Since the arrival of LLMs and the race for benchmarks, am I the only one who has the impression of being constantly behind, of never managing to keep up with the frantic pace of advances? Every week, I come across new tools, models or frameworks that already seem more relevant than those I had chosen the day before. Enough to drown, with the impression of drinking water with each LinkedIn post. In this context, how to build a solid data strategy, without exhausting your team to develop a solution that the market, research or our peers will have already surpassed, or even rendered insignificant, before it is even put into production? Faced with this acceleration, two traps await organizations. The first: remaining locked in an infinite cycle of tests and POCs, fueled by the fear of missing "the right tool", aka FOMO. The second: planning nothing, passively waiting for a perfect solution to emerge elsewhere. In both cases, we lose sight of the essential: our role is to create value in the short, medium and long term without becoming obsolete. In this talk, I first propose to take a step back to get our heads above water. We will start with a historical review: what strategic approaches have worked in the past? What mistakes have made some companies obsolete? And above all, what, in this precise moment, fundamentally changes the game compared to previous technological waves? I will then share a series of concrete advice for building a data strategy in this changing context, illustrated by historical cases, examples of tech companies, and recent feedback: put into production quickly in small pieces and improve afterwards, build thinking that everything can be deleted tomorrow, always measure, channel monitoring. This talk is aimed at CDOs, CTOs, tech leads, managers, data strategists and all those who design or lead data/AI projects.