Nine AI startups, one final: from smart promises to real value
On 12 November, the finalists take the stage at Evoluon. Their challenge: to show why their technology deserves a place in everyday life.
Published on November 2, 2026
Bart, co-founder of Media52 and Professor of Journalism oversees IO+, events, and Laio. A journalist at heart, he keeps writing as many stories as possible.
A baggage handler lifting hundreds of suitcases every day. A family caregiver looking for answers in the middle of the night. A factory collecting operational data but struggling to turn it into improvements. Behind the entrants in the AI Pitch Competition are concrete problems that artificial intelligence could help solve. On 12 November, nine startups must convince the jury of their approach during the final at AI Summit Brainport, held at the Evoluon in Eindhoven.
Over the past few weeks, IO+ has profiled all nine participants. Together, they demonstrate the breadth of AI entrepreneurship emerging in Brabant. Some teams are developing physical machines, others business software or digital guidance. Yet the same questions keep coming up: does the solution work outside a test environment, does the user retain control, and is there a market willing to pay for it?
The pitch starts with the problem
During the recent How to Pitch workshop at the Koning Willem II Stadium, participants practised distilling their stories to the essentials. In thirty seconds, they had to explain what their company does and why it matters. Coaches helped them make complex technology understandable and articulate their distinctive value more clearly.
The warning from BOM venture developer Guido Gageldonk applies to every team: a working pilot does not automatically mean customers will pay. And, as Liesje Goldschmidt of Erasmus Enterprise emphasised, a pitch's value extends beyond the grand prize. Meeting a customer, investor or partner could mark the beginning of the next stage of growth.
That is the challenge these nine participants are taking into the final.
Rigidbody: turning factory data into better processes
Many medium-sized factories have years of operational data but lack the specialists to use it. Rigidbody wants to make process optimisation accessible to these businesses. Founded by three TU/e engineers, the startup combines time-series models that incorporate physical principles with locally deployed language models that help interpret documentation and equipment codes.
The aim is to give operators clear recommendations to improve processes, potentially reducing energy and material consumption. The team is exploring the semiconductor supply chain as its initial market. As it prepares to grow, it faces a clear challenge: earning trust on the factory floor and demonstrating that each new customer can be onboarded without months of bespoke work.
UNSEQ: doing the heavy lifting in aircraft holds
A soft sports bag requires a different grip from a hard-shell suitcase. In a cramped aircraft hold, under time pressure, that variation makes automation particularly difficult. UNSEQ is developing robots that use vision, language and action models to handle these unpredictable objects.
For co-founder and CEO Robert de Bruijne, the value starts with the baggage handler: “A flight should not come at the expense of someone’s back.” The ambition is a system that safely takes over heavy lifting and fits into day-to-day ground handling. Bringing it to market requires reliable operation throughout an entire shift, backed by maintenance, support and trained human operators.
One Horizon: automating while retaining control
Businesses entrusting processes to AI want to decide which models are used, where their data stays and when a human intervenes. One Horizon is building an independent platform that brings together AI assistants, workflows and human oversight.
Founders Gijs van de Nieuwegiessen and Tijn van Daelen previously built Flow.ai. With One Horizon, they want to help organisations automate without becoming dependent on a single model provider. AI can also run on their own hardware. Specialist partners will help bring the platform to different sectors. The pitch addresses a familiar business need: benefiting from AI while retaining control over critical processes.
YourNeuroTune: background music with a different business model
For gyms, cafés and shops, music is part of the experience, but the associated licensing fees are a recurring expense. YourNeuroTune develops functional, AI-generated music as an alternative to traditional repertoires.
The startup combines generative audio with its own musical similarity checks and human quality assessments. It aims to deliver suitable music while identifying potential risks involving existing works. According to co-founder Sam Jansen, the company now serves more than 450 customers in four countries. YourNeuroTune can therefore also demonstrate commercial progress in the final. Further growth will depend on legal due diligence in each market and greater awareness among business owners.
Kyvvu: setting boundaries for AI that takes action
An AI agent may be allowed to read customer data. The same agent might also be allowed to send an email. But should it always be permitted to perform those actions in immediate succession? Kyvvu focuses on the risks arising from the combination and sequence of actions that are individually authorised.
The startup is building an Agent Security Kernel: software that checks proposed actions against fixed rules before execution, taking the task’s earlier steps into account. This check does not rely on a second language model acting as referee. Co-founder and CTO Maurits Kaptein aims to add a fundamental security layer to autonomous software. As agents gain more permissions, such boundaries become increasingly important.
Be-Inc: a coach for everyday communication
Company culture also takes shape in job advertisements, performance reviews and the next email to a colleague. Be-Inc, founded by Eline Schoonen, is therefore bringing a language and culture coach into Microsoft 365. It helps users recognise unintended exclusion, ambiguity and inappropriate tone in their writing.
The system provides explanations and alternatives, leaving the writer to decide. Be-Inc does not make automatic corrections or assign individual scores to employees. Organisational context helps keep its suggestions relevant. Under the banner of ‘Work EQ’, the company focuses on the human side of the digital workplace: improving communication as the words are being written.
Offensys: continuously testing digital defences
When attackers adapt their methods ever faster, periodic security tests quickly become outdated. Offensys is developing a platform that helps organisations continuously test their defences against realistic attack scenarios.
The startup combines knowledge of attack techniques with information about the customer’s environment. AI helps turn current threat intelligence into executable test scenarios. A next development step is an autonomous copilot that selects, runs and interprets simulations. Co-founder Cor Verhoeven emphasises explainability and human control. The company wants to help security teams keep pace with attackers through controlled testing in permissioned environments.
Synthgen: training for the rare exception
The very anomalies an inspection system needs to recognise can be extremely rare. Think of a crack in a weld or the early signs of a crop disease. Without enough examples, training an AI model to detect them is difficult. Synthgen therefore creates synthetic images of these exceptions, complete with automatic labels.
Using customer images, the startup generates additional variations to improve existing inspection models. Reality remains the benchmark: generated training data must deliver better performance on real images held back from training. CTO and co-founder Martin Tomov sees opportunities in European manufacturing and agriculture. The commercial challenge is to earn the trust of customers who have already heard plenty of AI promises.
Alzheimer.app: support when the question arises
Wildcard entrant Alzheimer.app brings everyday dementia care to the stage. Peter Spijkers’ initiative combines a language model with dementia-specific knowledge and personal context to support families and care professionals when questions arise, including outside office hours.
Its connection to the residential care facility De Gelukshoeve gives the team direct access to practical experience. At the same time, its boundaries are crucial: the app does not diagnose conditions or replace a care professional. When there is uncertainty or a potentially urgent situation, it should recommend appropriate human help. Growth requires more than translation: care practices, family relationships and cultural expectations also vary between countries.
Who will make the move to sustained adoption?
The nine stories reveal different stages of development: from preparing industrial pilots to running a music service with hundreds of customers. That makes the final compelling. Technical depth, societal significance and commercial progress will each take a different form on stage.
The profiles also reveal a shared entrepreneurial challenge. A factory must trust a young supplier, an employer must be willing to introduce a new communication tool, and a care organisation must find room for sustained adoption. That requires capital, early customers and evidence from practice.
On 12 November, the participants have an opportunity to bring that next step closer. Their task is as easy to describe as it is difficult to execute: make clear which problem you solve, show why your approach works, and give the audience a reason to keep talking after the pitch.
The AI Pitch Competition is organised by Erasmus Enterprise, BOM and IO+, with support from the Province of Noord-Brabant. The final takes place on 12 November 2026 during AI Summit Brainport at the Evoluon in Eindhoven.