2026-04-20

Table of content

1. The Shift Begins
2. Work Redefined
3. Declining roles
4. Evolution, not extinction
5. Shift in skills
6. Conversion Path
7. Learning the AI Way
8. Where Companies Get AI Wrong
9. The Role of AI Tools – ENGAI
10. This Is Not the End of Your Career
11. Advice for Engineers
12. What Comes Next

The Shift Begins

Just a few years ago, automotive software development was limited by the number of engineers available. Today, that constraint is rapidly disappearing. Code can be generated, unit tests can be written and executed automatically, and even requirements can be partially created or refined with the support of AI. As a result, development cycles are accelerating at a pace that would have been impossible a decade ago.

But this transformation comes with a structural consequence: the demand for engineers is changing – and in some areas, decreasing. Roles focused on repetitive, execution-heavy tasks are becoming less critical at scale. Positions such as classic AUTOSAR developers, system integration testers, or requirements engineers are no longer needed in the same volume as before. At the same time, we are seeing a clear rise in AI-related roles – from ML engineers to AI validation and integration specialists.

This trend is already reflected in the data. According to Forbes, AI-related job postings have grown by approximately 38% in recent years, while roles such as “AI Engineer” now rank among the fastest-growing jobs globally, consistently taking top positions in labor market rankings (check out the full ranking here: These are the 10 fastest-growing jobs in the US, LinkedIn says). At the same time, traditional engineering roles are no longer scaling at the same pace.

The implication is clear: this is not just a technological shift, but a structural one. AI is not simply changing how engineers work – it is redefining which roles are needed, and in what proportion.

Work Redefined

When asked about the impact of AI on engineering teams, he doesn’t hesitate:

The rapid development of AI means that engineering teams are now working at a significantly higher pace. Many tasks – from prototyping to code analysis – can be automated or accelerated using AI. Developers are increasingly taking on the role of “integrators”, combining code generated by AI with traditionally written code.

Teams must also learn how to evaluate the quality of outputs generated by models, because AI can accelerate work, but at any moment it may introduce subtle errors. The importance of integrating AI tools into CI/CD pipelines and DevOps processes is growing.

At the same time, there is increasing pressure on documentation and standards, because AI performs best in structured and well-organized environments. As a result, teams are becoming more interdisciplinary and more focused on architecture, data quality, and automation

In practice, this shift is already visible across everyday engineering activities. AI is significantly accelerating code generation, unit testing, documentation, and error analysis, fundamentally changing how work is distributed within teams. The engineer is no longer just a creator of code – but increasingly an integrator and validator of AI-generated outputs. This transformation is also reflected at the product level. Around 20 years ago, a typical OEM would release 4–6 models, each with a few body variants. Today, it’s not uncommon to see 20 or more models, each with multiple configurations – all supported by increasingly complex software systems. Without AI-driven acceleration, this scale simply wouldn’t be feasible.

This transformation is not only visible in the way engineers work – it is deeply embedded in the structure of modern vehicles themselves.

^{source:www.linkedin.com}

As illustrated above, modern vehicles are no longer built around isolated electronic control units, but around layered software architectures, spanning from hardware abstraction and embedded control units to cloud services and connected applications. This shift dramatically increases system complexity, the number of integrations, and the volume of data flowing through the system.

In such an environment, manual, repetitive engineering work simply does not scale. AI becomes not an enhancement, but a necessity – enabling engineers to manage complexity, automate low-level tasks, and focus on system-level decisions.

AI does not eliminate work. It eliminates repetitive work – and reduces the demand for junior-level, execution-heavy tasks.

Declining roles

This is where the shift becomes more uncomfortable – but also more real.

There are growing indications that demand for some traditional engineering roles is decreasing. Positions such as AUTOSAR Developers,QA/Automation Testers, Requirements and Validation Engineers, and parts of classic Embedded Software Engineering roles are no longer needed at the same scale as before.

The reason is structural. Code can now be partially generated, tests increasingly automated, and requirements supported or refined by AI tools. Tasks that were once time-consuming and resource-heavy are becoming faster and less dependent on large teams.

This trend is also reflected in broader market data. According to recent industry analyses (based on: It’s a tough time to be a tech graduate — AI and layoffs have made it a competitive job market):

• employment of software developers has been declining since around 2020, with a particularly sharp drop observed in early 2022, followed by continued volatility and a downward trend
• according to data from the LinkedIn application, competition per role is increasing, with more candidates applying for each position in the tech sector
• computer science graduates are facing higher-than-expected unemployment rates, in some cases exceeding those of non-technical fields
• since the first wave of layoffs in 2022, a longer-term trend has taken shape through 2024, and the market has not returned to its previous hiring dynamic

However, this does not mean these roles are disappearing. They are evolving. The engineer is no longer just executing tasks – but increasingly working at a higher level of abstraction, closer to systems, architecture, and AI-supported decision-making.

Evolution, not extinction

When asked whether roles like Embedded Software Engineer, Systems Engineer, or Software Test Engineer might disappear, Sławomir is very clear:

These roles will not disappear, but they will change very significantly over the next 3–5 years.

An Embedded Software Engineer will increasingly work with models running locally (SLMs, edge AI), which requires new competencies in optimization and integration. The Systems Engineer will likely become the most critical role. On one hand, AI increases system complexity and the need for coherent architecture; on the other, it automates code generation based on requirements. The difference between Systems Engineer and Software Engineer will increasingly blur, eventually merging over time as AI models improve. The Software Test Engineer will not disappear – quite the opposite. Testing will become even more demanding. Engineers will need to validate non-linear, probabilistic, and hard-to-predict behaviors. Many testing tasks will be automated, while the human role will shift toward scenario design, validation, and risk analysis.

The change will be significant – but it is evolution, not elimination. As a civilization, it would be a mistake to merge all three roles into one

This perspective reflects a broader industry pattern. The nature of engineering roles is shifting – not disappearing – and with it, a new map of roles is emerging.

Across the market, we are already seeing consistent growth in positions related to AI, data, and system-level integration. Based on hiring trends and job market data, the fastest-growing and most in-demand roles in automotive and mobility today include:

• AI / Machine Learning Engineer (especially in automotive and embedded contexts)
• Computer Vision Engineer (ADAS, perception, LiDAR, camera systems)
• MLOps / Edge AI Engineer (deployment and optimization of models on hardware)
• AI Validation / Verification Engineer (testing probabilistic systems and model behavior)
• Autonomous Systems Engineer (system-level design of self-driving functions)
• Data Engineer (vehicle data, fleet data, telemetry pipelines)
• AI Safety / SOTIF Specialist (increasingly relevant with ISO 21448 and AI risk)

These roles are not theoretical – they are reflected in current hiring demand. Positions such as AI Engineer, Machine Learning Specialist, and Data-related roles consistently rank among the fastest-growing jobs globally, according to LinkedIn workforce reports and labor market analyses.

At the same time, what remains clear is that deep embedded knowledge does not lose its value – it becomes a differentiator. Understanding real-time systems, hardware constraints, communication protocols, and safety requirements is exactly what allows engineers to successfully transition into AI-driven automotive roles.

Shift in skills

This is not a change of industry. It is a shift in the center of gravity. To better understand which competencies are gaining importance, we asked Sławomir:

One of the most underestimated skills today is the ability to precisely define problems – so-called problem framing – which is crucial when working with AI.

Solid foundations in mathematics and statistics are also becoming increasingly valuable, as they allow engineers to better understand the limitations of models.

Contrary to popular belief, classical systems engineering is gaining importance again: requirements analysis, interface design, and managing complexity.

Skills related to safety and reliability are also often underestimated, as AI introduces new risk vectors.

Soft skills are becoming more important as well – communication, working with uncertainty, and the ability to ask the right questions.

Finally: critical thinking, because AI generates confident answers, but not always correct ones

^{source:www.iot-analytics.com}

The shift described above is clearly reflected in market data. As shown in the chart from IoT Analytics, Generative AI is the fastest-growing skill cluster, while AI/ML combines both high demand and strong growth, making it one of the most strategically important areas. Cloud remains highly demanded but is entering a more mature phase, while several traditional engineering skills – especially those disconnected from data and AI – show slower growth or decline.

This confirms a key transition: the market is not eliminating skills – it is reorganizing them around AI, data, and system-level complexity.

What skills need to be built today?
LAYER	KEY COMPETENCIES	WHY IT MATTERS
AI & Data Foundations	Python, NumPy/Pandas, PyTorch/TensorFlow, model training, evaluation metrics, dataset bias	Understanding how AI works – and where it fails
Systems Engineering	Architecture, requirements analysis, interface design, system integration	AI increases complexity – systems thinking becomes critical
Automotive Context	Perception systems, sensor fusion, edge inference, latency optimization, SOTIF	Applying AI in real, safety-critical environments
Decision & Thinking Skills	Problem framing, critical thinking, interpreting AI outputs, asking the right questions	The key differentiator in AI-driven engineering
Human & Collaboration Skills	Communication, working with uncertainty, cross-functional alignment	Teams become more interdisciplinary

What stands out is that the most critical skills are no longer purely technical – but cognitive. What becomes clear is that AI in automotive is fundamentally different from general-purpose AI tools.

It is not about prompts and quick outputs. It is about safety, determinism, validation, and responsibility. And this is exactly where the biggest shift is happening.

In an AI-driven world, the real advantage is no longer writing code. It is asking the right questions.

Conversion Path

The most important realization is this: for most engineers, this is not a career change – it is a change in how they work.

The transition toward AI does not require abandoning existing experience. On the contrary, backgrounds in embedded systems, testing, or software engineering remain highly relevant, especially in automotive, where real-time constraints, hardware limitations, and safety requirements still apply. What is changing is not the role itself, but the level at which the engineer operates and the tools they use on a daily basis. In practice, this transition happens step by step and can be understood across three levels:

🔍 Level 1 – AI-aware Engineer

At this stage, the goal is not to build models, but to work effectively with them:
– using Python as a working language
– understanding basic machine learning concepts and model limitations
– working with datasets (cleaning, labeling, interpreting outputs)
– using AI tools to support coding, testing, and documentation
– understanding where AI can fail (bias, hallucinations, edge cases)

⚙️ Level 2 – AI-integrated Engineer

Here, AI becomes part of the system – not just a tool:
– deploying models on embedded or edge systems
– optimizing performance (latency, memory, hardware constraints)
– working with platforms such as NVIDIA, Qualcomm, or similar accelerators
– validating model behavior in real-world and safety-critical scenarios
– integrating AI into existing architectures (ECU, middleware, pipelines)
– working with CI/CD pipelines including AI components

🧠 Level 3 – AI-specialized Engineer

This path is not for everyone – and it doesn’t have to be:
– training and tuning models
– designing neural network architectures
– building data pipelines and MLOps workflows
– working on AI safety, robustness, and explainability
– optimizing models for production environments (edge/cloud balance)

^{source: created from author’s notes}

What makes this transition realistic is that it can be incremental. Engineers do not need to “jump” into AI – they can grow into it. Today, we already see clear and practical paths: embedded engineers moving toward edge AI, testers evolving into AI validation roles focused on scenario design and risk analysis, and software engineers working increasingly in AI-assisted development environments.

In many cases, reaching a solid, AI-integrated level is possible within 6–12 months of focused learning and hands-on practice.

Learning the AI Way

If there is one practical question every engineer is asking today, it’s this: where do I actually start?

The good news is that transitioning toward AI does not require going back to university or becoming a data scientist. What matters is not the number of courses completed, but building the right combination of skills and applying them in practice. In reality, most successful transitions follow a simple pattern: learn → apply → integrate. Below is a curated set of resources that reflect this approach – not as a checklist, but as a toolbox for building real, applicable competence.

^{source: created from author’s notes}

What is often misunderstood is the scale of this shift. Entering the world of AI does not mean abandoning engineering – it means extending it in the right direction. In practice, most engineers do not need a PhD or a full ML stack. What they actually need is Python, ML basics, and one real use case – such as computer vision, edge AI, or data-driven engineering. This focused approach allows engineers to start delivering value quickly, without getting lost in unnecessary complexity.

Where to start – curated learning paths

🧠 Foundations (AI & Machine Learning)

• Machine Learning Specialization (Coursera)
  Structured, beginner-friendly, great for understanding how models actually work.
• Practical Deep Learning for Coders (fast.ai)
  Very hands-on, less theory, faster results – ideal if you want to build something quickly.
• Machine Learning Crash Course (Google)
  Quick and practical, perfect as a low-commitment entry point.
• 📖 Hands-On Machine Learning with Scikit-Learn & TensorFlow
  Highly practical book, very engineering-oriented, with real code examples.

🚗 Automotive AI & Perception

• Self Driving Car Engineer Nanodegree (Udacity)
  Close to real automotive use cases – perception, sensor fusion, ADAS.
• Computer Vision Full Course (freeCodeCamp)
  Full end-to-end introduction, great free resource with lots of practical examples.

⚙️ Edge AI & Deployment

• NVIDIA Deep Learning Institute
  Very practical and hardware-focused, ideal for embedded engineers.
• Get Started with NVIDIA Jetson
  A very good bridge between embedded engineering and AI deployment, with access to documentation, courses, and hands-on resources.
• TensorFlow Lite Guide
  Focus on lightweight, production-ready models.

🛡 Safety, Validation & Explainability

• ISO 21448 (SOTIF) overview
  Critical for automotive – AI must be safe, not just accurate.
• AI Functional Safety (Perforce eBook)
  Shows how to approach safety and risk when working with AI-based systems in automotive.

🧠 Systems & Engineering Thinking

• Introduction to Systems Thinking (MIT)
  Helps manage complex systems – increasingly critical in AI-driven environments.
• Machine Learning in Production (Coursera)
  Focuses on deploying, validating, and monitoring AI systems in real production environments.

A realistic six-month roadmap shows that the transition does not happen overnight – but it also does not take years. In practice, many engineers follow a structured and achievable path like this, gradually building their skills and moving closer to their target role.

Period	Stage	Description
Months 1–2	Foundation	You build your foundation. Python, basic machine learning concepts, and working with data. The goal is not depth, but understanding and familiarity.
Months 3–4	Specialization	You begin to specialize. Most often in computer vision, perception, or data-related problems. This is where theory meets reality – through small projects, experiments, and hands-on work.
Months 5–6	Integration	You move toward integration. Deploying models, optimizing performance, or embedding AI into existing systems. At this stage, AI becomes part of your daily engineering workflow, not just something you study.

The key is not to do everything. The key is to move forward with intention. Because ultimately, the goal is not to collect courses or certificates. It is to become an engineer who can actually use AI – in real systems, under real constraints, and with real responsibility.

Where Companies Get AI Wrong

When asked where companies make the biggest mistakes when implementing AI in software development processes, Sławomir points to a pattern that is both common and underestimated:

The most common mistake is treating AI as a “magic accelerator” without analyzing the processes it is supposed to improve. Companies often implement tools without preparing data, standards, or documentation, which results in poorly conditioned models and chaotic outputs.

Another issue is the lack of responsibility – as users get used to AI suggesting solutions, they begin to trust it blindly and stop verifying results. Many organizations also ignore security and compliance, which can lead to data leaks or the introduction of vulnerabilities. There is often a lack of training, meaning employees use AI inefficiently or do not understand its limitations.

Finally, companies do not measure the results of AI implementation, so they cannot assess whether it actually delivers value

In practice, these issues usually come down to a few core gaps. Lack of data preparation. Lack of structured processes. Lack of ownership. Lack of measurable outcomes. AI does not fix these problems – it amplifies them. This is not just a theoretical observation. Industry data confirms the same pattern. According to Forbes analysis on AI model failures (see: Council Post: Why 85% Of Your AI Models May Fail), up to 85% of AI projects fail, most often due to poor data quality or lack of relevant data. At the same time, Gartner predicts GenAI project abandonment that at least 30% of generative AI initiatives will be abandoned after the proof-of-concept stage, primarily due to unclear business value, weak risk controls, and the inability to scale solutions into real production environments (check out the full article here:Gartner Predicts 30% of GenAI Projects Will Be Abandoned By 2025).

AI performs best in structured, well-organized environments. If the organization is chaotic, AI will not bring order – it will scale the chaos. Companies that approach AI successfully treat it not as a shortcut, but as a transformation:

– invest in reskilling instead of reducing headcount
– create structured AI transition programs
– combine embedded engineering and AI capabilities within teams
– build AI literacy across the organization
– implement tools that support real engineering workflows (e.g., solutions like ENGAI)

Companies that invest in competence transformation today will be the ones leading tomorrow.

The Role of AI Tools – ENGAI

This is exactly why the role of tools that structure engineering work is becoming increasingly important. If AI performs best in well-organized environments, then the focus shifts toward solutions that bring order to areas such as documentation, data analysis, logs, and testing – the very domains where engineering teams spend a significant amount of time on repetitive, low-value tasks.

AI is not meant to replace engineers. It is meant to remove the friction around their work.

In practice, this means automating activities like requirements analysis, documentation generation, log interpretation, test reporting, or change impact assessment – allowing engineers to focus on what actually creates value: system-level decisions, architecture, and validation.

At RSB Automotive Consulting, this is exactly the direction we are exploring with ENGAI (see: https://rsb-ac.com/products/engai/) – a solution designed to support engineers directly within their existing workflows, without disrupting the tools and environments they already use. ENGAI integrates with platforms such as requirements management systems, issue trackers, and documentation tools, enabling automation of tasks like traceability tracking, verification reporting, defect analysis, or compliance documentation. It can process large volumes of engineering data – from requirements and test results to logs and change requests – and transform them into structured, actionable outputs.

What makes this approach critical is not just automation, but context awareness – the ability to operate on real engineering data, within real project structures, while maintaining full control over sensitive information. This is especially important in automotive, where data security, compliance, and traceability are not optional, but fundamental.

The goal is not to replace engineering expertise, but to augment it, reducing manual effort and enabling teams to operate at a higher level of abstraction – where engineers focus less on generating artifacts, and more on understanding systems, validating outcomes, and making decisions.

Because in an AI-driven environment, the competitive advantage is no longer how much work a team can execute – but how effectively it can focus on the right problems.

This Is Not the End of Your Career

If there is one thing worth making clear, it’s this: AI does not mean the end of your career.

If you come from an embedded, systems, or software engineering background, you already have something extremely valuable – a strong foundation. You understand real systems, constraints, architecture, and complexity. These are exactly the competencies that AI-driven environments need. The shift is not about starting over. It is about evolving your profile.

^{source: created from author’s notes}

You do not need to learn everything at once. You do not need to become a data scientist overnight. What matters is consistency. One hour a day, over six months, is enough to create real change. Step by step, you move from understanding AI, to working with it, to integrating it into your daily engineering workflow.

AI is not your competition. It is a tool. The engineers who succeed will not be the ones who know everything – but the ones who adapt, stay curious, and build hybrid skill sets that combine engineering depth with AI awareness. Do not panic. You are not behind. But you do need to move.

The biggest risk is not AI. The biggest risk is staying still.

Advice for Engineers

As we’re slowly coming to the end, one thing becomes clear – this shift is already happening, whether we’re ready for it or not. And naturally, this leads to one question: what should you actually do as an engineer today? Instead of speculation, we want to leave you with something more grounded. Sławomir – someone who has been working in IT and engineering since the early 90s and has already gone through multiple technological transformations – has a clear, experience-based message for you:

The most important advice is not to be afraid of change – technology has always evolved, and those who stay curious and adaptable handle it best.

It is crucial to develop analytical skills and critical thinking, because they allow you to distinguish good solutions from bad ones. Strong foundations are also essential: knowledge of algorithms, system architecture, complexity, operating systems, and networks – these competencies remain relevant regardless of trends. At the same time, you need to learn how to work with AI, treating it as a tool rather than a threat.

It is also worth developing soft skills – communication, teamwork, and the ability to explain complex problems clearly.

And above all: stay curious and keep learning, because that is the only constant in this industry

His perspective reinforces a broader truth that runs throughout this transformation. The engineers who will thrive are not necessarily those who specialize the fastest – but those who build strong fundamentals, adapt continuously, and learn how to navigate change instead of resisting it.

What Comes Next

“It is not the strongest of the species that survives, nor the most intelligent that survives. It is the one that is most adaptable to change.” – Charles Darwin

This is exactly what defines the current transformation in automotive engineering. AI is not just accelerating development – it is reshaping roles, redefining value, and shifting the focus from execution to decision-making. Repetitive tasks are disappearing, while systems thinking, validation, and integrationare becoming the core of engineering work.

This is not the end of engineering. It is a change in its center of gravity. The direction is already clear: software-defined vehicles, data-driven development, and AI-enabled systems will continue to increase complexity – and with it, the demand for engineers who can adapt, think critically, and work with AI.

The change is happening.

The only question is: what will you do with it?

If this resonates with you – whether as an engineer or an organization – let’s continue the conversation. Because the decisions made today will define who leads tomorrow.

Sources:

These are the 10 fastest-growing jobs in the US, LinkedIn says

5 AI Jobs You Can Land With No Coding Experience Required | Forbes

How to deal with exponential complexity in automotive engineering

It’s a tough time to be a tech graduate — AI and layoffs have made it a competitive job market

What tech skills companies recruited for in Q1 2024? AI, Gen AI, and 5G

Council Post: Why 85% Of Your AI Models May Fail

Gartner Predicts 30% of GenAI Projects Will Be Abandoned By 2025