There are valid uses for AI, and valid AI models you should use. (LLMs are rarely one of them, having only a handful of reliable applications, but, when you push them aside, there are lots of other AI technologies that actually work if you don’t get blinded by the hype.) But there are invalid cases, and AI models you shouldn’t use. So to make it easy-peasy for you, here’s a simple guide!
| USE AI WHEN | DON’T USE AI |
| It’s a well constrained use-case where AI has been successfully deployed in industry, where the confidence is proven, and where you have access to the right technology. | It’s a poorly defined use case, AI has not yet been successful for the use case, the confidence is unproven, and/or the tech you have access to is still in alpha! |
| It works as well as previous gen tech but with significantly shorter training cycles and easier integration and utilization. | Previous gen tech still works better, costs less, and/or has no impediments to integration or UX. |
| It can bring value beyond what last generation tech can bring. | All of the value can be achieved using traditional rules-based (A)RPA, (decision) optimization, analytics, or (classical) machine learning. |
| It’s a cost effective solution that can be run predictably based on a predictable cost model. | It’s based (primarily) on LLM(-based) models that have unpredictable compute costs and, with the wrong request, can eat up thousands of dollars on a single request. |
| You have a valid use case for agentic tech! | You think you have a valid usre case for agentic tech. (If you think you’re ready for AI, you’re NOT ready for AI.) |
| You’ve mastered current generation tech. | You’re still a generation (or three) behind on tech. |
| You have in-house expertise on what AI is, and isn’t; where it can, and can’t be successfully deployed; and what “AI” is typically appropriate in a given situation. | You’re relying entirely on (junior) consultants from the Big X promising it’s gonna “change your life“. |
| It’s designed to augment human performance and make your employees more productive and more effective super humans (able to do the work of 3, 5, 7, and even 10 regular humans). | It’s designed to replace humans. (This doesn’t mean it can’t reduce the number required to do a task, just that at least one is still maintained to handle exceptions and make decisions.) |
| The firm is selling augmented intelligence! | The firm is selling AI Employees. (There are none! And any firm that makes this claim is dehumanizing your employees! But hey, it’s your choice if you want to lose all your money.) |
Get it yet?