---
id: "concept-side-quests"
type: "concept"
source_timestamps: ["§ What Leaders Can Do"]
tags: ["innovation-frameworks", "culture"]
related: ["action-legitimize-experimentation", "entity-anthropic"]
definition: "Sanctioned, self-directed AI experiments conducted outside the official roadmap, designed to legitimize tinkering and reduce the stigma of shadow AI."
sources: ["execution"]
sourceVaultSlug: "hbr-seg-execution"
originDay: 8
articleStem: "hbr-cl-76-employees-not-transparent-ai-usage"
sourceUrl: "https://hbr.org/2026/06/why-employees-arent-transparent-about-their-ai-usage"
sourceTitle: "Why Employees Aren’t Transparent About Their AI Usage"
---
# Side Quests (AI Tinkering)

A term adapted by [[entity-anthropic-d8|Anthropic]]'s Claude Code team to describe self-directed experiments that engineers, designers, and product managers run *outside* the official product roadmap.

The strategic value is naming: by giving the behavior an official, sanctioned label, organizations convert AI tinkering from perceived 'corner-cutting' ([[concept-blameworthy-deviance]]) into a legitimate, recognized category of work. This reduces the stigma and fear driving shadow experimentation and lets standard surfacing mechanisms (team demos, repositories) function. The 20% time model at Google is offered as a precedent.

This is the concept behind the action [[action-legitimize-experimentation]] and is the direct enabler of protecting [[concept-praiseworthy-exploratory-testing]]. It is the fourth of the five [[framework-leadership-commitments-for-disclosure]].