McKinsey: most enterprises overrunning AI budgets on coding agents despite lackluster results
2026-09-22
Most enterprises are exceeding their AI budgets as software teams increase reliance on automated coding tools, CIO Dive reported Sept 22, 2026, citing McKinsey data, even as coding-agent results are described as lackluster.
Significance 3: A McKinsey-sourced, cross-enterprise adoption-outcome finding directly on the enterprise AI ROI question, meaningful even without a new capability or score attached.
Implications · machine-drafted, not owner judgment
This is direct, named-source evidence for the persistent-ROI-gap thesis: enterprises are increasing coding-agent spend beyond budget while McKinsey's own framing calls the results lackluster, mirroring the individual-level cost-overrun pattern already seen this month (Databricks, Gas Town). For the owner's org-readiness/GTM consulting angle, this is a citable instance of the technology lane (buying more agent seats) outrunning the operational lane (workflow redesign, incentive alignment) needed to realize the spend — the technology-illusion breakpoint in practice.
- McKinsey publishing the underlying report with methodology
- A named enterprise disclosing coding-agent ROI figures that confirm or contradict the budget-overrun pattern
Sources
- press Enterprises keep betting on coding agents despite lackluster results retrieved 2026-09-22