Enterprise AI crosses the pilot-to-production threshold amid governance growing pains

Moments like this echo the late-1990s ERP boom—when SAP installs jumped after GM, Nestlé and others went ‘all-in’—or the 1870s switch from water to steam in factories: once a few flagship adopters proved scale economics, diffusion accelerated and regulation scrambled to keep up (e.g., Britain’s 1878 Factory Act). Today’s burst of concrete metrics and mid-tier acquisitions signals that ‘agentic’ AI has moved from shiny demos to the plumbing of supply chains, security and data labeling. Long-term trends on display: vertical integration of data + models, the shift of security’s control point from perimeter to dataset, and the age-old skills bifurcation that accompanied every automation wave. Whether this milestone ultimately resembles electrification (broad, productivity-raising) or 1980s expert systems (over-hyped, niche) will hinge on the governance gap now evident in the EU’s first-ever delay of AI-Act high-risk provisions; history suggests that codifying guardrails during the scaling phase, not after, shapes societal dividends for a century.