To an old-school automotive purist, driving is a conversation between driver and machine: the clutch pedal, the gear shift, the vibrations through the steering wheel, and the satisfying experience of the drive itself. To the busy, tech-savvy modern commuter, that experience is a chore someone finally automated: a vehicle that parks itself, predicts maintenance issues, and doubles up as a personalised digital cabin that updates and evolves.
Both camps have a point.
Love it or loathe it, we are in the era of intelligent vehicles, and it’s one of the most exciting yet disruptive engineering challenges in a century of automotive history.
Decoding the Tech vs Transformation Conundrum
A commercial aircraft runs on roughly 15 million lines of code and takes 7-10 years to design, test, and clear global safety audits. SDVs require over 100 million lines of code, yet market pressures demand that OEMs compress entire engineering lifecycles into just 2-3 years. Closing that gap demands a systems engineering mindset, one that leans on automation, virtualisation, and continuous validation to hold pace with market pressure.
Many treat the SDV shift as a software headcount problem: hire more developers, ship faster code releases, and layer more AI. That misses the point entirely. The reality is more nuanced, affecting organisational structures, operating models, supply chains, and how value is created long after the vehicle has been sold. This friction shows up in three areas.
People: Hardware engineers think in multi-year lifecycles focused on “right first time,” while software developers sprint in 2-week iterations, validating continuously and updating frequently. Compounding this cultural collision is a talent deficit, where the global demand for specialised automotive software engineers outstrips the supply.
Process: Legacy structures are unequipped for modern zonal architectures. The physical vehicle and its software stack must evolve independently. Buying advanced tools or AI models without rewriting the organisational operating model is a losing game.
Ecosystems: Building SDVs demands partners who understand the domain and digital infrastructure alike. It requires sharing data and co-developing the vehicle’s digital brain from day one. The multi-disciplinary expertise required now is simply too vast for any one manufacturer to absorb internally.
What Aircraft and Banks Figured Out First
Modern aircrafts rely on Integrated Modular Avionics, a zonal architecture that separates safety-critical systems from non-essential functions. For instance, if the in-flight entertainment crashes at 30,000 feet, the autopilot is unaffected. Automotive manufacturers are increasingly borrowing this philosophy as they transition to centralised electrical and electronic architectures.
Banking learned the hard way that every new digital touchpoint is a potential vulnerability, and the cost of breach is trust. It is the same story with intelligent vehicles. They contain dozens of wireless interfaces and as they begin handling payments for EV charging, smart parking and tolling, trust is as important as functionality, requiring BFSI-grade encryption and secure data pipelines. This level of cross-pollination is unlike anything seen in automotive history.
If AI Changes the Pace, Assurance Sets the Guardrails
Now, for the part that is genuinely exciting. AI is compressing SDV development timelines in ways that would have seemed implausible five years ago. General Motors, with physics-aware AI and advanced simulation, compressed vehicle timelines from 4-5 years to roughly 2; while Renault cut transmission development time by nearly 50% using AI to test design iterations virtually before building prototypes.
At Expleo, we developed AI-driven test engineering frameworks for leading ADAS providers that significantly improved test coverage, reducing validation effort and timelines by 60%. Tools that surface software and memory management issues much earlier in the lifecycle, reduce development times by up to 30%. We also built a portable cybersecurity assessment platform to help OEMs integrate cybersecurity by design.
The benefits are undeniable. However, AI models make highly sophisticated, best-guess predictions. But automotive engineering requires guaranteed outcomes for steering, braking, and powertrain. This is where AI assurance wraps probabilistic AI models in deterministic safety cages to ensure vehicles behave safely, predictably, and auditably under all operating conditions, throughout their lifecycle.
AI can accelerate testing, simulation, and decision-making, but accountability remains human. Engineers still need to validate outcomes, challenge assumptions, and determine whether a system is safe for the real world.
India’s Moment in the Driver’s Seat
You cannot talk about global automotive engineering without talking about India. With over 60 automotive GCCs and revenues projected to reach 9 billion by FY30, Indian engineers are writing the AUTOSAR code, training computer vision models, and building cloud architectures powering tomorrow’s global fleets. Domestically, cost-conscious consumers, unique infrastructure environments and training ADAS algorithms reliably across highly diverse road conditions means, adoption will follow a phased trajectory.
The future intelligent vehicles will be engineered like an aircraft, updated like a smartphone, protected like a banking platform, and powered by AI. Mechanical. Electrical. Software. AI. The boundaries between them are disappearing. Real-world problems do not arrive labelled that neatly anymore. Neither should engineering.
Unnikrishnan Anilkumar is COO at Expleo Solutions Limited. Views expressed are the author's personal.