The AI Adoption Flight Map — All 5 Levels, the Tell That You're Stuck at Each One, and the Exact Move That Gets You to the Next
The companion to the video. Boris Cherny — creator and head of Claude Code at Anthropic — published his 'Steps of AI Adoption' map in July 2026: five levels, from Level 0 (your company blocks the tools) to Level 4 (1,000+ agents steered by intent, where he says he personally lives now — while Anthropic itself sits at Level 3). THIS MAP: the full ladder on one page (level, agent count, role, unlock), then a page per climb — LEVEL 0 Gated (the bottleneck isn't technology: legacy security process, cost-per-token thinking, no technical voices — and the escape, which is politics: an executive sponsor, blockers escalated by name, a secure-launch framework, plus the guardrails that already exist: SSO, budget caps, existing approvals), LEVEL 1 Assisted (one agent, your attention is the bottleneck — and the climb: a self-verification loop you actually trust — tests, build, lint, e2e — auto mode with safe commands pre-approved, automated code + security review, then a second session on an isolated worktree), LEVEL 2 Parallel (~10 agents, you review diffs not keystrokes, and YOU become the slow part — the climb: give the model context, break work into loops and routines, and let Claude kick off Claude), LEVEL 3 Supervised autonomy (~100 agents, the question flips from 'did you read the code?' to 'what context was the model missing?' — plus the trap: scaling agent count before the loop has earned trust, and the cost filter: 'is this something an engineer would have done?'), LEVEL 4 AI-native (the quarter-long migration becomes a workflow you kick off and check on). Plus: the TELL for every level (the sign you're stuck), the THIS-WEEK move for each, the ROI worksheet that measures return instead of activity ('would you have spent engineering effort on this anyway?'), and the sequencing law that explains why skipping levels crashes. Framework: Boris Cherny, July 2026 — retold with independent field notes from Hyperautomation Labs. Not affiliated with Anthropic.
Free. No spam. Unsubscribe anytime.