Most "AI agents in production" talks skip the part where you have to drag tribal knowledge out of 100+ country SME teams and turn it into something an agent can execute safely. This is that part. How Maersk ships AI agents handling 100K customer cases a day across global logistics, and why extracting and aligning the tribal knowledge was 10x harder than the agent itself. - Why SOPs-as-code (versioned markdown, per-country) beats prompt engineering at this scale - The SME alignment loop: how corrections become SOP changes without breaking 99 other countries - Guardrails that matter in production: write-gating, loop breakers, classifier vs. SOP-body routing layers - Where agents under-deliver against the demo, and how we measured it honestly - Org/process patterns for the Applied AI / Forward Deployed Engineer stack across 100+ countries
AI-Native Enterprises sessions at AI Engineer World's Fair 2026 in San Francisco.
Thursday, July 2, 2026
11:10 AM - 11:30 AM·20m
Leadership 1 · Room 3016
Capacity: 550 attendees
Sign in to add this talk to your schedule.

Dmitry Buykin
Maersk, Applied AI Builder
Maersk
Dmitry designed, built and runs the agentic customer-experience system that processes 20 million emails a year for one of the world's largest logistics operations across 100+ countries.