What Luxembourg's small teams told us about adopting AI.

Introduction

The Luxembourg government has dedicated a huge amount of resources to the promotion and development of AI access and skills in Luxembourg (RTL Today, 2025; Luxembourg Times, 2026), including the most well-known version of this concept, generative AI. This includes more than 226 million euros committed to AI in Luxembourg, some cofunded with the EuroHPC Joint Undertaking (Paperjam, 2025).

This serious commitment to sovereign compute on Luxembourg soil is the right one. However, an understanding of who is using AI effectively and how accessible it is amongst teams of fewer than 20 people has not yet been undertaken either independently or by a private industry organization. Crystallized Intelligence aims to remedy this through conducting this survey.

With this survey, we will examine in public a picture of how small teams use AI in Luxembourg, including solopreneurs, small businesses, and NGOs (also known locally as ASBLs). As a disclosure, Crystallized Intelligence sells AI services to the types of teams it describes. In this report, we will attempt to follow the data, and share transparently where our point of view as a business may influence its interpretation.

We are providing this data freely to the public with the goal of improving AI’s effectiveness overall for local businesses, from your corner store bakery to the former big-4-turned-solo-consultant, and from the human rights educator at an NGO to the next Luxembourg unicorn startup. The backbone of Luxembourg’s economy is supported by locally-owned businesses, and it’s worth its own dataset.

Getting small teams to talk about AI is hard, partly because of the uncertainty around what “counts” as AI. But getting to the answer of “no AI use at all” at an organization is interesting for its own reasons. As ubiquitous as it is, knowing who isn’t using AI, or who is using it less impactfully than others, and why, can help all of us navigate the coming months and years as this technology continues to develop and impact our day to day lives, inside and outside of work.

In addition to the business case around using AI for small teams, this survey aims to measure the sentiment around AI in society. It’s becoming more apparent globally that AI is causing uncertainty in employment, environmental protection, privacy, and other aspects of life that are essential to a free and healthy society. If we only look at posts on LinkedIn, or only listen to the leaders on stage at tech conferences, we may be missing out on a fuller picture around AI’s impact on society, and hopefully, this small dataset can be absorbed as one of many points taken into consideration for policy and real-life AI practices and investments of time and money by employers and employees, independents, civil servants, activists, and politicians.

Methodology

The survey was administered via the website crystallized.lu during the period of 6 June 2026 to 31 July 2026 for teams of 1-20 people in Luxembourg. Through location- and company-size-targeted LinkedIn advertising, posts on WhatsApp community groups, and emails and private messages to various solopreneurs and small team members, a dataset of 25 anonymous responses was gathered across 7 sectors:

• Professional services (legal, accounting, consulting)
• Tech / software
• Creative / marketing
• NGO / non-profit
• Retail / hospitality
• Healthcare
• Education

“Financial” was also an option, but there were 0 respondents, likely due to the typical team size in Finance in Luxembourg, which according to the CSSF, represents 52,146 staff across only 669 institutions (Commission de Surveillance du Secteur Financier [CSSF], 2026), for an average of ~78 people per institution, though some will likely be much higher or much lower. While data on how AI impacts the Financial sector of Luxembourg is likely very interesting for the government and other actors, it doesn’t represent the target concept of this research, which is to better understand the impact of AI for smaller businesses, NGOs, and solopreneurs.

Finally, for future surveys we have identified the following refinements:

• Separating Civil Societies from Open Source Communities which overlap but are not understood the same way by all people
• To determine what type of data is being entered into the referenced system. Despite its anonymity, the decision was made to avoid this type of question which could potentially open respondents to probing on data-privacy-related issues. For the first time publishing this survey, it was decided to establish trust by proving the respondents’ data could be handled in a sensitive manner.
• To dive deeper into AI tooling such as models vs types of technology (coworking tools, coding CLIs, or other applications)
• To ask further questions about awareness of Luxembourg's funding for SME AI projects
• To better clarify uncertainty on the time saved question: do respondents feel they haven't saved time at all, or are they simply unable to estimate a positive number?

Change log

• 1.2, 16 September 2026: headline and report title name Luxembourg. No figures changed.
• 1.1, 14 September 2026: corrected summary for 'Who responded' to name the relevant sectors. No figures changed.
• 1.0, 14 September 2026: first publication.

Challenges and Next Steps

The Luxembourg government clearly sees the value of AI for our economy. However, most of the research thus far has focused on larger companies, leaving a gap of understanding for an important pillar, even a backbone of our economy: small teams.

The interesting aspect of small teams is how locally embedded they are in their own communities: these are people who live here, plan to stay here, and have data on local contacts, local culture, and the “secret sauce” to doing business in Luxembourg. A frequent complaint by those who try to set up shop in Luxembourg is lack of understanding on its bureaucracy, getting a business license, establishing a bank account, and other essential tasks.

A major risk if sovereignty is not established is not just that the data held by local residents may be exploited, but also their methodologies. Foreign AI hyperscalers are actively training on not just the content being shared with their tools, but how users are actually accomplishing their tasks, and this drives their own product development.

While there are promising signs that those who are aware of the topic of data residency are therefore choosing GDPR-friendly tools, those who are not informed are much less likely to consider this element, which is a must for all EU businesses, including here in Luxembourg. Better information on this topic for business owners will be essential to navigating the future of Luxembourg’s economy if we are to protect our own knowledge and unique advantages.

Luxembourg continues to expand its relationship with Mistral (AFP & Mouzon, 2026), but Mistral is not the first AI model provider on most respondents’ list. It may be worth examining why, and seeing how to improve the uptake on this trusted French provider.

For businesses hoping to make an impact providing AI services or consulting, looking at the current AI budgets of small teams is also an area to consider. Right now, they don’t spend a lot on tooling. However, tooling in the age of AI is cheap and likely to stay that way by using open weight models instead of hyperscaler ones. Open source projects connected with AI models can deliver a huge amount of value. The existing Fit4AI and SME Packages – AI are a good fit for helping reduce sticker shock to small teams.

In terms of impacts on jobs, this survey skews positive. There are more growing faster than those who have paused hiring, and the majority have not yet perceived an impact. Keeping this momentum, if it is reflective of the greater population, is an important element of our economic strategy to protect local employment and diversity of employment opportunities.

Acknowledgements

As the researcher and author of this project, I would like to extend my heartfelt thanks to my network in Luxembourg for helping promote this survey in their own networks as well as those who gave me their time and responded both anonymously and with helpful feedback.

In particular, I would like to thank my friend Nora who proofread the French version of the questions before they were published.

As part of this project, 2€ to for each survey response was pledged to Stëmm vun der Strooss. As the number of survey responses was limited to 25, this pledge was increased 2x and 100€ was donated. I would like to extend a heartfelt appreciation to Stëmm vun der Strooss for agreeing to be named as the benefactor for this charitable aim, and for their tireless work supporting the most vulnerable members of Luxembourgish society.

Finally, thank you to the country of Luxembourg for becoming my home, like many others who were not born here. I created Crystallized Intelligence with help from several institutions in Luxembourg, with the goal to help the Luxembourgish business and NGO communities benefit from the technological marvel that is generative AI, as well as lessening its risks and drawbacks. To that end, I also thank my husband for bringing me to his wonderful home country and supporting my work.