How Dutch NLP and speech AI startups choose their market and generate revenue

A handful of Dutch startups sell language or speech technology as their primary product, not as a by-product of something else. They target a range of sectors: recruitment, media, retail, customer service. What they share is that they use large language models or speech processing as the core of their proposition.

Funding rounds vary considerably, from a seed of a few million to an acquisition worth hundreds of millions of euros. That difference also reflects the stage each company is at and the maturity of the market it serves. A comparison of eight companies illustrates how diverse the choices are in terms of technology, customer segment and revenue model.

Two extremes: from niche market to broad HR infrastructure

Textkernel, founded in 2001 in Amsterdam, is by far the oldest player in this overview. The company spent years building CV parsing and semantic matching for recruitment professionals and staffing agencies. In June 2024 it was acquired by US-based Bullhorn; the deal is recorded in the database at €300 million. Prior to that, Main Capital Partners had already made a strategic investment, in 2020. Textkernel grew partly through acquisitions: it took over Sovren in December 2021, Akyla in June 2022 and Joboti in March 2023.

The business model is B2B licensing to HR software vendors and large recruitment agencies. The technology is deeply embedded in third-party platforms, which makes the company less visible to end users but structurally anchored in the market.

At the other end of the spectrum is Vox AI, founded in October 2023. This Amsterdam-based company targets drive-through and quick-service restaurants with a voice AI platform that handles orders. The system supports more than 90 languages and adapts to accents and ambient noise. In August 2025, Vox AI raised a seed round of approximately $8.7 million, led by Headline with participation from True, Simon Capital and Souschef Ventures. The narrow focus, one sector, one use case, makes the proposition concrete, but also limits the addressable market at an early stage.

Media and entertainment as a growing outlet

Three companies focus on media production and content distribution. Dubformer, also based in Amsterdam and founded in 2023, offers AI dubbing: from script to broadcast-ready audio. It raised a seed round of €3.3 million in 2025. The target market is production companies and broadcasters that want to localise content without the full cost of traditional dubbing.

AmberScript, likewise Amsterdam-based and active since 2017, converts audio and video into text through transcription and subtitling. The company raised a Series A of €8.7 million in 2021. Its customers are media companies, legal service providers and government organisations that require accessibility or archiving solutions.

XS2Content from Hilversum, founded in 2020, takes a different angle: repurposing existing content and automatically distributing it via AI-driven workflows. The €18.4 million Series A2 round in 2023 indicates that investors see potential in making content archives more efficiently accessible, a challenge that is becoming increasingly pressing for broadcasters and publishers as content volumes grow while budgets do not rise proportionally.

Customer service and content creation: human oversight as a selling point

Chatbots.expert from Tilburg builds and deploys chatbots for customer service, primarily in the Benelux market. The company was founded in 2017 and, according to the database, raised a Series C of €92 million in 2021, a notably high figure for a relatively regional player. The revenue model is the implementation and management of conversational AI for companies that do not want to develop this in-house.

CopyRobin from Mijdrecht, active since 2015, combines AI-generated text with editorial review by humans. It positions that combination as a guarantee of brand consistency and quality, an argument that resonates with clients who consider fully automated output too risky. In 2024, the company closed a Series B of €15.6 million.

Textmetrics from Arnhem, founded in 2014, monitors text quality and AI usage within organisations in real time. It targets HR and internal communications with a SaaS subscription model. The Series A of €10.1 million was closed in 2025. The proposition is less focused on content production and more on ensuring consistency and adherence to writing guidelines.

Technology differentiation and questions of scale

Technology choices reflect the target market. Vox AI emphasises full autonomy without human intervention and multilingual capability as distinguishing factors for operational environments. Dubformer and AmberScript build on speech models specifically trained for professional audio output. CopyRobin and Textmetrics, by contrast, place human oversight at the centre of their model.

Textkernel is the only player in this overview to have been fully integrated into a larger international organisation. For the remaining companies, scale remains an open question. The B2B SaaS models of Textmetrics and AmberScript offer predictable recurring revenue, but require a substantial customer base to become relevant at European level. Niche players such as Vox AI and Dubformer are betting on sector depth rather than breadth.

In the broader European AI landscape, the question is whether specialised language and speech models can remain competitive alongside the large generic models from US providers. Dutch startups appear to be answering that question by opting for vertical focus, deep integration into work processes or a hybrid approach with human quality control. These are positions that generic platforms are less easily able to replicate, and which, as the acquisition of Textkernel illustrates, can be of interest to investors and strategic buyers alike.