‘AI, and especially Agentic AI, is worse than heroin. And I’m not saying that just to cause a sensationalist’, says Gartner Chief of Research Daryl Plummer.
‘Heroin is addictive, and so is AI. And heroin use can lead to people harming themselves. AI use is leading to people harming themselves or the company. Heroin is something that will cause you to do things you might not otherwise do. ‘AI users – we call them “Hero innovators”’ do the same. They do things they thought were good, but end up doing things they didn’t otherwise want to do. The parallel is quite clear.’
‘So the solution suggested in my research is that we need a framework for AI usage within an organisation. We’ll say, look, this is a Class One or Class A AI use case. It gets as many tokens as it wants. This is a Class D use case. It can only have a limited number of tokens. You can apply this to the number of tokens, the cost of the tokens, and the models you use. It would be a smart move for organisations to implement this today. In the US, nobody wants to do that. They hate regulation. In Europe, we might see it happen. That would be a good thing,’ says Daryl Plummer.
