Smart Manufacturing & Materials-2.jpg

<p><strong>Most applications of artificial intelligence are impressive, but they have one major drawback: no one can explain exactly how they arrive at an answer. For many applications, that’s not a problem. But when AI is used in decision-making regarding safety, emergency response, or government processes, transparency is essential. That’s why the Enschede-based startup Thoughtful Oasis is developing a new generation of AI that not only provides smart answers but also explains how those answers are arrived at.</strong></p><p><strong>With this approach, the young company is now collaborating with organizations such as the North and East Gelderland Safety Region (VNOG) and the Ministry of Justice and Security, where the first applications are currently running in the pilot phase.</strong></p><figure class="image"><img alt="Thoughtful Oasis.jpg" src="https://cms.aihub-oost.nl/uploads/Thoughtful_Oasis_8813cf4018.jpg?token=QKPAGTnog62h-f7LebemGFy9vjAjT6xGOIbopLIsnkw&amp;expires=1787832802" srcset="https://cms.aihub-oost.nl/uploads/thumbnail_Thoughtful_Oasis_8813cf4018.jpg?token=TErrn7qe2UiAli-Kn0BTX3_C2B1QIcZfzFwESrm_a88&amp;expires=1787832802 234w, https://cms.aihub-oost.nl/uploads/small_Thoughtful_Oasis_8813cf4018.jpg?token=S9JZLkT4Cz-EGhS-maxlsEN7BzXwkrEypbHc7DQWyBk&amp;expires=1787832802 500w, https://cms.aihub-oost.nl/uploads/medium_Thoughtful_Oasis_8813cf4018.jpg?token=N9UQZhH7resTMm2ROA36Q16DqF9PrhssujAwbph3_gM&amp;expires=1787832802 750w, https://cms.aihub-oost.nl/uploads/large_Thoughtful_Oasis_8813cf4018.jpg?token=gqUKRmgTjghw6BOhw4AaFU92gmRFGYcwFNLmaGGBs40&amp;expires=1787832802 1000w" sizes="100vw" width="1000"></figure><h2>From a Black Box to Explainable AI</h2><p>Many AI systems convert information into complex computational models that humans can no longer follow. This makes the technology powerful, but also difficult to verify. Thoughtful Oasis saw an opportunity here. Instead of having AI operate exclusively with incomprehensible computational values, the company developed an intermediate layer that uses concepts chosen by subject matter experts themselves. Think of terms such as urgency, type of danger, or authority.</p><p>“The computational power remains the same, but the result changes,” says Rens Dommerholt, co-founder and CEO of Thoughtful Oasis. “Every answer can be traced back to the concepts that produced it. So not a score of 0.83, but high overlap on urgency, average on hazard class. This makes AI verifiable—not by subjecting it to an audit afterward, but because it’s built that way.” According to Dommerholt, this not only makes AI more transparent but also more suitable for environments where decisions must be justifiable after the fact.</p><h2>Scientific Basis as a Foundation</h2><p>This approach—which Thoughtful Oasis itself calls an ontological vector database—did not originate from an existing software product, but from the company’s own scientific research. The company has elaborated the method in three scientific publications and made them publicly available. The applications currently running for customers are a direct translation of that research.</p><p>The biggest challenge was not in devising the concept, but in demonstrating that it actually works. For a long time, it was assumed that powerful AI and full explainability were incompatible. Thoughtful Oasis wanted to prove that the two can indeed be combined. To do so, they not only had to develop an understandable conceptual framework but also demonstrate that a language model consistently evaluates the same situation in the same way. They succeeded, and the first pilots are already up and running.</p><p>“An added benefit is that the data layer we developed is independent of the AI model. This allows organizations to switch to new language models in the future without having to start from scratch,” says Dommerholt.</p><h2>Faster and Better-Informed Decisions for the Fire Department</h2><p>One of the first applications is MissionOS, developed for the North and East Gelderland Safety Region (VNOG). “During incidents, commanders often have to make decisions within seconds, while the correct protocol is contained in a manual hundreds of pages long that no one has time to flip through while riding in a moving fire truck,” explains Dommerholt. MissionOS analyzes the situation based on specified characteristics such as the type of incident, available resources, and authorities, and then identifies the most relevant protocol.</p><p>“The major difference from traditional AI solutions is that every recommendation can be directly traced back to the sources and concepts used. This allows users to verify why a particular recommendation is being given,” Dommerholt continues. The pilot is currently underway with fifteen VNOG employees. More than 250 simulations have been conducted so far, and a live rollout is planned for after the summer.</p><figure class="image"><img src="https://cms.aihub-oost.nl/uploads/to1_9772ccfbe7.jpg?token=R5XB24ItOFf2xntDxLc6cJ-iClJgwvdLshOqBvVX1iY&amp;expires=1787832606" alt="to1.jpg" srcset="https://cms.aihub-oost.nl/uploads/thumbnail_to1_9772ccfbe7.jpg?token=0xU7lh5wRkim5OgyuTRbRsCgE6khdChOaL7mZvHQtVY&amp;expires=1787832606 245w, https://cms.aihub-oost.nl/uploads/small_to1_9772ccfbe7.jpg?token=kete43kEkPaH5lgV3Z8RcG2d4E5HV7xX_mkQytfBp2o&amp;expires=1787832606 500w, https://cms.aihub-oost.nl/uploads/medium_to1_9772ccfbe7.jpg?token=6YuUXkA8xi41qr68Yt_qtMSX3v52mkD79GOHog6Ndu0&amp;expires=1787832606 750w, https://cms.aihub-oost.nl/uploads/large_to1_9772ccfbe7.jpg?token=tLEB9uYUDpJ2W2gNGGNxRW7-nIs7uNK0LmlRnEeKNJo&amp;expires=1787832606 1000w" sizes="100vw" width="1000"></figure><h2>Fewer Woo requests by providing citizens with faster responses</h2><p>The Ministry of Justice and Security is also a client of Thoughtful Oasis. VraagMijnOverheid is running a pilot there related to the Open Government Act (Woo), which allows citizens to request government information. Many citizens submit Woo requests because they do not know where public information is already available. Processing these requests takes up a significant amount of time and resources for government organizations.</p><p>VraagMijnOverheid tackles the problem at the source. Citizens ask their questions in plain language, after which the system immediately directs them to the answer if the information is already public. As a result, in many cases, there is no longer a need to file a formal Woo request. “The pilot has been open to citizens since July 15. Since then, 51 chats have been initiated, resulting in 12 search reports being downloaded, and three submitted information requests have been received. An information request is more advantageous for the ministry because they can be processed more quickly than official Woo requests,” explains Dommerholt.</p><p>The system offers benefits for both parties; citizens receive faster responses in understandable language, and the government avoids unnecessary administrative procedures. At the same time, the solution contributes to greater transparency and accessibility of government information.</p><h2>AI as a Support Tool</h2><p>For Thoughtful Oasis, AI is not about replacing human expertise, but rather about supporting it. By making decisions explainable, subject matter experts remain involved and can assess, verify, or challenge AI recommendations. This not only increases trust in the technology but also simplifies its implementation within organizations.</p><p>Explainable AI will play an increasingly important role as artificial intelligence is used more frequently in critical processes. Not only because regulations require it, but mainly because professionals want to understand the basis for a recommendation before acting on it.</p><h2>Practical Tips for Entrepreneurs</h2><ul><li data-list-item-id="e2f045035c2959ae3272aeed179c3a6bb">Involve subject matter experts from day one. They determine which knowledge and concepts are truly relevant to real-world applications.<br>Start with a single, specific problem. A well-defined pilot project yields insights faster than a broad implementation.<br>Think ahead. Design solutions so that new AI models can be easily integrated later on.<br>View AI as a support for people. The greatest added value arises when technology and human expertise reinforce each other.</li></ul><p><br>Would you like to get started with AI? Then please contact us.</p>

Thoughtful Oasis makes AI understandable and controllable for the public sector

Meet our frontrunners

Thoughtful Oasis.jpg

<p>Most applications of artificial intelligence are impressive, but they have one major drawback: no one can explain exactly how they arrive at an answer. For many applications, that’s not a problem. But when AI is used in decision-making regarding safety, emergency response, or government processes, transparency is essential. That’s why the Enschede-based startup Thoughtful Oasis is developing a new generation of AI that not only provides smart answers but also explains how those answers are arrived at.</p><p>With this approach, the young company is now collaborating with organizations such as the North and East Gelderland Safety Region (VNOG) and the Ministry of Justice and Security, where the first applications are currently in the pilot phase.</p>