Norman Deschauwer

Norman Deschauwer's public voice over the past year reads as a running experiment in applying agile thinking to the arrival of AI in the workplace, conducted in real time and largely in public. The throughline across his posts is that adoption numbers for AI-written code are already staggering and moving fast, and that most managers have no idea where their own organization stands. He cites the range he saw "en préparant ma session vibecoding à la Journée agile" — "environ 90 % chez Anthropic, environ 30 % chez Salesforce" — and contrasts it with Arthur Mensch's claim that Mistral engineers "n'écrivaient plus une ligne de code" . When he polled a room of practitioners on whether 30% of their own company's code was machine-written, "très peu de mains" went up, and he draws the blunt conclusion: "soit on ne sait pas, soit on a peur de dire oui" . For Deschauwer this isn't a curiosity, it's a management blind spot, and his recurring call to action is simply to measure it: "il est temps d'en parler" .

He is careful, though, not to let this read as uncritical AI boosterism. Drawing on his podcast conversation with Olivier Vergeynst, he pushes back hard on the tidy narrative that AI handling "les cas simples" frees humans to grow into "les cas complexes." He lays out the downside in detail: juniors skip the learning terrain that easy cases provide and land directly on senior-level problems they're unprepared for, while seniors, no longer exercising their skills on routine work, get rusty too — "on a une ba[isse]" across both ends of the skill spectrum . This is characteristic of how he handles guest ideas: he doesn't just relay them, he turns them into a checklist or a mirror question for the reader. The same move shows up when he discusses his own coding practice, admitting he started "comme un débutant" doing copy-paste into ChatGPT before learning to delegate "par paliers" rather than all at once. His operating rules are concrete and quotable: "Sheet in, sheet out," "coder par intention, pas par instruction," and the reminder that "l'IA invente parfois des bugs qui n'existent pas" so you always have to test for real . His conclusion is personal and slightly wry: "mon boulot n'est plus d'être développeur. Et ça me va" , and he frames delegation not as a switch but as "un curseur" you move deliberately .

Deschauwer treats AI tooling itself as something to interrogate rather than accept at face value. Covering Anthropic's Fable 5 release, he's attentive to the marketing language around "moins d'interventions" and asks directly whether that's "un vrai gain de productivité, ou juste un argument marketing," while flagging the practical cost that "Fable 5 mange votre quota plus vite qu'Opus 4.8" . He follows the same skeptical-but-engaged posture when Anthropic opened access to the "Mythos"-class model with automatic safety switching to Opus 4.8 for risky queries, noting the company's own admission that "des requêtes légitimes... seront bloquées à tort" . Elsewhere he insists that meaningful AI use isn't about volume: "la maturité, ce n'est pas «utiliser plus d'IA». C'est utiliser juste ce qu'il faut," pointing to frugal, lower-token tools as a deliberate choice rather than a compromise . He also values AI's capacity to restore human connection over sheer capability, quoting a guest's line about a voice tool that is "pas juste un synthétiseur de voix, c'est un lien" .

Beyond AI specifically, Deschauwer's agile convictions center on authenticity over performance. He's blunt that "l'agilité n'est pas morte," but that what actually dies is companies using agile as camouflage, "utilisée comme une rustine sur un pneu crevé ou... pour cacher les merdes sous le tapis" . He repeatedly frames failure diagnosis as a matter of culture, not method: whether rituals changed but not mentality, whether people can speak up without fear, and whether teams treat causes or just "collent un post-it dessus" . This same insistence on substance over metrics-theater appears in his summary of a conversation about KPIs and OKRs, where he warns against "l'indicateur pastèque," green outside and red inside, and lands on the line that "ce n'est pas un problème d'outil, c'est un problème de mindset" — measurement should serve learning, not judgment .

A related, recurring theme is that transformation starts with the self, not the system. He turns a colleague's provocative workshop title, "Suis-je un équipier idéal?", into a personal checklist about whether he himself proposes direction, speaks first, and takes autonomy rather than merely demanding it . He extends this into a broader philosophy of change borrowed from guests: "la plus grande chose qui m'empêchait de me réaliser pleinement, c'était moi-même," paired with the idea that freedom "vient avec le choix, et le choix vient avec l'espoir," and that real change is "un travail d'une vie" that begins with an uncomfortable but liberating realization . In a similar register, he stresses returning to fundamentals before adding complexity, citing a guest's advice to work on "l'intention d'abord" since "le processus suit," and to "reconsolider" foundations rather than build another floor on sand .

Finally, Deschauwer treats entrepreneurial risk-taking and imperfect starts as validated strategies rather than cautionary tales. He credits a young, self-taught entrepreneur's rejection of conventional wisdom, summarizing lessons like shipping before finding a business model and being your own first customer, because "je l'utilise, ça marche" convinces more than a pitch . He applies the same lesson to his own history, recalling that his flagship event began with only 32 people and a moment of panic, resolved by a co-founder's reassurance that "c'est exactement comme ça que ça commence" . From that experience he distills durable, practical takeaways: nobody is ever ready, so you start anyway; heartfelt execution beats polish, since "c'est fait avec le cœur" was the most common feedback he received; and doubt never fully disappears, it just gets outrun year after year . Across all these threads, from AI adoption to team culture to entrepreneurship, his consistent method is the same: take a guest's idea, test it against his own practice, and hand the reader a short, usable checklist rather than an abstract lesson.