The ROI of AI 2026: Who Is Pulling Ahead, and What Sets Them Apart?
The AI conversation is shifting from access to outcomes. The more useful question for executives is no longer which model or platform to choose, but what must be in place for AI to deliver measurable business value.
Google Cloud’s ROI of AI 2026 report, based on a survey of 2,403 executives conducted with the National Research Group, points to a clear pattern. Together, 84% of respondents report financial returns from AI that are either steadily increasing or accelerating year over year. But only 26% fall into the group Google Cloud calls AI ROI Leaders: organisations where returns are actively accelerating. Three factors stand out: clear ownership, AI embedded into core business processes, and continuous AI capability development.
What do AI ROI Leaders do differently?
The first difference is ownership. 48% of AI ROI Leaders describe ownership and decision-making authority for AI agent initiatives as “extremely clear”, compared with 27% of other organisations.
The second is integration into the business. 48% of leaders say AI is embedded into core business processes and revenue streams or is enabling new business models and revenue opportunities. Again, only 27% of other organisations say the same.
The third is capability. 38% of leaders have comprehensive, ongoing AI capability development embedded into roles with required training, compared with 18% of their peers.
Technology alone does not explain these gaps. The stronger pattern is operational: someone owns the outcome; AI is part of a real workflow rather than a side project, and employees are equipped to work with it.
Where the ROI is coming from: agents
The report’s clearest answer is AI agents. 94% of respondents report that agents contribute to both cost savings and revenue. In comparison, 86% agree that AI enables them to scale revenue or output without proportional increases in operating costs. The customer examples make that more tangible. Tata Steel deployed more than 300 specialised AI agents across its global operations in nine months. Highmark Health reports $27.9 million in value from its Sidekick AI assistant in 2025. Elanco estimates $1.9 million in ROI from agents supporting processes including pharmacovigilance and customer orders.
But agents are not the whole story.
When executives were asked what enabled them to scale AI from pilot to production and deliver measurable value, the top answer was improved security, compliance, or regulatory readiness at 46%. Workflow redesign was followed by 43%, while workforce training, change management, and cloud or infrastructure modernisation were each cited by 41%.
That is an important reality check. The agent may be the visible layer, but its value depends on what lies beneath it.
AI ROI is moving beyond productivity
Another shift is happening in how organisations define AI value.
Faster strategic decision-making is now the most-cited measurable outcome of AI investment, at 55%, ahead of increased workforce capacity at 52%. Other reported outcomes include faster innovation, improved customer lifetime value, new-product revenue, better risk intelligence, and revenue expansion within existing offerings.
That matters because many AI business cases still begin and end with time saved.
Productivity remains important, but the next stage of AI ROI is broader: better decisions, faster innovation, stronger customer value, new revenue, and improved risk intelligence.
And this is where the conversation about infrastructure changes. If AI is expected to act inside core workflows, data quality, identity, access controls, governance, monitoring, and accountability stop being supporting topics. They become part of the AI system itself.
The real AI gap is a readiness gap
This is the pattern we see in Applied AI projects as well.
An agent can perform impressively in a controlled pilot and still struggle when it encounters the organisation’s real environment: fragmented data, unclear process ownership, inconsistent access rules, or workflows never designed for an agent to operate within them.
The agent is rarely the hardest part. The foundation it must run on is. Before adding another AI agent, the better questions are:
Who owns the outcome?
Can the data and identity layer support an agent that acts?
Is the workflow ready for it?
Can the organisation measure whether it is creating value?
Get those foundations right, and AI can move from an isolated experiment into an operating capability.
The report’s final signal is telling: 97% of respondents plan to increase AI spending next year. The areas expected to deliver the greatest returns include AI-powered analytics and decision intelligence, workflow automation within existing systems, customer-facing AI, and data and infrastructure modernisation.
The investment is continuing. The advantage will come from how well organisations turn that investment into AI systems that can operate, scale, and deliver measurable value.
Source: Google Cloud, ROI of AI 2026; Google Cloud, How AI ROI Leaders prioritise investments for real business outcomes, July 2026.
