“I wanted to create technology that will work hand-in-hand with workers, rather than replace them.”

In today's competitive landscape, manufacturers are more and more pressed with the need to modernize and streamline their factory processes and workflows to meet the demands of the market and stay ahead of their competition. Berlin-based start-up Deltia.ai wants to fill that demand and empower companies to enhance the efficiency of their manual assembly lines drawing on sophisticated predictive analytics and data visualization solutions. We talked to CEO and co-founder Max Fischer to find out more about the interplay between human and machine, the importance of data privacy and the key challenges working in an industry in transformation.

Hello Mr. Fischer, your team believes that humans are critical to manufacturing processes, and your AI solutions are designed to empower and support humans in their work. What’s the origin of your mission? Why is it especially important in manufacturing?

On digitizing over 40 factories, we started noticing similar patterns. Any cutting-edge technology that we bring to the shopfloor, they are only effective when workers realize its potential and are willing to adopt it. Since the workers spend so much time close to these machines and processes, I believe they are the best judge of our innovations. Human involvement is critical to resolve complex issues and to make decisions that involve multi-dimensional factors that may be beyond a machine’s comprehension. With the rise in various AI-based tools and features in many SaaS solutions, most industries have noticed a paradigm shift. It was these realizations that motivated me to create technology that will work hand-in-hand with workers, rather than replace them.

As we delved more into the solution, we realized there was a clear pain point to solve. To learn more about the manual process in production, process and manufacturing engineers relied on outdated methods like MES, and manual notetaking. This discontinuous and often unreliable data created a black box for process optimization where they had no visibility into what exactly happens on the shopfloor. With more value-adding tasks in their to-dos, they shouldn’t spend their time observing the manual assembly processes to figure out where optimizations can be made. This is exactly where we can leverage technology; to perform a repetitive, mundane yet important task. Our technology would act as the co-pilot for manufacturing where the leadership can go through suggested improvements based on granular data and prioritize implementing them to boost productivity.

Can you describe the core functionalities and innovations of your AI solution, and how it differentiates from other AI solutions on the market?

Deltia digitizes shop-floor processes with computer vision, supporting continuous improvement and shop-floor management. By using cameras and AI, we track cycles and non-value adding activities to identify improvement potential for manual shop-floor processes. We currently have features that help detect bottlenecks, process anomalies and productivity drops. Data from our platform is useful for root cause analysis and other continuous improvement projects. Deltia can also automatically track machine processes to notify automatic interruptions and perform predictive analysis for machine failures.

Making continuous improvements and generating cost and efficiency savings require reliable data. Deltia bridges that gap. Deltia’s infrastructure gathers context-rich, real-time data and uses AI to do the monumental processing work. This data is then used to make suggestions and identify actionable improvements. What sets Deltia apart, is the fact that our solution is flexible to adapt to any use-case on the shopfloor. As long as the organization has semi-standardized processes with similar products, our model is able to learn to track processes and identify deviations, be it manual or machine-based. From planning and scheduling to prescriptive analysis, we offer a comprehensive solution that encompasses all facets of manufacturing.

How does your AI solution integrate with existing manual processes, and what are some of the key challenges you face during this integration, especially in the context of industry transformation?

We set up a mix of station-mounted cameras and bird-view cameras to generate thousands of data points every day, calculating the insights you need for further analysis. These video streams are continuously analyzed to detect workpiece movements, cycle times, and work step sequencing. Process data is aggregated per article and production in a cloud-based software analytics application. Factory managers and process engineers define, implement and measure process improvements to increase productivity and quality in their assembly line.

One key technical challenge was to cope with the huge variety of different processes in the manufacturing industry. To give you one example that might sound simple but is technically quite hard in practice. We are tracking the cycle, i.e. the time between a product entering and leaving a station. To do this reliably, we need to accurately identify product movements. However, every product at every line and every customer is very different. Our system needs to cope with this variety, without requiring a lot of domain input from experts or costly labeling work.

Another challenge was obviously the topic of data privacy and IT security, since cameras are usually used for surveillance purposes. Our usage of cameras as sensors to generate process statistics required a bit of explanation, especially our measures to ensure data privacy. Our approach of processing videos only on the edge and then automatically deleting them really helped with that.

Could you provide specific examples or case studies where your AI solution has significantly improved efficiency and outcomes within an industry or for a respective enterprise?

One of our bigger earlier successes was that we had found our initial buyer before the company was founded. This showed us how grave the issue was, and how crucial our solution was going to be. Recently with one of our customers, we helped to increase output of a line by more than 40% by focusing on eliminating bottleneck processes and optimizing by training everyone on best-practices. Through our suggested changes to their standard work model, they were able to increase their total revenue by 2 million euros.

We also helped another customer improve their efficiency by 30%, by tracking an undocumented shop-floor best practice and including it in the standard. At the end of the day, seeing our solution making a positive difference for these workers is the most rewarding success. All said and done, it's the most rewarding to see our solution enable workers to put out their best. We also started to introduce ergonomic assessment in our product, since a lot of shop-floor workers complained about back and wrist pain. We will automatically give feedback to workers on unergonomic movements helping to improve health and well-being.

The EU AI Act introduces stringent regulations. How is your company preparing with CertifAI to comply with these new rules, and what challenges have you encountered in this process?

While the Deltia product enables access to shop-floor data via state-of-the-art computer vision, we, as a company, never compromise on our privacy principles. Deltia anonymizes all video footage, for the security and comfort of workers on the floor. Videos are also only processed on a PC at the production line, with the raw-video footage being automatically deleted after a few seconds. This means that customers get unprecedented insight into their manual processes, without wading into the murky waters of GDPR and AI Act compliance issues. We have a dedicated team internally working on security and compliance with major standards like GDPR, and ISO. We keep an eye on making sure our product is always compliant. The security of our customers and those working on the shop floor has always been one of our top priorities. That’s why we’re working together with CertifAI to make sure Deltia’s products meet the highest standards through comprehensive AI Product Testing and certification.

Berlin is a unique hub in the German AI landscape, what is for you that makes Berlin a profitable location?

Berlin is definitely a great place for our venture. The city is young and energetic. In no other city have I seen an initiative like the Merantix AI Campus, which supports AI ventures in every stage. In fact, I met my co-founder, Silviu Homoceanu at Merantix. It has also been a great place to build an international team who are super motivated and bring merits from different backgrounds to the table.

What advice would you give to young entrepreneurs who plan on setting up their own AI company in Berlin?

I may not be the best person to give advice, but I can certainly share what I’ve learned, and which mistakes I would avoid repeating. At Deltia, we are dedicated to solving significant and impactful problems for our customers. One crucial lesson I’ve learned is to not let my love for technology blind me into believing that every problem we address is extremely important, that customers will pay a premium for our solutions, and that the market demand is infinite.

It’s vital to focus on the fundamentals and ask critical questions such as: Which KPIs are we actually impacting? How much money can our customers genuinely save with our solution? How much better is our product compared to alternatives, like in our case using paper for data collection? We learned at Deltia for example that customers are not paying us because we provide automated digital data, but rather we are one of the only systems that allows you to understand WHY certain things are happening on the shop-floor.

By addressing these questions, you can better determine whether to build a product and which customer segments to target. This clarity is essential for making informed decisions and ensuring that your efforts are truly aligned with customer needs and market opportunities. It helped us focus and at times even not pursue some customer opportunities, since we didn’t believe the value we could create was big enough for them.