AI is not new - but maturity is
When Melina Katkic first became interested in artificial intelligence, the field was far from mainstream. Back in 2015, she founded a network for AI enthusiasts - an initiative that later became the consulting company NordAxon.
"I followed my passion for AI. Those were the courses that were most interesting during my education in computer science and economics at Lund University," she says.
The history of AI is much older than many people think. The technology's roots stretch back to 1956, but its development has been marked by recurring 'AI winters' - periods when interest waned. It was only around 2012 that the technology really took off, thanks to breakthroughs in deep learning and access to large amounts of data.
Transformers, ChatGPT and AI democratization
In recent years, developments have accelerated significantly - not least through technologies such as the transformer architecture (the basis of e.g. ChatGPT) and platforms such as OpenAI and Hugging Face. These have made advanced AI accessible to more people, not just the tech giants.
"The democratization of AI is a milestone. When the tools became available to more people - both open models and cloud-based platforms - new opportunities opened up," Melina explains.
But at the same time, she warns against reducing AI to mere generative tools.
"ChatGPT is AI - but AI is so much more than ChatGPT."
How to create real value with AI
What can AI actually be used for in industry? Melina is clear: it's always about creating business value.
"We usually start from three values: increasing revenue, reducing costs or improving quality. Then we look at the type of data available - image, text or time series - and choose the method accordingly."
The three major AI fields she refers to are computer vision, language learning (NLP) and time series analysis - all with different uses. Identifying defects in the production line, predicting demand or integrating AI into customer service are just a few examples.
But it's not just about technology - it's about mindset.
"The companies that succeed are those that have people in leadership roles who are not afraid of AI, but want to understand it. That takes courage, curiosity and a willingness to learn," says Melina.
Five paths to AI implementation
According to Melina, there are five main ways to implement AI in an organization, depending on needs and resources. From using ready-made apps, to building your own models from scratch.
"The most exciting thing for us is when there is no ready-made solution - when we can build a model from scratch based on the customer's own data," she says.
But the cloud is not always the solution. Data protection concerns are common in industry, and with the right architecture, AI solutions can be built entirely on-premises.
"You don't have to upload the data. We can download an open model and train it on a local server. It's about finding the right technical path from the start."
AI as a tool for good
The conversation ends with an important reminder:
"AI should be used to help people, society and our planet - not the other way around," Melina emphasizes.
As AI continues to fundamentally change the industry, how it is used is up to us. It's not just about algorithms and code, but about vision, values and the courage to think beyond the next quarterly report.
Final reflection
AI is not a silver bullet - but used correctly, it can be a catalyst for real transformation. It requires both technology and trust.
And perhaps most importantly, people who dare to lead the way.
