Montréal, Canada
17–18 June 2026
Wednesday to Thursday
08:00–17:00 local time (GMT-4)
Microsoft Montréal
2000 Avenue McGill Collège Montreal, QC H3A, Canada
Open in Google Maps (opens in a new tab)This event took place on 17–18 June 2026.
Ce qui rend AICD Montréal unique
Communautaire avant tout
L’événement est organisé par la communauté, pour la communauté, avec une approche pratique et ouverte.
Bilingue FR/EN
Montréal est une ville naturellement bilingue. Nous accueillons des sessions en français et en anglais afin de maximiser l’accessibilité et la richesse des échanges.
Responsable et concret
Nous valorisons les usages réels de l’IA: productivité, automatisation, gouvernance, sécurité, données, agents, copilotes, adoption et transformation des façons de travailler.
Ancré localement, connecté globalement
AICD Montréal s’inscrit dans un mouvement international de partage de connaissances en IA, tout en mettant en valeur les besoins, initiatives et talents de notre écosystème local.
Programme 2026: deux journées complémentaires
17 juin 2026 — Workshops (journée pratique)
Une journée de workshops pour aller en profondeur, manipuler des outils, tester des approches, et repartir avec des compétences concrètes.
18 juin 2026 — Conférence (journée communautaire)
Une journée de conférences, retours d’expérience, tables rondes et rencontres pour découvrir des idées, comparer les approches et connecter avec la communauté.
À quoi s’attendre
Conférences et sessions d’experts sur les tendances, méthodes et usages de l’IA
Workshops pratiques pour apprendre en faisant
Retours d’expérience terrain (succès, défis, erreurs, leçons apprises)
Sujets de gouvernance et d’IA responsable
Opportunités de réseautage avec pairs, experts et leaders de communauté
Mise en valeur de projets locaux et d’initiatives inspirantes
Pourquoi participer
Apprendre vite grâce à des contenus concrets et actionnables
Comparer les approches entre équipes techniques, métier et formation
Rencontrer la communauté IA/Microsoft de Montréal et au-delà
Développer ses compétences sur des sujets qui impactent déjà le quotidien professionnel
Contribuer à une communauté ouverte, inclusive et orientée partage
Langues
Les sessions peuvent être offertes en français ou en anglais. La langue de chaque session sera indiquée dans le programme.
Times in Montréal time (GMT-4)
9:00 AM
12:00 PM
1:00 PM
Times in Montréal time
This workshop is a full build journey, not a slide deck safari. In about 5 hours, participants go from architecture whiteboard to a deployed, production-style RAG agent running across Azure and Microsoft 365. The format is intentionally step by step: each module adds one real capability, so by the end you have a complete solution, not a pile of disconnected demos. Developers can implement the solution using either Node.js/TypeScript or Python. The workshop supports both paths, so participants can follow the stack that best matches their team skills and production context. 1. Module 1, plan before you code: participants map the end-to-end architecture, choose the right services, and make practical tradeoffs around model choice, cost, compliance, and enterprise constraints. Think of it as the part where we save everyone from expensive rework later. 2. Module 2, build the foundation: participants provision the Azure backbone, ingest data, and create the retrieval pipeline with indexing, chunking, embeddings, and vector search. This is where the magic stops being magic and becomes engineering. 3. Module 3, create the first working agent: participants configure an Azure AI Foundry agent, connect it to retrieval, and validate grounded responses. First real answers appear here, and yes, this is where people start smiling at their terminal. 4. Module 4, bring it to users: participants wire the agent into Teams and Copilot with the Microsoft 365 Agents SDK, test streaming and non-streaming interactions, and iterate in a developer workflow that feels like building a real product. 5. Module 5, make it secure and useful in the enterprise: participants add authentication and SSO, then connect to Microsoft 365 data through Retrieval API and Graph patterns, so responses are personalized, permission-aware, and business relevant. 6. Module 6, ship it like you mean it: participants package and deploy to Azure and Microsoft 365 channels, moving from local prototype to operational solution with a repeatable release path. By the end, developers leave with three things: a working reference implementation, a reusable step-by-step delivery pattern for future agents, and a much clearer sense of how to build AI systems that survive real-world enterprise conditions. In short, fewer science projects, more shipped solutions.
Organizations invest heavily in data platforms, yet many teams still struggle to turn data into timely decisions. Data pipelines prepare datasets and dashboards provide insights, but business users often need faster answers, contextual explanations, and automated actions. In this hands-on workshop, we’ll explore how Microsoft Fabric can bridge this gap by combining modern data engineering pipelines with Fabric IQ Data Agents. Using a realistic service-operations scenario based on public city service request data, we’ll walk through how raw operational data can be transformed into intelligent, queryable, and action-ready insights. Participants will build a complete solution step-by-step: -Ingest and transform raw operational data using Fabric Data Pipelines -Curate and model the data in a Lakehouse and semantic model for business-ready insights -Create a Fabric IQ Data Agent that can reason over enterprise data -Expose this intelligence through Copilot Studio and MS Foundry, enabling natural-language interaction, insights, and automated actions
Unlock the full potential of AI-driven automation in this hands-on workshop, where you’ll learn to build, deploy, and orchestrate AI agents using Copilot Studio and Microsoft AI Foundry. This session focuses on practical, real-world applications, empowering you to design intelligent agents that streamline workflows, enhance productivity, and integrate seamlessly with the Microsoft platform. The workshop will focus on building AI agents using Copilot Studio to drive automation, with guided exercises and interactive demos to explore how these agents can enhance real-time business workflows. You'll gain a deep understanding of how to configure intelligent actions, how Copilot boosts productivity, and how to scale AI solutions using the 5P Framework (Purpose, Pilot, Playbook, Production, Performance). Towards the end of the session, we will demo AI Foundry, highlighting its role in AI agent deployment and orchestration. By the end of the workshop, you’ll be equipped with the knowledge to drive AI adoption across your enterprise, with a focus on security, governance, and responsible AI practices. Key Takeaways: - Hands-on experience building and deploying AI agents using Copilot Studio and Microsoft AI Foundry - Configuring and orchestrating intelligent actions across the Microsoft platform for seamless automation - Insights into how Copilot can enhance productivity and support real-time automation (discussion-based, not a demo) - Mastering the 5P Framework (Purpose, Pilot, Playbook, Production, Performance) to scale AI agent adoption across enterprise environments - Best practices for responsible AI implementation within the Microsoft platform - How Copilot can optimize business workflows, driving smarter automation and efficient decision-making - AI Foundry demo towards the end, showing real-world deployment and orchestration Ideal for: - Developers and engineers eager to build AI agents - IT professionals and business leaders focused on automation and scaling AI solutions - Anyone looking to harness the power of Copilot Studio, AI Foundry, and the Microsoft platform to drive enterprise-wide productivity and innovation Join us to transform your understanding of AI automation and discover how Copilot Studio, the Microsoft platform, and AI Foundry can enhance real-time business workflows. This workshop will equip you with the expertise to design, deploy, and scale AI agents that are secure, efficient, and optimized for the future of intelligent business operations.
As Copilot Agents become central to enterprise workflows, prompt engineering is the key skill that separates helpful assistants from frustrating bots. This session dives deep into how Copilot Agents interpret prompts during a session, how to guide responses reliably, and how to design prompts that trigger the right actions at the right time.
Les solutions d’IA générative comme Microsoft Copilot offrent des opportunités de transformation majeures, mais leur déploiement nécessite des stratégies de sécurité robustes pour protéger les données et les systèmes de votre organisation. Lors de cette session, nous nous concentrerons sur la manière d’adopter un modèle Zero Trust afin de déployer Copilot et des solutions d’IA similaires de manière sécurisée au sein de votre système d’information. Je présenterai mon modèle personnalisé Zero Trust, qui place l’identité au cœur de la stratégie tout en abordant les principaux vecteurs d’attaque tels que la protection et la classification des données. Ce modèle garantit que vos données sont correctement sécurisées et structurées avant tout déploiement d’IA. Nous explorerons comment : • Mettre en œuvre des mesures de sécurité basées sur l’identité et l’accès conditionnel (Conditional Access). • Protéger les données sensibles grâce à une classification avancée des données et à une gouvernance efficace. • Exploiter les capacités de Microsoft Purview pour appliquer des politiques de protection des données et assurer la conformité. • Réduire les risques liés au déploiement de l’IA en utilisant des outils comme DLP pour Copilot, Insider Risk Manager et DSPM pour l’IA. Cette session vous fournira une feuille de route claire pour déployer des solutions d’IA en toute sécurité, en maximisant leur valeur tout en protégeant les actifs de votre organisation.
Do you feel like your company has too many AI agents? An HR agent, an IT agent, a “put_your_agent_name_here” agent… and probably a few more no one remembers creating? With how easy it has become to build agents in the Microsoft ecosystem—Declarative agents in Copilot Studio or pro-code, Agent Builder, SharePoint agents, and more—organizations naturally end up multiplying agents for very specific use cases. Unfortunately, this approach comes with a few not-so-small problems: • Agent sprawl leads to poor discoverability and monitoring, higher maintenance costs, and significant overlap or duplication—both in purpose and in the data those agents consume. • Technology choices are not always the right fit for the scenario. How many times have we seen a “quick proof of concept” quietly promoted to “production,” only to become a limitation later when more advanced requirements appear (hello, declarative agents)? But wait - what if you could rationalize all of these agents into a single experience, exposed through Teams or Copilot, while handling requests behind the scenes? And what if, at the same time, you could maximize your Copilot investment by leveraging Microsoft Graph Connectors? The idea is simple: one agent visible to users, backed by a scalable fleet of specialized agents, orchestrated behind the scenes to handle diverse, read‑and‑write tasks. The result? A shared technical platform for all your agents—dramatically simplifying governance, maintenance, monitoring, and many of the challenges introduced by agent sprawl in the first place. In this session, I’ll show how to design and implement an enterprise-grade, multi-agent system architecture using: • Microsoft Foundry Agents Service • Microsoft 365 Agents SDK • Copilot Connectors via the Copilot Retrieval API Drawing from real-world business implementations - and the "Agents League Contest" hackathon winning solution, “Whatever AI Assistant”—I’ll walk you through: • The overall architecture and implementation steps • Key design decisions and trade-offs • Common challenges you’ll face when building these solutions, including: ○ How do you test and evaluate your agents? ○ How do make secure from infrastructure to agent guardrails? ○ How do you track meaningful business metrics? ○ How do you manage versioning and deployments? ○ And more. If you’re serious about building enterprise-wide agent solutions with Microsoft Copilot, this is a session you won’t want to miss.
AI promises transformation, but organizations only see results when people actually change how they work. This session explores what it takes to build an AI‑ready culture using Microsoft 365 as your foundation. We’ll break down the strategic pillars of successful AI adoption, from leadership alignment to communication patterns that reduce fear and increase trust. You’ll see how SharePoint becomes the backbone of your AI knowledge ecosystem, how Copilot Create can rapidly produce tailored training videos that reinforce consistent messaging, and how to design adoption experiences that meet employees where they are. Whether you’re leading a rollout or shaping your organization’s AI strategy, you’ll walk away with practical frameworks, examples, and tools to turn AI excitement into everyday impact.
Generative AI tools are being adopted rapidly across organizations, often without formal approval, oversight, or governance. While this “Shadow AI” is typically driven by a desire for productivity, it introduces significant risks, including unintended data exposure, compliance gaps, and loss of control over sensitive information. With the rise of Microsoft 365 Copilot, these risks evolve. AI is no longer just external, it is embedded inside the digital workplace, operating on top of existing permissions, data structures, and collaboration patterns. This means organizations must rethink how they approach governance: not by blocking AI, but by understanding and guiding its use. This session explores a practical approach to managing Shadow AI in both external tools and within Copilot. It introduces a lightweight governance model focused on detection, risk understanding, policy-driven control, and user guidance. Attendees will learn how to balance innovation with control & enabling employees to safely leverage AI while protecting organizational data.
What This session explores how to design and implement multi-agent AI systems using MS Foundry Agent Service, where multiple specialized agents collaborate to solve complex real-world problems. The focus is on moving beyond single chatbots to role-based agents that work together with clear responsibilities. Why Many enterprise AI scenarios require coordination between capabilities such as data retrieval, reasoning, and task execution. A single agent often becomes difficult to scale and manage. Multi-agent architectures provide a structured way to divide responsibilities, improve reliability, and enable more intelligent automation. How Through a hands-on demonstration built using MS Foundry Agent Service, the session will show how agents can be designed with specific roles, connected together, and orchestrated to collaborate while maintaining context. The talk will highlight practical architecture patterns, communication flows, and key considerations for building scalable agent-based systems. For Whom This session is designed for AI engineers, developers, architects, and solution designers who want to understand how to build real-world agentic systems and implement collaborative AI agents in enterprise solutions using MS Foundry.
Most companies are still treating AI as a feature - a new chat interface, a smart suggestion, a productivity boost layered onto products built for a different era. Meanwhile, users are getting their first real taste of what comes next through agentic AI: experiences that are unified, personalized, and proactive…and indifferent to whose product or platform owns the data underneath. The gap between what users now expect and what companies are still shipping is where the next decade gets won and lost. For decades, products have been built from an org-chart lens, not a customer-needs lens. Companies fragment experiences along their internal silos and ask users to do the integration work themselves. Agentic AI breaks that pattern. As agents mature, on the user's side and across every step of their journey, they stop respecting the boundaries between products, brands, and platforms. They pull content, services, and data from wherever it lives, integrating on the user's behalf. The brand stops controlling the experience. Agents do. This talk walks through the three states of the new experience model — from today's fragmented experiences, to the unified brand experiences companies haven't yet built, to the integrated user experiences agentic AI is already previewing. We'll cover the principles that will define the new era — unified, personalized, proactive — and what it takes to design products for a world where your competition isn't another app. It's a network of agents across every step of a user's journey, including their own. And what we can see today is only the beginning. Session Goals By the end of this session, attendees will be able to: • Recognize the expectation gap between what users now experience through agentic AI and what most companies still ship — and why closing it defines the next decade. • Distinguish AI-as-feature from AI-as-experience-layer, and understand why most current AI investment is aimed at the wrong target — improving today's products rather than designing for what comes after them. • Apply the three-state experience model — fragmented brand experiences, unified brand experiences, integrated user experiences — to evaluate where their product sits today and where it needs to go. • Build toward the three principles - unified, personalized, proactive - that will define the new experience era and shape product decisions starting now. • Position products for an agent-driven world, where experiences are shaped by a network of agents across every step of a user's journey — and recognize that what's visible today is only the start of the shift.
Most enterprise AI initiatives don’t fail at the idea stage; they fail during implementation. In this session, I’ll share 5 key lessons from real-world AI adoption inside organizations, focusing on what happens after rollout. We’ll explore common breakdown points, including low user adoption, unclear data ownership, and AI tools that don’t align with daily workflows. I’ll walk through real scenarios where initial approaches didn’t work, the constraints we faced, and the adjustments that delivered measurable results. Attendees will leave with practical guidance on how to avoid common pitfalls, integrate AI effectively, and drive adoption across teams. This session is about decisions, trade-offs, and actionable strategies not theory/generic tool demos.
A typical production agent needs to read four kinds of data: Parquet files in S3, structured state in Postgres, document catalogs in MongoDB, and an embedding index for memory. The default approach is four SDKs, an ETL job to land it all in one place, and a pile of glue code that's quietly the most expensive part of the system. There's a sharper default: federated query as a primitive. One in-process engine that exposes Parquet, S3, Postgres, MySQL, MongoDB, Iceberg, Hudi, and 40+ data formats as SQL table functions. No SDK per source. No ETL job. The agent issues one SQL statement that joins data across sources at runtime. This talk is built on chDB — the in-process build of ClickHouse, which I created and lead at ClickHouse Inc. chDB inherits ClickHouse's federation primitives (s3(), url(), mysql(), postgresql(), mongodb(), iceberg()) and exposes them as table functions inside an embedded library you import. We'll cover: - The architecture: what "federated by default" actually means for agent data layers - The 40+ formats and 10+ external sources you get out of the box - Honest cost / latency math vs. the ETL-everything-first approach - Where federation is the wrong call (high-frequency reads on the same hot data) - Comparison with DuckDB's federation story — chDB is broader on sources, narrower in some places Closing live demo: an agent that joins a Parquet file in S3, a row from Postgres, and a document from MongoDB in one SQL statement — no ETL, no glue. Code published as an open repo for attendees to fork.
Building an AI agent is easy, but managing it across development, testing, and production environments is often the real challenge. As organizations adopt agents built with Microsoft Copilot Studio, they need structured processes to deploy updates, manage configurations, and ensure stable releases. This session introduces Application Lifecycle Management (ALM) practices for Copilot Studio agents using solutions in the Microsoft Power Platform. We will explore how agents, flows, and integrations can be packaged and promoted across environments while maintaining proper governance and change control. Through a live demo, we will build an agent in a development environment and promote it to test and production, demonstrating how ALM helps teams manage agent deployments reliably at enterprise scale. Key Takeaways • Understand Application Lifecycle Management (ALM) concepts for AI agents • Learn how to structure Copilot Studio agents using solutions • Safely promote agents across Dev → Test → Production environments • Manage configuration using environment variables and connection references • Implement deployment pipelines to support enterprise change management
Microsoft 365 Copilot can only generate insights from the data it can access. While emails, chats, and documents are already part of Microsoft 365, one of the largest workloads in the enterprise — voice conversations — is often disconnected from the AI layer. In this session, we’ll explore how Microsoft Teams Phone, Copilot, agents, and new extensibility APIs allow organizations to turn calls and meetings into actionable intelligence. We’ll look at real scenarios using intelligent call summaries, meeting insights APIs, Copilot Studio agents, and Microsoft Graph integrations to automate workflows, enrich CRM systems, and build AI-aware applications. We’ll also discuss practical migration strategies from legacy PBX systems, licensing considerations, and why bringing telephony into Teams is becoming a key step in any Copilot adoption strategy. If you're working with Microsoft 365, Teams, Copilot, or enterprise integrations, this session will help you understand how voice, AI, and APIs are coming together to form the next generation of the Microsoft productivity stack.
Governments are starting to use AI agents to improve how services are delivered—from permits and public works to citizen support. These “Civic Agents” are helping staff work more efficiently and making public services easier to access. In this session, we’ll look at real examples from local governments already using AI, and share a simple 5-step framework (Purpose, Pilot, Playbook, Production, Performance) for scaling these solutions. Whether you're in tech or public service, you’ll walk away with practical ideas to start or grow your AI journey.
Most organizations are experimenting with AI. Very few are building systems that can scale and produce true ROI for businesses. In this session, Samantha St-Louis draws from real enterprise work to explore what it actually takes to move from disconnected AI experiments to cohesive, enterprise-ready agentic platforms that drive business revenue. We’ll examine how agentic systems behave at scale, where most organizations lose control, and why architecture and standards matter before velocity. Attendees will learn how to design AI systems that coordinate multiple agents, preserve shared context, prevent duplication, and align technical decisions with long-term business outcomes. This is not a tool demo or a theoretical overview but a practical look at how modern organizations should be thinking about agentic AI as a platform, not a feature. Ideal for leaders and architects who want to scale AI responsibly without slowing innovation... or waking up to AI sprawl six months from now.
Microsoft 365 Copilot is more than an AI assistant — it’s becoming a platform. With the new Microsoft 365 Copilot APIs (Retrieval API, Chat API, Meeting Insights, Interaction Export, and Copilot notifications), developers can build applications that use Microsoft 365 as their grounding, reasoning, identity, and compliance layer instead of rebuilding those capabilities themselves. In this session, we’ll explore how the Copilot APIs are changing the architecture of Microsoft 365 solutions, and what this means for developers, architects, and ISVs building modern workplace applications. We’ll start with a quick overview of the Copilot extensibility model, then move into real examples showing how to use the APIs to build scenarios such as: - Retrieving grounded enterprise data with the Retrieval API - Building chat experiences using the Chat API - Triggering workflows from Copilot notifications - Exporting interaction data for analytics and automation The session includes live demos showing how these APIs can be used in real applications, and discusses when you should build inside Microsoft 365 versus outside of it. If you’re building solutions on Microsoft 365, Copilot APIs are not just another feature — they change the platform.
Microsoft 365 Copilot represents as much of a people and process shift as it does a technology deployment. This session focuses on the real-world questions organizations face when bringing Copilot into everyday work. Drawing on lessons learned from Copilot pilots and enterprise rollouts, the session combines structured insights with live, audience-driven discussion. We’ll go where the questions lead, with topics that may include training, communication, pilots, tenant preparation, and optimizing ROI. Whether you’re planning your first Copilot rollout or refining an existing strategy, this interactive session will help you cut through the noise, address concerns candidly, and walk away with practical next steps you can apply immediately.
As AI systems grow more autonomous, developers must ensure they remain aligned, transparent, and controllable. This session focuses on the engineering side of Responsible AI, how to embed alignment checks, human-in-the-loop review, and feedback tuning directly into your agent lifecycle. We’ll code live examples using Azure AI Foundry, Semantic Kernel, and telemetry hooks for self-evaluation. You’ll learn how to operationalize trust without slowing innovation. Key Takeaways Implement human-in-the-loop checkpoints for autonomous agents Add feedback tuning and scoring pipelines Use Semantic Kernel for controlled agent reasoning Apply telemetry and evaluation to sustain responsible autonomy
Most production LLM apps end up gluing together a vector database, a metadata store, and an analytics warehouse — three systems, three credentials, three latency budgets, three things to break at 3 a.m. There's a lighter pattern worth knowing about: a single in-process OLAP engine that handles embedding storage, vector search, and analytical SQL together — and federates to your production data cluster when you need to reach further. This talk is a practitioner's deep-dive built on chDB (the in-process build of ClickHouse, which I created and lead at ClickHouse Inc.). We'll cover: - Embeddings as a column type — what native vector indexes do for your retrieval latency budget, and what they don't fix - Hybrid retrieval (dense + BM25 + structured filters) in one SQL statement, and why "one statement" matters for predictable RAG latency - Hallucination diagnosis as an analytical workload — logging every retrieval/generation pair and treating eval like real analytics - Read-only-by-default and the federation pattern — how to give your AI app analytical capability without standing up new infrastructure or new credentials - Honest comparison: where this beats Postgres+pgvector, and where it loses to DuckDB-VSS Closing live demo: a working RAG pipeline in ~50 lines that ingests documents, generates embeddings, retrieves hybrid, runs a generation, logs the pair, and queries the eval log — all in one process. Full code published as an open repo for attendees to fork. This pattern works equally well as a foundation for Copilot Studio / Microsoft Foundry agents, LangChain / LangGraph, or your own framework — the talk is platform-agnostic.
Using the City of Montréal’s Copilot Studio implementation as a real-world case study, this session explores what it takes to deploy conversational AI at scale in a public-sector environment. We’ll examine how the city enabled natural-language access across 40,000+ webpages and connected APIs using Microsoft Copilot Studio and generative AI to improve citizen engagement, service accessibility, and operational efficiency. Beyond the technology itself, attendees will learn practical lessons around AI-friendly content design, Bing indexing, enterprise search, API integrations, governance, analytics, and responsible AI considerations. The session will also highlight the collaboration model between business and IT teams that helped support scalable AI adoption. Drawing from hands-on experience with Power Platform and Microsoft AI implementations, I’ll translate these lessons into actionable enterprise strategies for building compliant, impactful conversational AI experiences across organizations. Attendees will leave with practical implementation patterns, governance insights, and real-world lessons for deploying enterprise-ready copilots that improve user experience while balancing compliance and operational needs.
When we talk about AI security, the conversation usually focuses on models, prompts, and data. But what about everything underneath? The containers, packages, and pipelines that build and deliver your AI workloads are just as much an attack surface — and most teams aren't looking there. Project Hummingbird is a next-generation initiative to fundamentally change how application dependencies are packaged and delivered. By leveraging years of containerization expertise, Hummingbird produces ultra-minimalist, hardened, distroless container images for high-value open source languages, frameworks, and components — resulting in a dramatically smaller attack surface and near-zero CVE count in your base images. In this session, we'll explore what it really means to build AI workloads on a trustworthy foundation: why package provenance and signing matter even inside a container, how automated build pipelines enforce integrity at every step, and what "hardened" actually looks like in practice. You'll leave with a mental model for AI supply chain risk, an understanding of how distroless and minimal images change your security posture, and concrete practices you can bring back to your own pipelines — regardless of your stack.
L’intelligence artificielle offre de nombreuses opportunités pour améliorer l’expérience utilisateur. Se lancer, choisir un fournisseur de cloud de confiance et comprendre les différentes capacités du service peut s’avérer complexe. Au cours de cette introduction riche en démonstrations à Azure AI Foundry, nous découvrirons ce que le service propose, comment l’intégrer dans une application et quels scénarios il permet d’envisager. Cette session s’adresse principalement à un public de développeurs, mais les décideurs et administrateurs pourraient également y trouver des informations intéressantes.
Découvrez comment le nouveau connecteur SharePoint Embedded transforme le stockage des données et la gestion documentaire au sein de Power Platform. Cette session, destinée aux développeurs et aux professionnels IT, explore comment intégrer simplement SharePoint Embedded avec Power Apps et Power Automate afin de créer des applications métier modernes, sécurisées et évolutives. Vous aurez un aperçu concret de la manière dont le connecteur SharePoint Embedded permet aux applications Power Platform d’interagir avec des conteneurs de stockage de fichiers dédiés, en apportant toute la puissance des capacités de gestion documentaire de Microsoft 365 dans vos solutions personnalisées. Ce que vous allez apprendre : Présentation de SharePoint Embedded : comprendre son architecture et les avantages de son utilisation comme solution de stockage de fichiers sécurisée et orientée API. Introduction au connecteur SharePoint Embedded : découvrir comment ce nouveau connecteur relie Power Platform à SharePoint Embedded et offre un accès low-code à un stockage de niveau entreprise. Développement avec Power Apps : apprendre à créer des applications capables de téléverser, récupérer et gérer des fichiers dans des conteneurs SharePoint Embedded. Automatisation avec Power Automate : démonstration en direct d’un flux Power Automate déclenché depuis une application Power Apps, illustrant des scénarios d’automatisation réels. Cas d’usage et bonnes pratiques : explorer des exemples concrets dans différents secteurs (workflows documentaires, partage sécurisé de fichiers, stockage soumis à des exigences de conformité) et apprendre à les mettre en œuvre efficacement. À l’issue de cette session, vous disposerez des clés pour exploiter pleinement le connecteur SharePoint Embedded et enrichir vos solutions Power Platform avec des capacités de stockage de fichiers robustes, sécurisées et flexibles.
Conway's Law says any organization will design systems that mirror its own communication structure. For decades, that's been an annoying observation. In the AI era, it becomes a fatal one. It's not that we can't envision the experiences of tomorrow, it's that we can't envision how to build them in the organizations of today. Most companies are attempting to layer AI on top of organizations built for a different era - fragmented by business unit, product line, and acquisition history. The internet era taught us where this leads: innovation stranded on islands, integration prohibitively expensive, customers falling through the gaps. Companies are now repeating those exact mistakes with AI - bolting it onto the same fragmented structure and accelerating the chaos rather than solving it. In a shape like that, no company can move at the speed innovation now requires. The System-Led Organization is what comes after Conway's Law - a centralized, customer-aligned structure that learns from the internet era rather than repeating it. AI, data, applications, and product development live inside a unified system instead of scattered silos. This talk lays out what that structure looks like, the four big shifts required to build it (centralizing AI, applications and data, product development, and customer-aligned objectives), and the three principles it runs on — integration, data freedom, and customer alignment. We'll cover why most current AI initiatives are doomed by the structure they live inside, and what it takes to start building the shape of business that actually works for AI. Session Goals By the end of this session, attendees will be able to: • Diagnose the structure problem — why even clear AI vision fails when the company underneath isn't built to deliver it, and how Conway's Law makes today's fragmented organizations a liability in the AI era. • Recognize the internet-era pattern repeating with AI — how companies are bolting AI onto already-fragmented structures and accelerating the chaos rather than solving it. • Understand the System-Led Organization as a centralized, customer-aligned structure where AI, data, applications, and product development live inside a unified system rather than scattered across silos. • Apply the four big shifts required to build it — centralizing AI, applications and data, product development, and customer-aligned objectives — to evaluate where their company sits today. • Build toward the three principles that hold the structure together — integration, data freedom, and customer alignment — and start the transition before the next wave of competitors does.
Dans le contexte actuel, où l’environnement d’entreprise évolue à grande vitesse, le déploiement de solutions d’IA s’accompagne souvent de défis liés à la fiabilité et à la clarté, ce qui peut engendrer de l’incertitude et des résultats sous-optimaux pour les organisations. Cette session explore comment l’affinage (fine-tuning) et des pratiques d’évaluation robustes dans Microsoft Foundry peuvent répondre à ces enjeux, en permettant aux développeurs, aux décideurs et aux créateurs de concevoir des systèmes d’IA dignes de confiance. Les participants découvriront des stratégies concrètes pour adapter les modèles à des besoins métiers spécifiques et pour garantir des performances transparentes et mesurables. À travers des exemples réels et des recommandations actionnables, nous montrerons comment ces approches réduisent les risques et renforcent la confiance dans les déploiements d’IA. Rejoignez-nous pour découvrir comment transformer des intégrations d’IA imprévisibles en solutions fiables, créatrices de valeur pour votre entreprise.
What if your AI system could behave like a brain, perceiving, responding, and evolving, without being rebuilt every time the world changes? Welcome to NeuroModular AI Architecture, a framework developed by Microsoft MVP Samantha St-Louis for building living, thinking AI-native systems. In this session, Samantha helps you abandon the outdated model of AI as a static tool. Instead, you will reimagine it as a modular nervous system with reusable intelligence, dynamic feedback loops, and orchestration across the full Azure cloud ecosystem. You’ll walk through real-world architectures that connect short-term memory (vector DBs), long-term semantic storage (data lakes), shared cognitive models, and cross-system outputs like Copilot, APIs, and dashboards. Each “neuron” is built to scale, self-heal, and evolve making your AI infrastructure resilient by design. Whether you're a technical architect designing next-gen healthcare solutions or an executive looking to future-proof your AI investment, this talk will change the way you see intelligence in the cloud. Key Takeaways: Discover how to design modular, orchestrated AI systems that act like a living brain. Architect AI ecosystems that are resilient, reusable, and regulatory-ready. Use Azure AI Services, Foundry, Cognitive Services, and MLOps in harmony. Build event-driven pipelines with automatic retraining and drift correction. Implement multi-output intelligence without data duplication or model sprawl. Learn how governance, observability, and cost-efficiency form your AI’s immune system.
Dans cette présentation très pratique, Je montre comment passer de “Copilot = moteur de recherche poli” à un véritable allié stratégique, grâce à 4 leviers concrets : Personnalisation de Copilot : adapter l’assistant pour qu’il réponde efficacement aux habitudes de travail de chaque utilisateur Modes & modèles : (choisir le bon niveau de rapidité/profondeur selon la tâche), Prompts planifiés : pour automatiser les tâches répétitives. Agents Copilot (utiliser des agents existants ou créer le sien). Les participants repartent avec des méthodes applicables immédiatement pour améliorer la qualité des résultats, réduire la charge mentale et transformer l’IA en levier de décision.
16 speakers shared what they know.
ClickHouse, Tech Director
Spoke in Montréal
Software Engineering Director | SaaS Architect | Microsoft & AI Platform Builder
Spoke in Montréal
Microsoft 365 architect
Spoke in Montréal
aimpact Consulting , Co-founder, CEO
Spoke in Montréal
AI Automations & Agentic Solutions Architect | Digital Transformation Leader | Enterprise AI & Data Strategy | Power Platform & Fabric Architect | Program & Delivery Management | Cloud Modernization M
Spoke in Montréal
Chief Executive Officer, Ignatiuz Inc.
Spoke in Montréal
Senior Developer Advocate | Microsoft MVP Microsoft 365 & SharePoint | International Speaker | Community Leader | Product Evangelist
Spoke in Montréal
Best-Selling Author, Speaker, and Founder of Workshops by Kristen
Spoke in MontréalExpert en Solutions - Microsoft
Spoke in Montréal
Sr. Data & AI Engineer, Long View Systems
Spoke in Montréal
Systems & Functional Lead - D365, Power Platform & Copilot | AI Enablement & Adoption | Speaker | Creator
Spoke in Montréal
VP of Ai App Innovations @ Atmosera
Spoke in MontréalSenior Principal Software Engineer, Red Hat Hardened Images
Spoke in Montréal
Cloud Security Architect -Microsoft Regional Director & Microsoft MVP Security
Spoke in Montréal
Founder / CEO - Neon Navy
Spoke in Montréal
Développeur principal et conférencier
Spoke in MontréalEvery AI Community Day is run by volunteers from its city, with help from the global team.
The organizers of AI Community Days Montréal appear here once they confirm their place on the event.
Run the next one in MontréalThe people who run AI Community Days around the world.
CEO · Cloud and Security Architect
Cloud and AI Engineer · Tech Trainer · Tech HealthCare Expert · Women in Tech Advocate
Chief Information Security Officer at FORTEIA
AI, Microsoft 365 and Power Platform Architect · MCT and MVP · Trainer · Speaker
Founder & Independent Technical Consultant · Microsoft Certified Trainer (MCT) · Power Platform, M365, AI & Copilot
Solution Architect on Microsoft Dynamics & AI
The recap
This event took place on 17–18 June 2026. Thank you to everyone who came, spoke, sponsored and volunteered.
16
speakers took the stage
33
sessions on the agenda
Every certificate has a unique code. Anyone can check that it is genuine on our verification page, and you can add it to your LinkedIn profile from your account.
Tell us you would like the next edition, or help organize it with the AICD platform behind you.
Upcoming events
Nov 21–22
Azure Bots - MedTech SMU Hackathon
Nov 24–26
.NET Africa Conference 2026
