ALL IN 2026 wrapped up in Montréal, Canada, on 17 September after two days bringing together industry leaders, researchers, investors and startups. The programme ranged from infrastructure and enterprise adoption to robotics and AI’s place in society.

For us at Mitigate, the quality of representation was the strongest part: major companies, well-selected startups and interesting products. Overall, we enjoyed it.

Six sessions from the programme

These selections reflect the published programme, rather than a claim that we attended each talk. Session titles link to the official descriptions; roles are those listed for the event.

  • OpenAI on adoption — 16 September. The Age of AI in Action paired George Osborne, managing director and head of OpenAI for Countries, with Bloomberg’s Rachel Metz. Its focus was the distance between experimenting with AI and getting useful work done, including in science and cybersecurity.

  • NVIDIA on open ecosystems — 16 September. Accelerating Canada’s Industry Advantage With Open Model Ecosystems, with NVIDIA vice president of global initiatives Calista Redmond, connected open models with software, infrastructure and Canadian research. The interesting question: how do those pieces give organisations more control over adoption?

  • Agents beyond the pilot — 16 September. Agents in Production: Turning Autonomy Into Results brought together Valérie Bécaert, VP of AI research at ServiceNow; Razat Gaurav, CEO of Kinaxis; Marc Klingen, co-founder and CEO of Langfuse; and Jaime Tatis, EVP and chief data and AI officer at CIBC. Moderated by MIT Technology Review’s Grace Huckins, the panel addressed realistic expectations, integration and oversight when agents enter production.

  • Unilever on procurement — 17 September. Competing Smarter: AI-Powered Competitive Buying at Unilever featured Gary Bogdani, head of Horizon3 AI Innovation Labs, and Jenn Maclean, its partnerships lead. Unilever’s session described a Scale AI-supported initiative combining forecasting and specialised agents to help buyers weigh cost, suppliers, sustainability and risk: a concrete business problem with competing priorities.

  • Google DeepMind on responsible deployment — 17 September. The Gemini Project: Responsibly Deploying Frontier AI at Scale paired Tulsee Doshi, senior director of product management at Google DeepMind, with Superhuman AI co-founder Hassan W. Bhatti. The discussion was framed around keeping utility, privacy and safety central as advanced models reach products used by billions.

  • Yoshua Bengio on reliability — 17 September. Engineering Honesty and Reliability for Safer AI featured Bengio, co-president and scientific director of LawZero, in conversation with Handelsblatt’s Larissa Holzki. Its published focus was honest agents, dependable guardrails and safeguards against frontier AI risks.

What we wanted more of

Much of the programme stayed at a high managerial level. What we missed was more field experience from the people repeatedly testing, fixing and refining agents and chatbots until they reliably perform.

We wanted concrete KPIs, achievements and failures: what worked in production, what broke, how many iterations it took, and what changed in the business. The programme did include technical and deployment-focused sessions; our point is about the balance and depth, not their absence.

We left with a positive impression of ALL IN. For the next edition, more candid accounts from hands-on experts would make an already enjoyable event more useful to teams doing the work.