Build Otto: Ship an AI Agent Without Redeploying

Hardcoded prompts and models mean every AI improvement waits on a release, every bad response is a live incident, and model costs are locked in. This hands-on workshop shows how LaunchDarkly’s AgentControl turns prompts, models and targeting into runtime configuration, so your team iterates in minutes, not sprints. You’ll build Otto, the shopping assistant for ToggleWear, a fictional LaunchDarkly-branded apparel retailer: give him a personality, reuse brand voice across agents, route premium shoppers to a more capable model, and monitor cost, latency and quality in production. By the end, Otto is live, on-brand, tier-aware and measurable, with zero redeploys. Bring a laptop; we’ll provide access to the labs.