Tech & the future of AI

Most organisations are experimenting with AI, but few get past the pilot; analysts expect over 40% of these projects to be cancelled. This is the story of one that made it. Philips Hue brought AI to millions of smart-home users and has kept building ever since, growing from a simple assistant to a platform where users create their own automations in plain language. We share the pragmatic lessons behind moving from experiment to production, and how to keep delivering value once you're there

Most organisations are experimenting with AI, but few get past the pilot; analysts expect over 40% of these projects to be cancelled. This is the story of one that made it. Philips Hue brought AI to millions of smart-home users and has kept building ever since, growing from a simple assistant to a platform where users create their own automations in plain language. We share the pragmatic lessons behind moving from experiment to production, and how to keep delivering value once you're there

• What happens to work and social status if machines become better than humans at much of what we currently consider skilled, creative or intelligent activity?
• What happens to human relationships when AI can become our colleague, friend, therapist, romantic partner—or even a more attentive companion than another human being?
• What happens to truth, identity and meaning when much of the world around us is generated, mediated and interpreted by artificial systems?

Customer-facing AI agents are moving from static chatbots to autonomous assistants that manage accounts, integrate with user tools, and execute purchases. But exposing (autonomous) agents directly to external users introduces critical architectural challenges: How do you enable agents to safely act on behalf of a user? How do you let agents connect to third-party and internal tools without mishandling user credentials? And how do you verify a real human is behind the agent and empower it to complete purchases without risking unauthorized actions?

This breakout treats AI agents as work horses on a running platform: what they take over, and where the same idea shows up in other fields. My angle is security: I used AI agents to reconstruct architecture documentation that never existed, straight from code and history, then verify every compliance claim against evidence instead of trusting the paperwork. It's told as a story, not a slide deck: come with your own examples of AI doing the tireless, repetitive work in your domain.

This breakout treats AI agents as work horses on a running platform: what they take over, and where the same idea shows up in other fields. My angle is security: I used AI agents to reconstruct architecture documentation that never existed, straight from code and history, then verify every compliance claim against evidence instead of trusting the paperwork. It's told as a story, not a slide deck: come with your own examples of AI doing the tireless, repetitive work in your domain.