The 15th edition of Big Data & AI was held in Paris on 15-16 September. This year, 149 exhibitors participated in the event including start-ups, companies specialised in developing AI architecture, major group companies – many opportunities for clients were given to grow their network and identify technical solutions for their business needs.
Introduction
The 15th edition of Big Data & AI was held in Paris on 15-16 September. This year, 149 exhibitors participated in the event including start-ups, companies specialised in developing AI architecture, major group companies – many opportunities for clients were given to grow their network and identify technical solutions for their business needs.
The state of AI market in 2026
Nearly four years after the launch of ChatGPT, AI remains a top priority across every industry, from insurance and cosmetics to travel.
Companies now clearly understand the opportunity AI represents. To seize it, they have launched a wave of IT projects to bring AI into their core systems, investing heavily in data quality and robust platforms. That is why many of them are looking to leverage external expertise from IT services firms to help them accelerate their data and AI initiatives while bridging internal skill gaps.
From Proof of Concept to Scale: The Theme of Big Data & AI Paris 2026
The organiser RX France through the voice of Emilie Pierre-Desmonde, Big Data & AI Paris Director announced the main theme of the event: How to support companies in succeeding in going beyond POC's stage (Proof of Concept).
In the interview with French media Alliancy, she discusses the challenges companies face when moving AI from tests to large-scale deployment. In addition, the event's pursuing its vow to help clients to go forward in their reflection to support AI transformation with a clear mindset.
To that end, the conference programme was organised around three tracks: "Modern Data Architecture", "Governance & Regulation" and "Data & AI for Business". Key topics covered in the keynotes and panel discussions included digital dependencies, preparing organisations for agentic AI, AI systems security, and scaling AI.
#1 key takeaway: Moving on from POC to AI scaling
While navigating between the alleys of the event, I noticed one thing that I confirmed what I was already seeing in previous talks and discussions. AI is not anymore in a testing phase.
Indeed, most companies, if not all, want to integrate AI technology in their internal processes. They have understood the value AI could create, so their employees may focus on more valuable tasks rather than repetitive ones. Consequently, enterprises have reshaped their IT roadmaps around AI, now seen as a core long-term ambition.
#2 – User-Centric AI: Where Companies Now Look for Value
Furthermore, companies have the desire to build AI tools centred on user needs.
This means building solutions that are not one-off deliveries, but products that evolve alongside business needs.
Such an approach relies on continuous feedback from users, regular iterations and close collaboration between technical and business teams. Putting users at the centre also means involving them from the start
#3 – A data/AI platform is useless without business buy-in.
Finally, for me, the most important factor concerns the business teams themselves. Building a powerful, robust data and AI platform is a wasted investment if business teams do not get on board.
Bringing them to the table from the outset ensures that technical solutions include features that meet their real needs and fit into their day-to-day work.
As a result, this approach builds trust and ownership. When people have helped shape a tool, they are far more likely to adopt it and suggest improvements on a daily basis.
Conclusion
Moving beyond the POC stage is as much a human challenge as a technical one. By building AI with and for the people who use it, companies can turn experimentation into lasting value and finally go beyond POC's stage.
Of course, it is a long process and certainly not the easiest one. According to the RAND Corporation (2024) and the Harvard Business Review (2023), more than 80 per cent of AI projects fail to achieve their initial objectives – twice the rate of traditional IT projects.