ELC Annual 2026

To realize meaningful returns on AI investments, leadership must take accountability and ownership of establishing best practices, enabling engineers, measuring impact, and ensuring proper guardrails are in place.

When prompting practice and reflexive AI use is driven from the top down, engineers can align on the highest value use cases and experience peak productivity gains. When coupled with DX's AI Measurement Framework, leaders can gain a clear picture of AI's true impact, identify the real bottlenecks in the SDLC that can be augmented with AI, and drive improvement.

In this session, Justin Reock, Deputy CTO at DX, and author of DX's Guide to AI Assisted Engineering, will explain what the most effective leaders of AI enabled engineering organizations are doing to drive satisfactory utilization, augmentation, and psychological safety across their teams. Based on interviews, use cases, and data, leaders will walk away with an understanding of how to best lead their teams through mature AI rollouts.

Within every team, there's the culture we say we have (e.g. a culture of feedback, excellence, or experimentation), and then there's the culture we actually have. That gap widens fastest when the ground shifts: AI changing how work gets done, a reorg reshuffling priorities, an exec or investor issuing an edict from above. In those moments, your team's default behaviors get written fast, via behavioral pattern matching and perverse incentives. The best leaders know that culture isn't a values doc or a set of gameable productivity metrics — it's the specific behaviors you model, reward, and reinforce every day. In this talk, you'll walk away with a concrete, repeatable framework for closing that gap: defining the team culture you actually want and starting to shape it before the next chapter writes itself. Note: Attendees of this talk are invited to join Jill’s follow-up working session for live coaching and a group breakout session.

The rapid pace of generative AI development presents a unique architectural challenge: how do you build production-ready systems when the underlying infrastructure components change every few months? Box CTO Ben Kus asserts that to succeed with AI, it’s more critical to be prepared for change than to make the correct choice now. The engineering advantage is no longer just picking the perfect stack, it’s building systems, abstractions, and teams that can absorb the next change without starting from zero. Ben will discuss lessons learned from managing model neutrality, deploying AI across billions of files, and why building an adaptable, directed-graph architecture today is the best way to scale your enterprise AI solutions tomorrow.

Every engineering leader putting AI into production is making a set of decisions that will outlast any single model release. How much to build internally vs. buy, how much control to trade for convenience, who owns the model, the data or the token cost curve as the AI market shifts. Each of these decisions shapes team structure, vendor dependencies, and how quickly an organization can adapt.

Drawing on his work helping teams at companies like Uber and Shopify deploy AI at scale, Benny will walk through the tradeoffs behind these calls, including where a general-purpose model makes sense vs. when customizing open models wins, and how to build an architecture and a team that keeps evolving instead of locking into one choice. Attendees will leave with a framework to reason through these decisions for their own organization, weighing flexibility against operational load and how to lead a team through the daily decision-making of AI.

Building an AI agent is easy now. Running thousands of them safely in production is a different problem entirely.

AI agents have gone from demos to writing code, handling customer questions, automating workflows, and making real decisions. The hard part is no longer making agents smarter. It's keeping them reliable once they're operating at scale.

As a former VP of Engineering at Meta and now CEO and Co-Founder of Guild.ai, James Everingham has seen exactly where production systems break down. The failures come from everything around the model: context management, permissions, observability, cost controls, versioning, evaluation, and governance.

This session looks at what changes when agents stop being prototypes and start being production systems. Attendees will walk away with practical frameworks for designing agent systems that are observable, secure, reusable, and built to evolve.

A practical look at the SDLC mechanics that separate teams getting real output from agents from teams generating a lot of code that doesn’t hold up. This talk covers spec-driven development as a way to give agents better constraints, large language model (LLM) gateway and routing strategies for managing which model handles which task, test-driven development adapted for an agent-first workflow, and verification practices built for the volume and pace of AI-generated code. This session is a deep dive on the actual mechanics of building software well with agents in the loop.

Your next user might never visit your website, fill out a form, or open your dashboard. It might be an AI agent trying to get something done. Yet most products still require a human to verify an email and copy an API key before an agent can start working. Adding an API or MCP server isn’t enough. What’s missing is a way for agents to sign up and become legitimate users of your product.

Drawing on our work on auth.md at WorkOS, I’ll show how to build that path: letting agents discover your service, establish identity, register, and receive the permissions they need to get started. We’ll explore agent-native onboarding built on existing standards, including how a person or organization can later claim the account. Engineering leaders will leave with a practical framework for opening their products to agents, not just as another integration to support, but as a new way to reach and serve customers.

Engineering has always had a quantification problem: we could measure activity, but never truly connect investment to output. Ironically, AI agents are solving it. Tokens are metered, agent work is traceable end to end, and for the first time the inputs to engineering are as measurable as the outputs. This talk shows how leading companies are turning that visibility into a return-on-token-spend model: which productivity metrics hold up in agentic workflows, which quietly break, and how to tie token consumption to shipped outcomes. You'll leave with a framework you can implement with your existing tooling, and the numbers to answer the ROI question your CFO is already asking.

Baseten is a hypergrowth company with an unusual trajectory. In the last 12 months alone, the increases in revenue, usage, and headcount have put Amir in a position that very few engineering leaders have had to face.

Join Amir and Corinne for a conversation covering:

• Maintaining uptime, responsiveness, and reliability through hypergrowth
• Landing teams well through acquisition
• Choosing which problems to solve when engineering talent is stretched thin
• Sustaining high retention in a high-growth environment