OPEN
In 2023 and early 2024, some high-performing organizations were already reporting measurable EBIT impact from gen AI within business operations, with some pathways to value on roughly annual—not decadal—horizons.
Source: McKinsey & Company ·
Analyst note
Enterprise ROI claims are a battleground of incentives: vendors want logos, consultants want repeat engagements, and CIOs want proof that budgets were justified. McKinsey-style surveys are useful priors but are not receipts.
We model this claim as a distribution: some workflows yield clean before/after metrics; many do not because success requires data hygiene, process redesign, and evaluation discipline. Model routing—Cohere vs Anthropic stacks, etc.—matters chiefly through deployment fit rather than leaderboard aesthetics.
Evidence timeline
McKinsey’s annual AI survey highlighted a minority of organizations reporting material bottom-line impact, with adoption concentrated in specific functions and workflows.
Independent academic replication attempts emphasized measurement frictions: cost allocation, overlapping productivity programs, and survivorship bias in vendor case studies.
By mid-2026, the ‘18-month ROI’ hypothesis for early-2024 deployments became empirically testable for first movers; public disclosures remained uneven, with stronger evidence in software and customer support than in heavy industry.