Den Atelier: using AI to better anticipate concert success

Founded in the mid-1990s, den Atelier has established itself as one of the key players in concert organisation in Luxembourg. Behind this well-known name, two complementary entities - L’Atelier and A Promotions - operate with a total team of 11 people.

This deliberately small size is seen as a strength. “We are a bit like cowboys and we are not organised like established companies,” says Michel Welter, one of the structure’s partners. “We keep things simple, in the best sense of the word. That has enabled us to remain small and navigate crises such as Covid with agility.”

With limited fixed costs and a flexible organisation, this model has demonstrated its resilience. But in a sector as unpredictable as live entertainment, with a huge number of parameters to consider, some of them entirely random, survival is not enough. You need to be able to anticipate.

Estimating demand: a major financial challenge for concert organisers

Concert organisation is based on a financial equation that is particularly exposed to risk. “At least 70% of the costs are linked to artists’ fees,” explains Michel Welter. “The venue, communication and staff come afterwards.” This risk falls entirely on the organiser’s shoulders. Artists, meanwhile, receive a guaranteed minimum fee, together with a share of the profits. They do not bear the commercial risk.

Everything is therefore decided upstream, during negotiations with the artist’s agent. The organiser must make an offer — including the fee and expected venue capacity — while assessing the artist’s potential in their market.

Following its near-bankruptcy in 2015, the company introduced several management tools, including an internal database containing highly detailed information such as sales timelines, ticketing data and concert-by-concert results. This is a valuable tool, but it is incomplete. “It all remains fairly empirical, and we are missing the ‘trending’ dimension,” Michel Welter acknowledges. How can you know whether a young artist who has never performed in Luxembourg can fill a venue? How can you draw comparisons between two artists with similar styles but whose careers are several years apart? Internal data is no longer enough.

This raised the question of how to cross-reference den Atelier’s historical data with external signals — streams, Shazam searches, geolocated data and so on — to refine estimates of future demand. The aim is to assess demand artist by artist, market by market and season by season.

Fit 4 AI: a structured audit to identify artificial intelligence use cases

Michel Welter does not describe himself as a technology enthusiast. “I know absolutely nothing about it. I was actually rather sceptical about artificial intelligence to begin with. But I also know that it can be a very powerful tool. So we needed support.”

The relationship with Luxinnovation was gradually established through an introductory training course at the Chamber of Commerce, followed by support from the consultancy Résultance and recommendations from another well-known Luxembourg company.

The Fit 4 AI programme represented a structuring step. It is open to any company established in Luxembourg, regardless of its size or sector, and is based on the intervention of an external consultant approved by Luxinnovation. Its objective is straightforward: to identify the most relevant artificial intelligence use cases, assess data maturity for each of them and define a quantified action plan.

This leads to a detailed and costed action plan covering the prioritised next steps, enabling the company to deploy AI-based tools and solutions within its processes.

“We were required to analyse ourselves, which is something we never do.” The audit therefore highlighted deeply rooted habits that had never been questioned. “A concrete example is that we do not keep records of our estimates. We aim to break even on every concert, but we are unable to measure whether we did a ‘good’ or ‘bad’ job.” Every booking offer involves an internal discussion about the artist’s commercial potential. Without traceability, however, it is impossible to learn from mistake… or successes.

The support proved decisive in structuring the approach and completing the administrative steps associated with financing. Michel Welter, den Atelier

The Fit 4 AI programme also provided a better understanding of the value of the data held by the organisation and the data it lacks. “We now appreciate the value of information from a different perspective, whether it is shared or not shared. Too much information can be harmful: you need to know what to extract from each piece of information.”

The audit has now been completed. Specifications for the development of a proprietary AI tool were submitted to the Ministry of the Economy, which has recently issued a favourable co-financing opinion. At the same time, an internal chatbot has already been activated as the first building block in the team’s gradual integration of AI into its working practices.

A bespoke predictive AI tool

The planned solution will not be an off-the-shelf product, but a tool developed in-house. “We know that prediction models exist, but there is no ready-made solution adapted to our specific needs.” For this reason, den Atelier has opted for a tailor-made solution, powered by its own historical data and enhanced with external sources: geolocated streaming data, Shazam statistics and trend indicators.

The aim is to aggregate this information so that a predictive model can estimate future demand for a given artist in the Luxembourg market. The tool will support decision-making, improve the quality of booking offers and reduce the financial risk structurally borne by the promoter. Operational implementation of the tool is planned for the first quarter of 2027.

Structuring the approach with Luxinnovation’s support

The relationship with Luxinnovation was neither immediate nor linear, but it proved effective. “At the very beginning, it is not easy to understand who does what,” Michel Welter admits. Once the right path had been identified, however, “the support proved decisive in structuring the approach and completing the administrative steps associated with financing.”

The rigorous and analytical approach provided by the programme’s experts enabled den Atelier to take a fresh look at its own operations and identify opportunities for improvement that the team would not have detected on its own. “We learned things. Some habits that we thought we would never need to question, particularly when it came to data analysis.”

The financial stakes are real, even though Michel Welter prefers to remain cautious in his projections. “Such a decision-support tool could help us limit loss-making commitments and generate savings of several tens of thousands of euros per year,” he estimates. “In any case, the potential is there. But I always prefer to remain very cautious, or even pessimistic, about any estimates.”