Businesses warned to define goals before scaling agentic AI

Summarised from 3 outlets · Updated 22 Oct, 22:03 · Archive

Just 39% of organisations reported financial impact from AI investment despite widespread adoption, according to McKinsey's 2025 survey.

Organisations are increasingly adopting agentic AI without defining the value they hope to achieve, leading to strategies that may not deliver results. The rush to deploy AI tools following the rise of ChatGPT meant many companies rolled out systems to their workforce at large with no defined goals or end point in sight. Without establishing a baseline before implementation, there is no way to know what has actually improved or whether the investment was worthwhile.

McKinsey's 2025 survey highlighted this problem, which it terms the 'gen AI paradox': just 39% of organisations reported any enterprise-level financial impact from their AI investment, despite mass adoption. The challenge lies in measuring the right things. Most organisations capture only adoption metrics, but miss the crucial 'behaviour layer' in between that shows causation rather than mere correlation. For example, employees who use AI may sell more, but measuring this properly requires understanding specifically how the AI changed their behaviour—such as giving them time to focus on client calls rather than paperwork.

Organisations need to be precise about what they want to measure before scaling agentic AI. This requires understanding not just whether people are using the tools, but what measurable business outcomes result from that use. Without such clarity, the hammer-and-nail problem persists: having access to AI as a tool can lead organisations to apply it everywhere without asking whether it actually solves the underlying business problem.

How it is being reported

• In Meta’s version of the future, I’ll have my Muse call your MuseMark Zuckerberg is barrelling toward a future in which an autonomous network of AI agents will run our lives - shopping for us, working for us, talking to other people's AI agents for us. Are humans ready for this very specific and unrelenting vision of a life in which we communicate primarily through bot proxies? […]The Verge · 22 Oct, 19:15
• ‘Know Your Agent’ will be the new rule of the AI ageAs autonomous software does more, we need to focus on how to control its behaviour and establish accountabilityFinancial Times · 22 Oct, 15:38
• Stop choosing the wrong Agentic AI use casesWhy businesses must rethink agentic AI before scaling costly, ineffective use cases.TechRadar · 21 Oct, 11:19

In this story: McKinsey · ChatGPT · Meta · Mark Zuckerberg

This summary was written by AI from the headlines and standfirsts above, and states as fact only what at least two outlets report. How we use AI · Report a problem