Strategy, data, platform, automation, and governance — ordered by what practitioners are reading right now.
AI test generation raises coverage metrics while the test oracle problem bakes bugs into passing tests. The diff-scoped, spec-first, mutation-filtered pattern that generates tests worth keeping — and the failure modes to screen before they reach human review.
AI ROI stalls when data quality is measured in null rates instead of decision outcomes. The five-layer causal map and accountability bridge connecting data governance to business outcome ownership — so quality failures have a named owner and a dollar cost before the quarterly review.
When production agents fail, teams default to prompt tuning regardless of structural root cause. This MAST-based triage protocol gives engineering leaders three speed-ordered checks — 30 seconds, 5 minutes, 20 minutes — each routing to a different structural owner before anyone changes a line.
MAST's 14 agent failure modes cluster into 3 structural categories, each preventable at a different pre-production stage. This playbook maps them to 12 deployment gate questions with pass criteria and named ownership.
Why frontier model defaults bloat inference bills, and the per-task quality SLO framework that makes model selection explicit, testable, and owned — instead of inherited from prototype defaults.
MAST (NeurIPS 2025, UC Berkeley) identifies 14 MAS failure modes across 3 structural categories. This playbook maps them to 3 diagnostic questions — and tells you which layer to fix before touching the model.
Field notes for the people putting AI into real organizations. Useful, specific, and built for readers who ship.
