As businesses increasingly adopt SaaS solutions, many are realizing that standard platforms often fail to meet their specific needs. In an environment where personalization, scalability, and smooth integration are growing in importance, custom software development is proving to be a crucial strategy for companies to maintain a competitive edge.

In a recent online discussion organized by Genia Xasis, Timspark’s CTO Alex Velesnitski, Carl Eidsgard, Founder & CEO FenxLabs and Head of AI Advisory & Solutions at Timspark, and Jimmy Nassif, CTO & Co-Founder at idealworks, Co-Founder at SORDI.ai, explored the role of emerging technologies—particularly AI—in shaping the future of custom software development and business processes. Here are some key insights from the talk.

How AI-powered robots are transforming BMW and beyond

As Jimmy Nassif mentions, ‘BMW was the first to collaborate with Nvidia and the first to deploy digital twins, unveiling this innovation to the world in 2021. We strongly believe that Nvidia’s technologies, along with those of Industry 4.0 and even 5.0, will enable not just advancements in logistics but also in large-scale production. This is how we began at Idealworks, a subsidiary of the BMW Group. The original goal was to introduce new technologies to optimize BMW’s internal logistics and, crucially, to reduce costs. Ultimately, any new technology must have a clear business case, which often revolves around cost optimization.

In 2016, one of our key initiatives was the development of a smart transport robot to eliminate repetitive transport tasks in production and enhance the flexibility and efficiency of logistics processes. The budget for logistics at that time was approximately 13 billion euros, and our objective was to reduce this cost. With the introduction of the smart transport robot, we managed to replace many forklift operations within warehouses and production areas. The robot could navigate autonomously from point A to point B, marking our first significant use of AI—specifically in autonomous navigation.

In addition, we incorporated perception and recognition capabilities into our robots, enabling them to drive autonomously without relying on any IT infrastructure. These robots could recognize and avoid obstacles. Back in 2016, this was groundbreaking in the industry. Previously, AGVs (automated guided vehicles) followed either a physical or virtual line, but with the advent of AMRs (autonomous mobile robots), we developed machines that could autonomously navigate from point A to point B without following a predetermined path, powered entirely by AI.

This solution was custom-built to meet BMW’s needs because no existing market solution at the time could satisfy our requirements. After initial tests in 2016, we now, eight years later, serve 25 clients worldwide—not just automotive companies like BMW and Toyota, but also retail and logistics providers (3PL and 2PL), first- and second-tier automotive suppliers, and appliance manufacturers. To date, we have implemented over 1,100 robots across production sites and warehouses, all using AI-powered autonomous navigation, perception, and recognition, utilizing AI both at the edge and in the cloud.

We employ AI in the cloud to optimize traffic and fleet management, while edge AI ensures functionality even when Wi-Fi or connectivity is lost, a common occurrence in industrial environments. By splitting AI tasks between the cloud and the edge, we ensure that our robots continue to operate smoothly and autonomously, even in connectivity-challenged conditions.’

The biggest potential of AI robotization

Jimmy Nassif strongly believes in the automotive industry for one main reason: it is already far ahead in terms of automation. This doesn’t mean that automation or the need for robots doesn’t exist in other industries, but in comparison, they are still lagging behind. Automotive businesses are actively optimizing both production and logistics processes.

The competition from Chinese manufacturers poses a significant challenge for European producers. To reduce costs, companies must minimize reliance on labor and automate as much as possible. This is why there is a strong push for automation and robotization—not just to replace human labor, but also to support workers on the production line, which is equally important. High-quality processes are crucial to ensure production runs smoothly. For instance, at BMW Group, a car leaves the production line every 56 seconds. If the line stops for just one minute, the result is the loss of one car.

Maintaining process quality is essential to keep production running seamlessly while ensuring consistent quality. Most manufacturers operate three shifts, 24/7, to meet demand. However, this does not mean that other industries don’t require automation. In Jimmy’s opinion, the automotive sector is ahead of the curve, investing heavily in automation, while other industries are following suit, and he believes that first- and second-tier suppliers to the automotive sector will be next, as their supply chains are interconnected and must maintain competitive pricing to serve the automotive industry.

The role of robotization in today’s IoT industry

Jimmy Nassif: While robots are the hardware component, most of the operation now takes place in the software world. Essentially, all IoT devices need to be interconnected.

For instance, Jimmy Nassif mentions that ‘In our industry [automotive], robots must be connected to every IoT device on the production line. When a robot delivers goods to a conveyor, that conveyor needs to be linked through the cloud to recognize when it’s receiving goods and to send them to the next station. We also connect to lifts, traffic lights, barriers, and fire alarms—every device in the warehouse communicates with the robots. Looking ahead, I strongly believe that in the future, factories themselves will become robots, with every component interconnected.

Taking it a step further, let’s consider the Nvidia use case. We use simulation extensively to optimize and test our processes because we cannot afford any downtime in production. It’s crucial to test everything in a simulation environment before implementing it in the real world. Moreover, when creating digital twins, we gather synthetic data from them, which we use to train our algorithms. It creates a loop: AI helps create digital twins, which are connected to real robots, and the synthetic data from these twins is used to train the robots, making them more efficient and autonomous in decision-making.

To achieve high accuracy in computer vision for perception and recognition, models need to be trained with vast amounts of data—potentially billions of images, which is nearly impossible to gather in real time from the real world. This is why we rely heavily on synthetic data. Currently, about 80% of our models are trained on synthetic data, with the remaining 20% on real data. Once trained, we evaluate these models using 100% real data to ensure they are reliable enough to make correct decisions in real-world production and logistics processes.’

Automation in the GCC region—what are the peculiarities?

When examining the ROI in the GCC region, it is significantly lower than in Europe due to high labor costs. Currently, labor costs in the area are rising, prompting businesses to seek automation solutions to maintain quality and scalability. To ensure both scaling potential and consistent quality, automation emerges as the optimal answer. Consequently, major companies in the GCC region are increasingly focusing on and investing in automation, particularly in AI.

Challenges of integrating AI automatization

Jimmy: The main challenge lies in gaining acceptance from people and managing the accompanying changes, a process we refer to as change management. It’s essential to ensure that employees understand why automation is being implemented, how it supports their daily tasks, and how they can contribute to its swift implementation and optimization.

The technology itself is not the issue. Instead, it’s crucial to identify the real problems that need solving and involve the people working on those problems in the journey. By leveraging their knowledge, you can automate your systems effectively and enhance overall efficiency. That’s how the process should work.

How automation and AI practices can be customized

Alex: I personally have a strong belief in the healthcare trend. In my view, this is a significant development because using AI for patients—such as implementing predictive analytics and creating personalized treatment plans before any issues arise—could be critical for society. The potential impact is truly transformative.

For one of our clients, a European healthcare provider, we developed a mobile application capable of diagnosing asthma. This project involved a dedicated team of six developers from our side. Technically, we used Python for the development. The team did an outstanding job creating an application that helps monitor asthma symptoms, provides analytics, and tracks potential triggers.

I believe clients will become accustomed to this trend in the coming years. Wearable devices are already on the market, and I anticipate an increasing number of applications and software with groundbreaking capabilities to detect health issues.

Second topic I see promising and evolving in the market is natural language processing (NLP) tools. My team is currently working with Google’s GMI and Facebook’s LLaMA open-source models, which are becoming industry standards. I personally believe that NLP may not yet be fully ready for mission-critical systems. However, with OpenAI’s recent release of O1 which I call GPT-5, I see potential for significant advancements in this area.

And the third point of AI application is Fintech and Banking, helping to detect unusual behavior during the transactions. A prime example of AI’s impact on Fintech can be seen in one of our projects on machine learning in banking. We helped a client develop a machine-learning solution for one of the leading banks in the U.S. to detect transaction anomalies and prevent fraud. The system uses deep learning to analyze vast datasets, identifying suspicious patterns in real-time. The back-end was built with Python and Scala, using tools like Apache Spark and Scikit-learn.

Speaking of proper AI implementation, I believe that the whole data should be prepared—labeled and normalized in some way. But again, we at Timspark are not doing only AI things. We can provide a full-cycle development. Why talk about AI only when you can have a full-cycle brilliance?

The rise of AI agents in business

Carl: When the new generation of generative AI models emerged in 2022, it was my signal to branch out and start something on my own. Initially, I collaborated with Tim Spark while building my own venture, which is now operational as FenxLabs Labs. What we do is use generative AI models to create specialized agents capable of performing automated tasks. Essentially, we offer automation platforms.

Unlike Alex’s point about customers providing data, our approach is different—we fully deploy models, infrastructure, and everything directly into the customer’s environment. The unique aspect we bring is our agent configuration. Agents allow us to scale an AI system’s capabilities by duplicating and fine-tuning model instances to perform specific tasks. When you combine all these elements, you create a system that can automate nearly anything. It acts as connective tissue between existing systems, enabling you to integrate with all apps, legacy systems, and infrastructure to automate tasks across the board. That’s our focus. So it’s the same alley, same street as Timspark, but a different type of shop.

The most effective application of AI

Carl: When you have this conversation about AI, it’s very important that you categorize the discussion or the topics correctly and assign them to the correct category. So when Jimmy said that he predicted that automotive would be the first ones to automate, I actually don’t think so. I think that the companies that manage to build this ecosystem using large language models, And I can do that, or effectively, they will be the first. And especially if they’re lightweight businesses, like, say, an agency of some kind. I think that that is true huge potential. And if you work for an agency out there, I would highly recommend that you start looking into how can automate.

The future of AI—what will it be?

Carl: Where we are in a year largely depends on whether a global recession occurs. If it does, companies will have a strong incentive to automate in order to cut costs. Without a recession, automation will rely on the pace of human adoption, which tends to be slower. Looking back at the past two years of generative AI, the technological advances haven’t significantly changed the landscape, largely due to slow human adoption. However, if technology continues to advance and automation accelerates, companies that embrace it early will outcompete those that don’t. That dynamic, though, may take more than a year to fully unfold.

Also, when we consider the future of AI, it’s not really about gimmicky features like AI avatars, content generation, or adding superficial enhancements to SaaS applications. In my view, these aspects are largely irrelevant. The real significance lies in what happens when models can do everything straight out of the box—like when you log into ChatGPT and ask it to write an application, and it just does it. What becomes important then?

The key to the AI revolution is the ecosystem. In the future, AI agents will be performing tasks for us, and the crucial question will be: who has control over that ecosystem? If everything is handled by OpenAI and ChatGPT, then OpenAI could become the most powerful company in the world, effectively controlling the market. I personally don’t agree with that approach. I believe the ecosystem should be diverse. While it may be hard for some to envision, creating diversity in the ecosystem is actually achievable. If you can spin up an automated AI agent system that you control, you’ve built the foundation for how you’ll interact with markets, society, and the economy in the future. This is why we advocate for building your own automation platform. We don’t dictate how to do it—if you need guidance, companies like Timspark and FenxLabs Labs are here to help.