How is AI being used by industrial companies like Sandvik, Valmet and Normet today – and what are their plans for tomorrow? Unikie’s recent breakfast event brought together industry professionals to talk about the utilization of AI in combination with data to improve processes, automation and productivity, for example.
The event was hosted by Unikie’s CTO Niko Haatainen and featured a panel discussion with Olli Snellman, Head of Automation Technology Programs at Sandvik, Olli Mylläri, VP, Technology at Normet and Hannu Lätti, Sales Director at Valmet.
Using AI to Build Better Products
The benefits of artificial intelligence, especially generative AIs, are often connected to things like efficiency and cost savings. While those are important, from Unikie’s perspective – as stated in Unikie CEO Juha Ala-Laurila’s opening words – the recipe for the ultimate AI payoff is a bit different: using AI to build better products. In the long run, what matters most is that the product is competitive and the very best solution there is.
In the industrial sector, introducing new technologies takes its time and requires careful planning for the years to come. Unikie’s role is to apply its deep know-how on the intersection of hardware, software and data to help its customers to practically get going with their development efforts.
AI Agents and AI Everywhere
In Niko Haatainen’s talk, the Unikie CTO highlighted two topical aspects to AI: AI agents that are independent AI operators focusing on a particular task and AI Everywhere, which means bringing AI capabilities from the cloud to the edge, closer to devices and hardware.
According to Haatainen, when individual AI agents are orchestrated to work and solve problems together, the potential for automation reaches a new level. This allows the human expert to interact with the system on a higher level, and possibly just oversee and review the AI agents’ plans before execution. This results in less manual work, faster execution and more meaningful work for the human expert.
Adding AI capabilities and edge computing to embedded solutions is partly being driven by smaller, more capable chips that are constantly becoming cheaper. Utilizing smaller, distilled AI models that are tailored for specific purposes, it is already today possible to include AI models that run locally and manage the industrial processes.
Lastly, Haatainen touched upon a topic very important to industrial AI applications: security. While the safe route is to use AI non-critical tasks like data analysis and reporting, developers should not sign off considering AI for mission-critical purposes. Proper monitoring, sandboxing and reviewing – possibly with other language models – can help introduce AI to deeper tasks as well.
Panel Discussion: Where Is the Industry Going with AI?
In the panel discussion, representatives from Sandvik, Normet and Valmet more or less agreed that currently in their organizations AI is being used on three separate levels.
On a personal level, individuals use AI to improve their productivity, through coding assistants or using tools like Copilot to manage and summarize information. On the higher level, AI is being used for making work processes more efficient, such as generating or analyzing documentation and test cases for quality assurance, for example.
The third level of using AI is in product development – incorporating AI to build better solutions. To this end, the companies are using various technologies, such as high-resolution imagery combined with machine vision, machine learning for process modeling and optimization, and making use of AI to combine and validate data quickly.
Edge computing is seen as important, as for example in mining operations network bandwidth is a consideration. Also, much of the data must be processed locally almost in real-time to allow autonomous machine actions.
The panelists agree that utilizing AI in industrial projects takes time. For example, in a new project, the first year could be simply about gathering and annotating data. The slow pace is seen in team development, too: you cannot simply drop 10 AI experts to the team; they have to learn the ins and outs of the domain they work in to be successful.
There is a consensus that the industry is still in early phases of the learning curve. Each project teaches a lot, and going forward step by step seems the best way forward, especially when enhancing existing products. Gradual progression also helps in proving the value of new technology and earning credibility with customers.
All in all, the panelists agree that organizations should invest in having the right competence to both make strategic decisions on the use of AI as well as expertise to deliver on those plans.
Learn more about Unikie’s offering for AI & Data Analytics
