How AI innovation in EV manufacturing can be cost-effective despite 80% of projects failing

EVs are reshaping the automotive landscape, offering significant environmental and economic advantages, from reduced emissions to lower maintenance costs. The global transition towards EVs is accelerating, with nearly 14 million new electric cars registered in 2023 alone. This rapid growth, coupled with the increasing demand for sophisticated features, is driving the need for AI-powered solutions across the entire EV value chain. 

AI is poised to drive the next phase of innovation in the EV sector, providing solutions to the unique challenges faced by EV manufacturers. Only this week, the UK government highlighted the importance of this technology by unveiling the AI Opportunities Action Plan, a strategy integrating artificial intelligence to stimulate economic growth. Within the electric vehicle industry, AI presents significant opportunities to help manufacturers address common obstacles and drive economic growth. From improving battery management systems to enhancing range estimation, it’s reshaping how EVs are designed, built and experienced. The impact is significant: according to the IBM Institute for Business Value, AI is projected to increase the perceived value of EVs by over 20 per cent.

As experts in R&D tax credits at RandD UK, we've witnessed firsthand the significant financial investment required for AI innovation in various sectors, including the burgeoning EV manufacturing industry. While AI offers immense potential, the high failure rate of AI projects can significantly impact a company's bottom line. Statistics show that 80% of AI projects fail and it's crucial to understand why this is and how you can mitigate the fallout of these failures. 

How AI can be useful and worthwhile in manufacturing electric vehicles

From battery safety and performance to in-car assistance and predictive maintenance, AI offers a multitude of solutions that will significantly impact the future of EV manufacturing.

Improving EV battery safety and chemistry

New research from the University of Arizona demonstrates the potential of machine learning to reduce the risk of "thermal runaway" in lithium-ion batteries. Inspired by weather forecasting frameworks, their AI model can predict and prevent temperature spikes, a leading cause of battery failures, fires, and even explosions. This proactive approach can detect potential hotspots before they escalate, potentially preventing catastrophic events.

Not only can AI make EVs safer, the algorithms can also improve their performance and sustainability. IBM Research is using AI and machine learning to determine the specific battery chemistry and size that creates the best-performing EVs. These advancements, according to Benjamin Boeser at Mercedes-Benz, could unlock “a billion-dollar opportunity” in the EV market.

Reducing energy consumption

Other researchers are thinking beyond the internal chemistry of cars and discovering how AI can optimise energy consumption in EVs. By integrating real-time traffic data, weather conditions, and even road gradients, AI algorithms can calculate the most efficient travel routes, minimising energy consumption as a result.

For example, AI-powered GPS systems, like the one developed by researchers at the Arab Academy for Science in Giza, can select the shortest, fastest, and most energy-efficient routes, potentially saving up to 46% in energy by avoiding steep inclines. 

AI and in-car assistance 

By processing data from cameras, radar, and LiDAR, machine learning algorithms allow EVs to navigate complex environments and make real-time decisions for safe driving. For example, Volkswagen has integrated generative AI to allow drivers to interact with their vehicles more naturally, such as requesting re-routing to the nearest charging station.

Equally, companies like Nauto are developing AI-powered systems to address critical safety concerns like distracted driving. Utilising dual-facing cameras and sensors, Nauto's technology monitors driver behaviour, identifies potential hazards like pedestrians and cyclists, and analyses road conditions. This has the potential to greatly improve road safety and prevent accidents.

These are just some of the many ways we’re increasingly seeing AI integrated into EV manufacturing. Its potential also extends into revolutionary developments like “smart charging” (the ability of the EV to suggest the optimal times to charge) and “predictive maintenance” (detection of early signs of wear and tear).

What could the future look like for businesses that implement AI and it fails?

Despite the exciting and varied potential of AI for EV manufacturing, it's important to recognise that AI projects within this sector face double the failure rate compared to traditional development approaches. This is owing to a range of factors, including the insufficiency of stakeholders to understand which problems should be solved using AI, and the lack of necessary data to adequately train an effective algorithm.

Tesla served as an example of what can happen when the new technology goes wrong in 2024. Its autopilot feature made headlines when 13 fatal crashes were identified as caused at least in part by the AI. The manufacturer later had to recall 362,000 US vehicles after the NHTSA said the vehicles did not adequately adhere to traffic safety laws.

The implications for EV manufacturing businesses when AI fails can be severe and far-reaching:

Incidents involving AI failures can trigger investigations by regulatory bodies, as we saw with Tesla. This can lead to costly recalls, fines, and even legal action. 

News of AI failures, especially those involving safety concerns, can generate significant negative press; this has the potential to damage both the company's image and consumer trust.

Financial losses will be suffered when AI fails, including recall costs, legal fines, lawsuits, and the costs of re-engineering and testing to fix the AI failures.

How can EV manufacturers minimise the risks associated with AI experimentation and ensure it’s business effective?

In the initial stages, manufacturers should be realistic and careful when identifying where to apply AI. In some cases, projects fail because the technology is applied to problems that are overly ambitious, or infeasible. 

Once the project is underway, rigorous testing and validation procedures are essential. Ai models should be thoroughly tested in both simulated and real-world environments to identify potential failures before they impact the production operation of the EVs.

Additionally, AI models are only as good as the data they are trained on. Many AI projects fail because the organisation lacks the necessary data to adequately train an effective AI model; manufacturers should prioritise collecting high-quality, unbiased data to ensure reliable systems.

How failed AI trials can still be cost-effective for a business

Organisations that experience failed AI trials should not be discouraged. These setbacks, when approached strategically, can provide valuable insights and improve future projects. We recommend treating each project as a pilot and focusing on the way failed attempts can help you refine subsequent efforts.

Firstly, analyse the reasons for failure, whether that be data quality issues, incorrect model selection, or unforeseen challenges. These lessons should be documented and analysed carefully for future projects.

Even if the initial AI model fails, the data collected and processed during the project becomes a valuable asset which can be repurposed in a number of ways:

Identifying patterns in equipment failures to optimise maintenance schedules.

Analysing production data to pinpoint defects and improve product quality.

Optimising your supply chain by identifying any bottlenecks or inefficiencies.

Failed projects provide opportunities for team members to develop crucial skills. They gain practical experience in data preparation, model selection, training, evaluation, and more; this is expertise that can be invaluable for future AI initiatives, and projects in general.

Moreover, should an AI project fail and your organisation face losses, utilising government incentives like R&D tax credits can help offset these costs. This credit can provide a significant financial boost, allowing you to reinvest in further research and refine your approach. It’s worth getting in touch with an R&D specialist who can help you make the best possible claim.

In summary

While the potential for AI in EV manufacturing is undeniable, the high failure rate of AI projects can understandably make businesses cautious. With a strategic approach and a focus on continuous learning, AI can be a powerful driver of innovation in the EV sector. 

We predict that, in the coming decades, AI systems will play a crucial role in manufacturing optimal-performing electric vehicles. To remain competitive and contribute to a more sustainable future, EV manufacturers should seriously consider investing in AI solutions to address the unique challenges within this industry.