The adoption numbers look spectacular. McKinsey's 2025 State of AI survey recorded the most dramatic single-year jump in enterprise AI use ever measured, from 55% to 78%. Gallup confirmed it from the workforce side. Deloitte's enterprise survey backed it up at 88%. But buried in the same reports is a far more interesting number: only 20% of organisations report actual revenue growth from AI. 2025 was the year everyone adopted AI. It was not the year everyone profited from it.
Executive Summary
- Adoption hit a ceiling-breaking milestone. 78% of organisations now use AI in at least one business function (McKinsey, May 2025), up from 55% in 2024. The tech sector leads at 77% total use with 57% using AI frequently. Retail sits lowest at 33%.
- Revenue impact remains concentrated. Deloitte's 2026 enterprise survey found 88% of companies report using AI, but only 20% attribute actual revenue growth to it. 74% say they hope AI will drive revenue. Hope is not a metric.
- Productivity gains are real but narrow. 66% of organisations cite productivity improvements (Deloitte). Three-quarters say they are meeting or exceeding ROI targets, but those targets are often set against narrow, easily measurable tasks rather than organisation-wide transformation.
78%
of organisations now use AI in at least one business function
Source: McKinsey, The State of AI, May 2025
20%
of organisations report actual revenue growth from AI
Source: Deloitte, State of AI in the Enterprise, 2026
77%
AI usage rate in the technology sector (highest)
Source: Gallup, Workforce AI Adoption Survey, 2025
66%
of organisations citing productivity gains from AI
Source: Deloitte, State of AI in the Enterprise, 2026
33%
AI usage rate in retail (lowest among surveyed sectors)
Source: Gallup, Workforce AI Adoption Survey, 2025
74%
of organisations that hope AI will drive revenue
Source: Deloitte, State of AI in the Enterprise, 2026
The Adoption Story
The raw numbers are hard to argue with. McKinsey surveyed over 1,600 organisations across industries and geographies. The jump from 55% to 78% in a single year is unprecedented for any enterprise technology, let alone one this complex. Gallup's workforce-level data adds texture: in tech, 31% of workers use AI daily. In retail, adoption barely registers.
AI Usage Rate by Sector
Source: Gallup, Workforce AI Adoption Survey, 2025
What drove it? Three things. First, generative AI tools became cheap and accessible. ChatGPT, Copilot, and their competitors gave every knowledge worker a reason to experiment. Second, executive pressure intensified. Boards that were curious in 2024 became impatient in 2025. Third, the cloud providers embedded AI into existing subscriptions, removing the procurement barrier.
The Value Gap
Adoption without value is just expense. And the Deloitte data paints a sobering picture of where most organisations actually sit.
88% report using AI. 66% cite productivity gains. But only 20% report revenue growth. The gap between "we use it" and "it makes us money" is the defining enterprise AI story of 2025.
The Enterprise AI Value Funnel
Source: Deloitte, State of AI in the Enterprise, 2026; McKinsey, 2025
Three-quarters of organisations say they are meeting or exceeding ROI targets. That sounds good until you look at what those targets measure. Most are scoped to narrow tasks: document summarisation time, customer service response speed, code generation throughput. Useful. Real. But not the transformational impact that justified the budget.
The organisations that are making money from AI did something different. They measured revenue impact from day one, not after the pilot was done.
Isaac Rolfe
Managing Director
What This Means for NZ and AU
NZ and Australian enterprises sit squarely in this global pattern. Adoption is high, particularly in financial services and government. But most deployments are productivity tools layered on top of existing processes, not structural changes to how the business operates.
The opportunity is clear. Organisations that move from "we use AI" to "AI changes how we deliver value" will pull away from competitors who are still measuring chatbot deflection rates.
The RIVER Perspective
We see this gap every week. Organisations come to us after their pilots worked but didn't scale. The technology performed fine. The business case didn't, because nobody defined what "success" looked like beyond the pilot metrics.
The fix is not more AI. It is better scoping, clearer measurement, and operational integration that connects AI outputs to revenue-generating workflows. That work is harder than deploying a model. It is also where all the value sits.