driving safety component cannot be managed with the same compliance file. The current amending text should be taken as the basis for this distinction.

Global companies should build an application inventory based on intended purpose, target market, data type, provider and their own role. A conformity assessment in a single country does not automatically ensure conformity in all other markets.

Usability is also part of safety performance

Euro NCAP’s 2026 assessment approach places importance on areas such as driver monitoring, the understandability of assistance systems and the usability of basic controls. This is a consumer safety assessment programme; it does not replace legal type approval. The 2026 protocol changes are a reminder that more functions do not always create a better experience.

The driver understanding the system’s limits, not facing unnecessary warnings and being able to take control at the right moment are fundamental to product design.

How is the success of SI investments measured?

Measuring success starts with model performance and is completed with operational and economic results. A system that produces more accurate predictions may not create value if it is not connected to the decision process.

Example investment calculation

The scenario below has been prepared solely to show the calculation method; it is not an industry average or a Webtures client result.

Assume that a service operation handles 60,000 eligible requests a year and that SI support can save an average of 4 minutes per request. This corresponds to 4,000 hours of capacity a year. If the fully loaded labour cost per hour is taken as 30 US dollars, the theoretical capacity value is $120,000.

If only 60% of this capacity turns into completing more work or a real cost reduction, the usable annual value is $72,000. If the annual operating cost is $30,000 and the initial setup cost is $50,000, the net annual benefit before the setup cost is $42,000. Under the simplified assumption that full capacity is reached from day one, the payback period is approximately 14.3 months.

In a real calculation, the ramp-up period, human oversight, training, integration maintenance, erroneous transactions and changes in demand must be added. Not every hour saved is a direct cash saving. Capacity gains, cost reduction and additional revenue should be tracked separately.

From token cost to the cost of a successful transaction

In enterprise SI, tracking only the model’s token price is not enough. Data preparation, search, integration calls, evaluation, retries and human intervention are all part of the total cost.

For this reason, alongside the question “How much does a million tokens cost?”, managers should also ask “What is the total cost of a correctly completed service request?” If a cheaper model requires more corrections, the process can become more expensive overall.

How should employees and the organisation prepare?

The adoption of SI applications depends on employees knowing in which decisions they can use the system. A quality specialist questioning a model result, a service employee flagging an incorrect answer and the product team correcting a wrong feature are all part of operating the system.

Training should not only teach how to write prompts. It should also cover checking sources, interpreting uncertainty, understanding transaction permissions and passing a problem on to the right person. When the corrections made by early users are recorded, training and product development can benefit from the same feedback.

The business owner should take responsibility for the expected outcome, the data owner for the accuracy of the information and the technical team for operating the system. Security, legal and relevant engineering specialists should join the process at the stages the use case requires. If these responsibilities remain unclear during the pilot, the problem grows as the application grows.

How capacity gains will be used, whether for better service, new tasks or cost reduction, should be planned separately. Seeing time savings in a task does not lead to the conclusion that a workforce reduction of the same proportion is possible or appropriate.

Roadmap for 2027 readiness

Preparation should not start with the goal of transforming all processes at once. A limited number of priorities should be selected by assessing business value, data readiness, integration difficulty and error impact together.

Work that can be carried out in the first 90 days

First 30 days: A business problem is selected and the current result is measured. Users, data owners, transaction permissions and exceptions are identified. Errors that will not be accepted are defined as well as the success criterion. The expertise required for sensitive customer data or safety-critical functions is brought into the process from the start.

Days 31 to 60: A prototype is prepared with representative data. Alongside normal flows, missing data, incorrect requests, outdated documents, system outages and unauthorised transaction attempts are tested. The reasons why human experts correct the results are recorded.

Days 61 to 90: The application is opened to a limited group of users. Evaluation is carried out against a control group or a comparable period; the effects of demand, product mix and seasonality are taken into account. Based on the results, a decision is made to scale, redesign or stop.

This 90-day framework is intended for enterprise applications with a limited scope. It cannot be used as a general delivery time for developing a safety-critical driving system or for an approval programme.

Which developments could stand out in 2027?

Rather than reading the future through a single date or market forecast, it is sounder to build readiness scenarios from today’s developments. The headings below are an editorial assessment by Webtures; they are not confirmed industry results.

The commercial importance of vehicle data may increase. Managing hardware, software, battery and service information together can create value in more processes, from the sale through to the used-vehicle market. Data standardisation may become a higher priority than developing a new assistant.

SI agents may become widespread in limited but completable tasks. Creating an appointment, retrieving verified product information or preparing an approved work order are more measurable starting points than systems that do everything with unclear permissions.

Task variety may increase in physical SI. This need not mean that factories will become fully autonomous. Safety, cycle time, maintenance and economic viability must be demonstrated again for each new task.

Autonomous freight may expand within regional and operational limits. The factors that determine scale will include permits, fleet services, insurance, maintenance and customer experience as well as technology.

SI investments may face more detailed economic scrutiny. Rather than the number of applications, management teams may demand indicators of successful transactions, reliable decisions, capacity used and sustainable profitability.

Readiness check for automotive companies

To assess a company’s level of readiness, the following questions can be answered before the technology inventory:

• Are the business problem we want to solve and our current performance clear?
• Does each application have a clear data owner, business owner and technical lead?
• Can we correctly match vehicles, parts, model years, markets and software versions?
• Are the data the model can access and the actions it can take restricted?
• Have missing information, system outages and misdirection scenarios been tested?
• Are we measuring critical errors and human intervention as well as success?
• Do employees know the system’s limits and the escalation path?
• Can we re-evaluate quality when the model or supplier changes?
• Are update, monitoring, security and maintenance costs included in the budget?
• Has it been decided in advance which result will lead to the pilot being scaled or stopped?

The answers to these questions make priorities visible without reducing SI maturity to a single score. A missing security or authorisation condition should not be considered offset by high performance in other areas.

Frequently asked questions

Where is SI used in the automotive industry?

It can be used in vehicle design, software development, driver assistance systems, quality control, predictive maintenance, supply planning, battery analytics, sales and customer service. The value of an application depends on its connection to the existing business process and on its verified results.

Is every vehicle that uses SI autonomous?

No. An in-vehicle voice assistant, energy forecasting or a driver assistance function may use SI. Their presence does not show that the vehicle can operate without a driver. The level of automation and the conditions of use must be assessed separately.

Is a software-defined vehicle the same as an autonomous vehicle?

No. A software-defined vehicle is an architecture and product approach in which functions are managed through software. Autonomous driving is the system taking over the driving task under certain conditions. A software-defined vehicle may still need driver supervision.

Can SI completely eliminate production defects?

No such general guarantee can be given. Detection performance depends on the product, sensors, data and operating conditions. SI can support quality control; it should be used together with process design, validation and appropriate human oversight.

What does agentic SI mean in the automotive industry?

It refers to SI applications that gather information towards a specific goal, interact with systems and carry out transactions with defined permissions. Creating a service appointment is one example. Which transactions require approval and what happens in the event of an error must be defined in advance.

How should automotive companies choose their first SI project?

They should choose a process whose business value can be measured, whose data is accessible, whose scope is limited and whose error impact is manageable. Technical document search or service request classification may be suitable candidates; the same starting point does not apply to every company.

Is a digital passport mandatory for all batteries in 2027?

No. The EU regulation has a scope defined by market, battery type and placing on the market or putting into service. The relevant start date is 18 February 2027. It should not be interpreted as if it applied with the same scope to all existing batteries or to products in every country.

How can automotive brands improve their SI visibility?

They can start by keeping product information accurate and current, clarifying model and market distinctions, publishing content supported by reliable sources and measuring how they are described in representative questions. Incorrect information and incomplete representation should be monitored as well as visibility.

The Webtures approach

SI strategy in the automotive industry should connect the company’s technical capabilities with the customer’s decision process. The accuracy of product information, how the brand is discovered, which questions customers find answers to and how these interactions are measured must be addressed together.

The Webtures view of this transformation takes concrete form in three areas of work: examining how the brand is represented in SI environments, making product and service information consistent across digital channels and measuring the commercial impact of new customer journeys.

SI visibility and Visibility Intelligence, Agentic Commerce readiness and SI & Agentic Analytics offer complementary frameworks for assessing these areas.

When preparing for 2027, a valuable starting point is to identify where the company will make its most important decision more reliable, before choosing where it will use the most technology. Let us assess your automotive brand’s visibility, data and customer experience priorities together.

Get in touch with Webtures →