career path · 13 chapters · ~10h

build the support agent, then catch what it gets wrong

ai is answering your customers now. the customer service path teaches you to build those agents and spot the wrong answer before a customer does.

who this is for

you run or work in support, and ai already drafts replies, routes tickets, and answers customers. you need to understand what it's doing well enough to build it, monitor it, and trust it. this path gets you there.

what you'll be able to do

build a support agent that looks things up and answers with what it found. structure tickets into clean data your system can route. read what the agent actually did, and catch a wrong or off-policy answer before it reaches a customer.

the route, in order

start in support studio by defining safe ticket routes, escalation triggers, and review boundaries for a reply drafter. the shared builder spine supplies the data, agent, and evaluation skills that keep the workflow honest. self-paced; the full route is listed below.

  1. 01support studioch 50 · ~1h 1m
  2. 02from worker to builderch 00 · ~1h 14m
  3. 03read the scriptch 01 · ~21m
  4. 04read the data shapesch 02 · ~23m
  5. 05read the failurech 03 · ~24m
  6. 06read the pipelinech 04 · ~26m
  7. 07llm apisch 13 · ~45m
  8. 08structured outputch 14 · ~38m
  9. 09agent loopsch 16 · ~1h 11m
  10. 10mcpch 15 · ~40m
  11. 11prompting cursor and claude code effectivelych 19 · ~59m
  12. 12eval-driven ai developmentch 21 · ~57m
  13. 13debugging broken ai outputch 24 · ~1h 1m

how it works

no videos, no fizzbuzz. every step shows a small piece of ai-written code. you read it, predict what it does, find the bug, fix it, run it. it all runs in your browser.

where you end up

builds and monitors support agents, catches bad ai responses before customers do, routes and structures tickets with code.

free to start, no signup. plus unlocks every chapter at once; the free path moves through them in order. teams get progress visibility and role-fit rollout — see for teams.