Shopify CEO Reveals Their Secret AI Developer

TL;DR

• Shopify’s big AI win is a workflow constraint, not just usage numbers — River was used by 5,938 employees across 4,400 Slack channels and opened 1,800 pull requests in one week, but the key design choice is that every interaction happens publicly in Slack.

• Most companies have a visibility problem, not a tooling problem — employees already use ChatGPT, Claude, Copilot, and coding agents constantly, yet the best prompts, corrections, and workflows disappear into private chats so teams keep reinventing the same solutions.

• The real knowledge to share is task, context, interaction, and review — a prompt library misses the messy but valuable parts: what context was loaded, how the model was pushed back on, what got rejected, and why a human overrode a plausible answer.

• This is an apprenticeship crisis in the AI era — junior people rarely get to watch how senior operators frame ambiguity, supervise models, and reject weak outputs, so Nate frames the issue as an “apprenticeship gap” that widens every quarter.

• Public AI work has to be deliberate, bounded, and led by senior people — the model isn’t ‘make every chat public,’ but create declared channels for reusable, nonsensitive workflows, with leaders like Shopify CEO Tobi Lütke showing their real work in public.

• The best metrics are about reuse and fewer repeated mistakes — instead of obsessing over token volume, teams should track reusable workflows, cross-team adoption, pinned examples, retired stale patterns, and whether public AI work reduces duplicate effort and recurring errors.

The Breakdown