Applied AI in Action: Where It’s Already Working, and What It Takes to Scale
Adoption is no longer the hard part. Turning AI into repeatable business value is.
McKinsey’s State of AI 2026 survey, published in late August, puts numbers on that gap: 44% of organisations now report AI scaling across their enterprise, up from 38% a year ago, and 80% of people using AI in their own role say it has made them more productive. Yet only 37% say AI has contributed to their organisation’s EBIT, while just 6% qualify as AI high performers. (McKinsey & Company)
It’s why Applied AI gets interesting department by department rather than model by model. In our work at TeraSky Europe, the use cases that hold up don’t start with a model and a question about what it can do. They start with a workflow: where work slows down, where skilled people spend time on repetitive tasks, and where a faster or better decision would matter.
Over the past month, we’ve walked through six of those workflows across HR, Legal, Finance, Retail, Product, and the C-suite. Here’s what they have in common, and what the wider research says about turning AI from a pilot into an operating capability.
From administrative load to time back
The next group of use cases moves closer to revenue and decision quality.
In Retail, a search and recommendation layer works directly with the existing product catalogue, with no custom model to build from scratch. It helps customers find more relevant products and gives the business another lever for improving conversion and average order value.
In Product, an AI agent removes a familiar bottleneck. Instead of filing a request and waiting for the data team, a product manager asks an A/B-test question in plain language, and the agent queries the existing data in BigQuery to return a clear answer.
At the executive level, the same pattern extends further: ask a business question in natural language and receive an answer, a chart, and the supporting source data from connected systems, without adding another report request to the queue.
Google Cloud’s ROI of AI 2026 research backs this shift at scale. Together, 84% of surveyed executives report AI returns that are either steadily increasing or accelerating year over year, while 94% report that AI agents contribute to both cost savings and revenue. But only 26% belong to the group Google Cloud calls AI ROI Leaders, where returns are actively accelerating.
What distinguishes that group is not simply access to better technology. AI ROI Leaders report clearer ownership and decision-making authority, deeper integration of AI into core business processes, and more systematic ongoing AI capability development.
More broadly, when executives were asked what enabled their organisations to scale AI from pilot to production and deliver measurable value, the leading answers were security, compliance and regulatory readiness at 46%, workflow redesign at 43%, and workforce training and change management and cloud or infrastructure modernisation at 41% each.
Why scaling is still the hard part
McKinsey’s numbers show the same divide by company size. Among organisations with more than $1 billion in annual revenue, 40% are now scaling AI agents in at least one function, up from 27% a year ago. Among smaller organisations, the figure remains at 22%. (McKinsey & Company)
And scaling introduces a different category of risk. Gartner predicts that by 2027, 40% of enterprises will demote or decommission autonomous AI agents because of governance gaps identified after production incidents. Its argument is not that agents cannot work, but that their autonomy, permissions, monitoring, and governance need to match the level of responsibility they are given. (Gartner)
Deloitte’s 2026 enterprise research shows the same tension from another angle. Two-thirds of organisations already report productivity and efficiency gains from AI, and 53% report better insights and decision-making. But only 20% currently report increased revenue from AI, even though 74% hope to achieve revenue growth from their AI initiatives in the future. (Deloitte)
The pattern suggests that the constraint increasingly sits around the model: governance, ownership, data, infrastructure, and workflows.
An AI agent can perform impressively in a controlled pilot and still struggle when it meets fragmented data, inconsistent permissions, unclear ownership, or a process that was never designed for an agent to operate inside it.
Four things worth checking before scaling a use case
1. Start from a bottleneck, not a capability. A clearly defined process gives you something concrete to improve and a metric against which to judge the result before anyone starts discussing the model.
2. Connect to the systems where the work already happens. The use cases above work because they connect AI to tools such as Gmail, Jira, BigQuery, HR platforms, and ERP systems rather than putting another isolated interface in front of employees.
3. Redesign the workflow, don’t just speed up every step. Making an inefficient process faster doesn’t necessarily make it better. The larger gain may come from removing a hand-off, shortening an approval cycle, or eliminating a step.
4. Decide ownership, governance, and measurement before scaling, not after. Someone needs to own what the system can access, what it can do, where human review remains necessary, and how success will be measured.
Applied AI is becoming an operating capability
The use cases may look unrelated: screening CVs, drafting legal replies, generating financial reports, recommending products, analysing A/B tests or answering executive questions. But underneath them is the same pattern: a clear business problem, the right data, integration with existing systems, strong governance and security, and a measurable outcome.
That is also why TeraSky Europe positions Applied AI as more than experimentation. The aim is to build production-ready AI that works with modern cloud, data and security platforms and is supported by robust data foundations and enterprise-grade protection.
The next competitive advantage is unlikely to come from simply having access to the newest AI model. It will increasingly come from an organisation’s ability to turn AI into a reliable part of how the business works: connected to real workflows, governed properly, trusted by users and measured against business outcomes.
So, the more useful question is no longer, “Where could we use AI?” It is: which workflow is valuable enough to improve next?
Sources:
TeraSky Europe, Applied AI Use Cases, July 2026. Internal presentation covering Applied AI use cases across HR, Legal, Procurement, Finance, IT, Marketing, Retail, Product, and Top Management.
McKinsey & Company, The State of AI in 2026: On the Road to ROI, August 25, 2026. McKinsey State of AI 2026.
Google Cloud, ROI of AI 2026. Research based on 2,403 executives covering AI returns, agents, scaling enablers, and business outcomes.
Deloitte, The State of AI in the Enterprise 2026. Deloitte, State of AI in the Enterprise 2026.
Gartner, Applying Uniform Governance Across AI Agents Will Lead to Enterprise AI Agent Failure, May 26, 2026. Gartner AI agent governance research.
