Community Spotlight: Alexander Maas and Marinus Israël

Bridging AI’s business gap

By Ioanna Lykiardopoulou

Every month, we spotlight a member of our community. At KickstartAI, we believe that meaningful progress starts with individuals — engineers, researchers, founders, and practitioners — who are applying AI in bold, practical ways. Through our Community Spotlight series, we celebrate the people shaping the future of AI in the Netherlands: their work, their challenges, and the lessons they’re learning along the way.

This month, we’re featuring Marinus Israël and Alexander Maas, co-founders of Algemi — a generative AI agency helping businesses bridge the gap between AI’s potential and its practical use. From their first collaboration on a refugee support platform to launching a company focused on practical AI, this is their story.

Alexander Maas and Marinus Israël met in 2022 while working together on the development of RefugeeHelp — an online platform which helps refugees find their way to the Netherlands.

Following a fruitful collaboration during that project (and an even more fruitful coffee meetup), the duo decided to build a company together. In 2024, they founded Algemi, a generative AI agency aiming to help businesses adopt AI solutions and harness the technology’s full potential.

“We saw a gap in what AI can do and what businesses are using it for. So we thought ‘well, let's start a company together to try and fill that gap,’” Alexander says. 

The journey towards Algemi 

Alexander’s passion for AI began early, but he chose to study aerospace engineering and spent several years at Airbus Defense and Space, contributing to satellite missions. He continued exploring software and AI as side projects, launching several startups alongside his aerospace work.

Eventually, this led him to shift careers and join a software consulting firm, where he worked as a senior software engineer and, later on, as tech lead and CTO. 

Marinus’ interest in digital technology also started early on in his life. He built his first website when he was around 12 years old, and created his first company — in web hosting — when he was 16. He has since founded multiple startups and is specialized in digital strategy and product management and development. 

“What we’ve both seen in our past,” Alexander says, “is that 90% of initiatives in software and digital product development have failed. The reason why has mostly to do with the fact that companies don’t think about the end customer.” 

“And we see that a lot of companies currently trying to use AI are making the same kind of mistake. They focus too much on small efficiency gains. They forget the huge impact it will have on how work is performed and what is considered valuable by customers in the future.” 

This is what the duo has set out to do differently with Algemi. The startup both helps companies identify AI use cases and builds custom AI and software solutions. Currently Algemi consists of a core team of four, supported by a vetted freelance pool.

Identifying AI use cases 

According to Marinus, Algemi’s approach to identifying where AI can bring value relies on combining knowledge and effort. 

“From our side, we know a lot about AI and what it can do. We also have a pretty broad understanding of how different types of businesses work,” he explains. 

However, the most important factor, he says, is gaining a deep understanding of the business from the inside, which can only be done in close collaboration with Algemi’s clients. “They know their processes. They know their customers. They know the pain points and what they envision for the future.”

To guide their work, Marinus and Alexander use a framework that divides AI use cases in three parts. The first is efficiency gains — using AI to optimize existing operations. The second part is growth, looking at previously unfeasible products or services that AI can enable. The third part focuses on where AI itself may put a business at risk. 

Once the team identifies the potential use cases, they pick the ones that can bring value in the short term. “We like to focus first on the low-hanging fruit,” Alexander says. “This provides customers with an easy first experience with AI. And with AI developing so quickly, what is now difficult to solve might be easy in the next few months.” 

The team then builds a proof of concept to test whether their AI solution is feasible. The next phase is developing a Minimum Viable Product (MVP), involving the end users as much as possible throughout the process. 

“We let them try it out, collaborate closely, and learn what tools they already use and how the solution can fit into their existing workflow,” Alexander explains. “A tool is used most often and effectively when it is easy to apply and fits well within your normal workflow.”

Following the MVP’s launch, the duo focuses on fine-tuning and improving based on user feedback. 

Putting LLMs to work 

When building AI solutions, Algemi leverages existing LLMs, selects the model that best fits the use case, and fine-tunes it accordingly. For most of the applications they’re building, the team uses OpenAI’s GPT models. 

“LLMs are starting to get their own personalities,” Marinus says. “For instance, OpenAI’s GPT models are strong all-rounders, Claude is great for software engineering, and Gemini is especially effective with health-related tasks.”

In some cases, they already know by experience which model will work best, but they have also developed a test suite to match the right LLM and the right use case. 

This test suite also helps evaluate use cases. “What we do is create a human-curated test set — a list of questions with expected answers,” Marinus explains. “Sometimes the answer is straightforward, like ‘yes,’ ‘no,’ or a number. Other times, it’s a bit more fuzzy.”

The team runs these questions through different models and compares their output to the expected answers. This allows them to see how well a model — or the overall use case — performs.

It also helps the duo fine-tune the models to improve accuracy and, importantly, gives them a way to show clients measurable results. “We can say, ‘This model gets 90% of the answers right, for example, and we can prove it,’” Marinus adds.

“In most cases, almost right is still wrong, however,” Alexander adds. For this reason, Algemi is conducting human or automated quality checks, aiming to get the correct result 100% of the time with minimum human effort. 

Advice to AI founders 

Building an AI company comes with unique challenges, one of the biggest being the speed at which the sector is moving. The premise is simple: what is an opportunity at the moment might no longer be an opportunity in just a few months from now. 

A significant learning for Marinus and Alexander has been that sometimes it’s better to procrastinate building your own generally applicable tools. With these tools it is better to wait and see what kind of groundwork gets developed, and make use of that.

Marinus advises founders to start with a niche they understand well — whether it’s a specific industry or process — and combine that domain knowledge with AI’s capabilities. “I’d say, don’t go big on general AI because you’ll need a lot of money and you’re up against fierce competition from Silicon Valley.” 

Alex shares a key insight (originally from Sam Altman) that he wishes he had known earlier: it's better to build products that benefit from more advanced AI, rather than compete with it. “For example if you build a product for a hospital, focus on integrating it into their systems using existing AI models over developing your own custom models. A more intelligent model is bound to come around in a few months which will only make your product more effective — instead of replacing it.”

The Dutch AI ecosystem 

While the Netherlands is strong in leveraging existing AI models, it falls behind in foundational AI development, Alexander and Marinus note. They both advocate for further investment into building data centers in the country and advancing core AI capabilities. 

“It’s something that’s going to be a strategic resource at some point,” Marinus says. “You don’t want half of your country to be relying on technology that’s hosted in a data center somewhere else. I also think it’d be a great idea to train a Dutch LLM.” 

He nevertheless remains optimistic, seeing excitement across the sector and the creation of various AI communities, which could eventually lead to a single, all-encompassing one. 

Marinus particularly appreciates the meetups of the KickstartAI community. “It’s a great opportunity to mingle, easily present what you’re doing, get to know what other people are working on, and keep up to date with developments.” 

Alexander also sees value in exchanging ideas during the meetups, especially when it comes to creating awareness about the necessity of quickly implementing AI  within companies. “It’s interesting to see how other businesses are approaching this,” he says. “Sharing this knowledge is very valuable, and I think the market is large enough for all companies to survive. It’s more about winning together by solving the issue [of creating awareness for AI adoption] than about giving away our competing advantage.” 

💡 Want to learn more about the KickstartAI community and our mission to accelerate AI adoption in the Netherlands? Check out our website, listen to our podcast, or follow us on LinkedIn!