Methodology

Collection policy: metadata_only. Content fields are denylisted by default. Token/cost fields are incomplete for some tools (notably Copilot daily metrics).

• Adoption funnel

  Unique actors grouped by highest stage reached this period. Each bar is a filter link. One person counts once at their best stage, not once per tool-week row.

• Activated

  First real session beyond hello-world. Threshold: at least one substantive session.

• Habitual

  Used on 3+ active days per week, mid-task. Habitual = 3+ active days/week with real tool use.

• Deep

  Agentic multi-file work ending in commits. Habitual week with commit outcome and agentic depth: ≥1 subagent or ≥5 tool calls (PostToolUse only — PreToolUse pairs are not double-counted).

• Using shallowly

  Short sessions, low tool depth, no commits.

• Using poorly

  High abandonment, errors, token waste.

• Tool mix

  Share of actor-weeks per AI tool in the current filter.

• Team × week stages

  Each cell shows where that team sat on the funnel that week. Modal unique-actor stage per team-week, with higher ladder stage winning ties. Full formula: Methodology page.

• Week over week

  Stage counts vs the previous ISO week (same team/tool filters).

• Verdict

  One sentence on whether adoption is stuck or moving, for the current filters. Built from the attention list and week-over-week Deep+ movement. Full formula: Methodology page.

• Cost per Deep user

  Estimated period spend divided by people at Deep or Multiplying. Blank when nobody is Deep+ yet, so we never divide by zero. Full formula: Methodology page.

• Copilot daily metrics

  GitHub reports one row per user per day — not session detail. On Copilot rows, “session count” means active days in the week. Full formula: Methodology page.

• Nudge outcomes

  Playbook actions leaders marked as sent, joined to stage movement 2 and 4 weeks later. Aggregate counts only — whether the nudged team or person moved up a stage. Full formula: Methodology page.

• Estimated spend

  Token/cost fields are incomplete for some tools (notably Copilot daily metrics). Spend rollups sum actor_week cost_usd for tool/team slices and session + daily model rows for the model table. Figures are labeled estimated until all tools report cost.

• Commit-window attribution

  PR is AI-assisted when a matched actor has a session overlapping [first commit − 2h, merge + 1h]. Actor match v1: git_prs.author_login ↔ actors.metadata.github_login. Directional only — no prompt or content analysis.