<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Benjamin Maillard — Articles (EN)</title>
    <link>https://benjamin-maillard.fr/?lang=en#blog</link>
    <description>Articles on AI, agentic architecture, and field experience.</description>
    <language>en-GB</language>
    <lastBuildDate>Mon, 15 Jun 2026 21:30:21 +0000</lastBuildDate>
    <atom:link href="https://benjamin-maillard.fr/feed-en.xml" rel="self" type="application/rss+xml"/>
    <atom:link href="https://benjamin-maillard.fr/feed.xml" rel="alternate" type="application/rss+xml" title="Benjamin Maillard — Articles (FR)"/>
    <item>
      <title>Generative AI for DataViz</title>
      <link>https://benjamin-maillard.fr/?article=ia-generative-dataviz-bi-traditionnel&lang=en</link>
      <guid isPermaLink="true">https://benjamin-maillard.fr/?article=ia-generative-dataviz-bi-traditionnel&lang=en</guid>
      <pubDate>Mon, 01 Jun 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[Can generative AI replace or outperform traditional BI tools like Tableau, Power BI, or Looker for data visualization? Field experience with OenoTrac: from the "tool" BI paradigm to "intent" BI, and a new split of roles between business and data experts.]]></description>
      <category>AI &amp; Data</category>
    </item>
    <item>
      <title>Think fast, reason slow: a two-level agentic approach</title>
      <link>https://benjamin-maillard.fr/?article=systeme-1-systeme-2-architecture-agentique-deux-niveaux&lang=en</link>
      <guid isPermaLink="true">https://benjamin-maillard.fr/?article=systeme-1-systeme-2-architecture-agentique-deux-niveaux&lang=en</guid>
      <pubDate>Wed, 20 May 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[Building OenoTrac led me to formalize an architecture pattern inspired by Kahneman's System 1 / System 2. A fast agent that converses, an expert agent that arbitrates, shared memory that connects them. When to use it, when to avoid it, how to operationalize it, and why it complements rather than replaces native "reasoning" models.]]></description>
      <category>AI &amp; Architecture</category>
    </item>
    <item>
      <title>&quot;Parole&quot; / Émile — when the voice agent becomes the natural interface of a business agent</title>
      <link>https://benjamin-maillard.fr/?article=parole-emile-agent-vocal-interface-agent-metier&lang=en</link>
      <guid isPermaLink="true">https://benjamin-maillard.fr/?article=parole-emile-agent-vocal-interface-agent-metier&lang=en</guid>
      <pubDate>Mon, 18 May 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[Field experience adding a real-time voice agent to OenoTrac. Why voice is not just audio output, but a conversational layer able to explore the cellar, keep context, propose a V3 analysis, and carry the decision forward in the conversation.]]></description>
      <category>AI &amp; Architecture</category>
    </item>
    <item>
      <title>From Agent V1 to agentic sommelier V3: making AI reason like a professional</title>
      <link>https://benjamin-maillard.fr/?article=agent-v1-sommelier-agentique-v3-raisonner-ia-professionnel&lang=en</link>
      <guid isPermaLink="true">https://benjamin-maillard.fr/?article=agent-v1-sommelier-agentique-v3-raisonner-ia-professionnel&lang=en</guid>
      <pubDate>Fri, 08 May 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[Field experience on the evolution of the OenoTrac sommelier: from an enriched prompt to an agentic architecture that inspects the cellar, uses the right tools, compares several options, validates candidates, then extends to a voice agent that dialogues with Sommelier V3.]]></description>
      <category>AI &amp; Architecture</category>
    </item>
    <item>
      <title>From chatbot to shared operational graph: building a real AI agent system</title>
      <link>https://benjamin-maillard.fr/?article=chatbot-graphe-operationnel-systeme-agents-ia&lang=en</link>
      <guid isPermaLink="true">https://benjamin-maillard.fr/?article=chatbot-graphe-operationnel-systeme-agents-ia&lang=en</guid>
      <pubDate>Sat, 28 Mar 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[Field experience with MindMapCompanion. Why chat-driven agent systems are structurally limited, and how a shared operational graph (persistent state + Markdown agents + closed loop) changes the game, from the individual workstation to IT governance.]]></description>
      <category>AI &amp; Architecture</category>
      <enclosure url="https://benjamin-maillard.fr/mindmap-companion-interface.png" type="image/jpeg"/>
    </item>
    <item>
      <title>Coding complex algorithms with AI</title>
      <link>https://benjamin-maillard.fr/?article=coder-algorithmes-complexes-ia-simulateur-retraite&lang=en</link>
      <guid isPermaLink="true">https://benjamin-maillard.fr/?article=coder-algorithmes-complexes-ia-simulateur-retraite&lang=en</guid>
      <pubDate>Sat, 14 Feb 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[Can you really code complex mathematical algorithms with AI without being a domain expert? Field experience building LongView, an algorithmic retirement simulator: Monte Carlo, correlations (Cholesky), dichotomy optimization, and what AI truly brings when translating mathematical concepts into production code.]]></description>
      <category>AI &amp; Algorithms</category>
    </item>
    <item>
      <title>Agentic AI Sommelier: Mono-Agent vs Multi-Agent Approach</title>
      <link>https://benjamin-maillard.fr/?article=creation-sommelier-ia-agentic-ai-mono-agent-multi-agents&lang=en</link>
      <guid isPermaLink="true">https://benjamin-maillard.fr/?article=creation-sommelier-ia-agentic-ai-mono-agent-multi-agents&lang=en</guid>
      <pubDate>Thu, 05 Feb 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[Field experience building the AI Sommelier for OenoTrac. Practical comparison between a mono-agent approach with dynamic prompts and a multi-agent orchestration approach. Analysis of trade-offs, latency, answer quality, and lessons for choosing the right architecture depending on usage context.]]></description>
      <category>AI &amp; Architecture</category>
    </item>
    <item>
      <title>Running a security code audit with AI: field experience</title>
      <link>https://benjamin-maillard.fr/?article=audit-code-securite-ia-retour-experience&lang=en</link>
      <guid isPermaLink="true">https://benjamin-maillard.fr/?article=audit-code-securite-ia-retour-experience&lang=en</guid>
      <pubDate>Wed, 14 Jan 2026 00:00:00 +0000</pubDate>
      <description><![CDATA[Field experience running a full security audit of the OenoTrac application with AI (Cursor + Claude). Discovery and remediation of 3 critical vulnerabilities (JWT secret, localStorage, refresh tokens), moving from a "medium" to "good" baseline on OWASP basics, for ~€10–20 of API tokens and one week of work.]]></description>
      <category>Security</category>
    </item>
    <item>
      <title>Modern scalable web architecture with integrated AI: field experience</title>
      <link>https://benjamin-maillard.fr/?article=architecture-web-moderne-ia-retour-experience&lang=en</link>
      <guid isPermaLink="true">https://benjamin-maillard.fr/?article=architecture-web-moderne-ia-retour-experience&lang=en</guid>
      <pubDate>Sat, 20 Dec 2025 00:00:00 +0000</pubDate>
      <description><![CDATA[Technical and strategic details on implementing a three-tier architecture (React, Node.js/Hono, Python/FastAPI) with a multi-agent AI service. Technology choices, AI cost management, resilience patterns, and concrete lessons for executive committees and technical leaders.]]></description>
      <category>AI &amp; Architecture</category>
    </item>
    <item>
      <title>Like Saint Thomas, I only believe what I code</title>
      <link>https://benjamin-maillard.fr/?article=comme-saint-thomas-je-ne-crois-que-ce-que-je-code&lang=en</link>
      <guid isPermaLink="true">https://benjamin-maillard.fr/?article=comme-saint-thomas-je-ne-crois-que-ce-que-je-code&lang=en</guid>
      <pubDate>Fri, 19 Dec 2025 00:00:00 +0000</pubDate>
      <description><![CDATA[Introduction to a series of articles on concrete experimentation with AI and modern web architectures. A technical leader deliberately returns to hands-on work to test what AI can actually produce, decide, and create in terms of value.]]></description>
      <category>AI &amp; Architecture</category>
    </item>
  </channel>
</rss>
