From environmental crime to fighting hunger: TechDays 2026 shows how AI is already making a difference in the real world

An AI tool helping police investigators uncover new leads. IKEA using AI to help customers redesign their living rooms. And a smartphone that could help identify malnutrition in children. TechDays 2026 showed just how widely AI is already being applied, including to tackle major societal challenges. But it also revealed how much work it can take to turn a promising experiment into a reliable, real-world solution.

Building an impressive AI demo is one thing. But how do you make sure the technology actually works beyond the lab or drawing board? That question ran through TechDays 2026, KickstartAI's annual AI conference, held on 15 October at Circa Amsterdam. Around 370 AI engineers, data scientists and technical leaders came together across 17 sessions to share their experiences of turning experiments into working AI applications. Here are some of the highlights.

Europe needs to take matters into its own hands

It starts with a fundamental question: who actually controls the technology? Lokke Moerel, Professor of Global ICT Law at Tilburg University, opened the day with a warning about Europe's growing dependence on American technology.

"We have to be realistic. We're probably not going to develop a language model here that can compete with the biggest models from the United States and China," Moerel said.

She also pointed to a new kind of dependency. For years, organisations have worried about vendor lock-in, where switching suppliers becomes difficult. Now there is also the threat of lock-out: the risk of companies suddenly losing access to critical technology altogether.

Still, Moerel sees opportunities, particularly in smaller language models developed for specific industries. Her message was clear: European organisations need to invest in alternatives together, such as GPT-NL, the Dutch initiative to develop its own language model.

"If we all keep making individual decisions to use certain Big Tech services, we'll never build alternatives of our own."

AI builds an app almost by itself

But having access to technology is only half the battle, as Pratyush Sinha from Decathlon demonstrated. Together with his colleague Abdul Rahman, he spent six months building a prototype app designed to help employees learn more about AI. The twist? They handed most of the development work over to AI agents. The agents created lessons, quizzes and animations tailored to users' existing knowledge. The result: 20 lessons and 193 animations. But also plenty of headaches.

The agents sometimes cited the wrong sources, failed to communicate properly with one another and produced animations that were only half finished. Sinha therefore developed a system in which different agents check each other's work and fix mistakes. "If you want to build reliable systems, focus on governance," he told the audience. AI can take over a great deal of work, but someone still needs to decide what to build and judge whether the results are any good.

New leads for police Investigators

That human judgement also plays an important role in Al Capone, an AI tool developed by KickstartAI's Henk Snepvangers and Sanne Wielinga together with the Dutch National Police. The application helps investigators uncover new legal angles in environmental crime cases, such as illegal waste shipments or the dumping of toxic substances.

These investigations are often complicated. Detectives have to work through vast amounts of legislation and cannot possibly be experts in every area of law. Al Capone therefore also searches for rules and regulations that investigators might not immediately think of.

The name refers to the notorious American gangster who ultimately went to prison not for murder, but for tax evasion.

Six investigators tested the tool using real, anonymised cases. Nearly two-thirds of its suggestions were considered relevant, while 8% actually provided a new insight. Whether such a lead holds up legally is still for the investigator to determine.

Read more about how the KickstartAI team built a retrieval engine for crime investigations.

IKEA and the oversized Sofa

It is not just complex police investigations that require careful checks. Even something as seemingly straightforward as furnishing a living room can prove surprisingly difficult for AI. IKEA demonstrated this during a plenary session. The company is testing an application in markets including the United States and the United Kingdom that lets customers see what new furniture might look like in their own living rooms. The problem? AI sometimes takes a rather creative approach to interior design. A sofa suddenly becomes much bigger than it really is, an extra window appears out of nowhere, or two pieces of furniture end up merging into one.

"We need to make sure that your room stays your room," said Shahin Shahkarami, Global Vice President AI & Automation at IKEA Retail. That is why IKEA has built in a range of checks designed to ensure that the original room stays the same and that furniture looks the way it does in real life.

Zero Hunger Lab

The importance of getting these details right was also clear in a very different field: childhood malnutrition. Researchers at Tilburg University's Zero Hunger Lab are developing a method to estimate children's body measurements using a smartphone photo. This could help identify malnutrition earlier, particularly in areas where healthcare professionals are in short supply. For their research, the team collected data from 512 children in Iraqi Kurdistan. The initial results are promising, but plenty of questions remain. How accurate are the measurements for children from different backgrounds? And how can their privacy be protected?

Learnings from TechDays 2026

How do you put AI to work in the real world? What obstacles do you encounter along the way? These are exactly the kinds of questions TechDays visitors are dealing with in their own organisations.Take Jeanine Vonkeman, AI Delivery Lead at Boskalis. She works on AI solutions for the company's engineers and came to TechDays to find out how other organisations are approaching similar challenges.

"What I like about TechDays is that it's not about the AI hype. It's about real technologies being applied in the world," she said.

Two insights from the morning particularly stayed with her. First, having high-quality data can be more valuable than having huge amounts of it. And second, as AI takes on more tasks independently, the role of people changes, but they do not become redundant.

"You can still control the quality of the work as a human outside the loop. It doesn't eliminate the human, it just changes the role you play."

That same confidence in people was reflected in the closing remarks from KickstartAI CEO Erick Webbe.

From rising healthcare costs to an overstretched electricity grid and growing mobility challenges, there is no shortage of major problems to solve. Webbe believes AI can help tackle them, provided people are willing to combine their expertise and take action.

"What stood out to me most today is just how much we can still shape our future.” Because ultimately, it's not just about what AI can do. It's about what we can achieve with it, together.

Want to keep the conversation about applied AI going?

TechDays is just the beginning. Over the coming months, KickstartAI will organise further events on topics including AI and food security, and building reliable AI agents. Another Data Science Summit and the next edition of TechDays are also in the works.

Check out the KickstartAI events calendar and join the conversation.