Dylan Lake + Daramola Benjamin

As AI tooling evolves at an incredible pace, one question keeps coming up: how do we build systems that last when the capabilities of the models themselves are changing every few months?

This month, we're joined by two speakers exploring different sides of that question. One looks at the reality of building sophisticated AI engineering workflows in a world where today's innovation can become tomorrow's built-in feature. The other looks further ahead, asking how AI agents will fundamentally change the identity layer of the web.

Whether you're building developer tools, AI-powered products, or thinking about the future of agentic systems, these talks will challenge some assumptions and provide practical insights into where AI engineering is heading.

Building on Borrowed Ground: What I Learned Letting Anthropic Absorb My Work
Dylan Lake - Engineer, Crown Gas & Power

As foundation models become more capable, many of the workflows engineers painstakingly build eventually become native features of the models themselves. Rather than seeing this as wasted effort, Dylan shares why the real value lies in the knowledge gained along the way.

Drawing from his experience building multi-agent systems modelled on real workplace structures, Dylan explores the evolution of agent orchestration, what happens when platform capabilities catch up, and how to decide when custom engineering still provides value.

You'll learn:

• When to build custom AI workflows versus relying on model capabilities
• Lessons from designing multi-agent orchestration systems
• How AI tooling has evolved over the last year
• Where frontier models still provide unique advantages
• A practical framework for choosing the right tool for the job

-------------------------------------------------------------------------------

The Intent Graph: Rebuilding Web Identity for the Agentic Internet
Daramola Benjamin - Founder, Levered

For years, web applications have been built around behavioural data—clicks, sessions, cookies and analytics. But what happens when AI agents become the primary users of our applications?

Daramola introduces the concept of the Intent Graph, exploring how identity, authentication and authorisation may evolve in a world where autonomous agents research, negotiate and complete tasks on behalf of users.

Rather than focusing on marketing or analytics, this session takes a software engineering perspective on what AI-native products might require from the next generation of web architecture.

You'll learn:

• Why behavioural tracking breaks down in agent-mediated workflows
• How user identity differs from agent identity and task identity
• The role of OAuth-style permissions and scoped delegation for AI agents
• How event schemas may evolve in AI-native applications
• What developers should consider when building products for an agentic web

-------------------------------------------------------------------------------

Who Should Attend

Software engineers, architects, platform engineers, engineering managers and technical leaders interested in AI-native software development, agentic systems, developer tooling and the future architecture of AI-powered applications.

Why It's Worth Attending

Practically AI is about sharing real engineering experiences rather than product pitches. This month's talks combine lessons learned from building production AI systems today with thoughtful discussion about where AI platforms and web architecture are heading next. Whether you're building with LLMs already or planning your next AI-native product, you'll leave with practical ideas, fresh perspectives and plenty to discuss over pizza afterwards.

Agenda

18:30 – Doors, pizza & mingling
18:50 – Introductions + AI spotlight
19:00 – Building on Borrowed Ground: What I Learned Letting Anthropic Absorb My Work
Dylan Lake
19:30 – Q&A
19:45 – The Intent Graph: Rebuilding Web Identity for the Agentic Internet
Daramola Benjamin
20:15 – Q&A
20:30 – Open social & refreshments
21:00 – Close