How to choose the right agentic AI use cases

Businesses should identify the work they want to improve before choosing agentic AI, according to TechRadar contributor Pete Evans. Without a clear goal and baseline, organizations may struggle to show what an AI investment achieved. McKinsey’s 2025 survey found that just 39% of organizations reported any enterprise-level financial impact from AI investment, despite widespread adoption.

Evans says measuring adoption and outcomes alone can show a correlation, but understanding how work changes helps explain whether AI caused an improvement. In one procurement process at Atos, a workflow described on paper as four steps was closer to 10 when examined with the people doing the work. Staff moved between Enterprise Resource Planning (ERP), Outlook and spreadsheets, copying and re-entering information. A four-step baseline would have measured the wrong process, and an agent given four steps could not complete a job that took 10.

He says fixed-rule tasks may be better suited to Robotic Process Automation (RPA), workflow automation or process changes. Agentic AI is more appropriate when work requires interpreting context, coordinating across tools, making recommendations or acting within defined boundaries. Evans recommends assessing use cases for financial impact, feasibility, process and technology requirements, human impact and risk, rather than ranking them by return on investment alone.

Organizations also need to budget for adoption and change. Gartner calls the extra effort a “transition mortgage”; AI adoption may require 25% more training effort and up to 200% additional change management. Evans argues that success should be measured by whether work changed and whether the organization can demonstrate that change, not by the number of agents deployed. TechRadar notes that the views are the author’s and are not necessarily those of TechRadar Pro or Future plc.