The 4 Levels of AI Proficiency

3 Apr 2026 · 1 min

Watch on YouTube

1:11

In this video, the presenter references Nvidia CEO Jensen Huang’s claim that companies are now prioritizing AI ability over traditional academic metrics like GPA when hiring new graduates. He then outlines his company’s internal framework for categorizing AI proficiency into four tiers: (1) Unacceptable – basic use of chat interfaces such as ChatGPT or Claude for simple research; (2) Capable – deeper engagement using tools like Claude CoWork and Claude Skills; (3) Adaptive – leveraging Claude Code to build custom solutions; and (4) Transformative – deploying autonomous agents to create end‑to‑end applications, citing platforms such as OpenCLAW and NemoCLAW for enterprise security. The speaker warns that without clear definitions of these levels, organizations risk hiring underqualified “bozos.” He concludes with a call‑to‑action, inviting viewers to comment “newsletter” for more AI, business, and marketing content.

Chapters

Nvidia CEO Jensen Huang said that when it comes to hiring new college grads, they are not even looking at GPA. They're looking at how good they are with AI. Now, if you want to adopt this, I think there are some caveats to it. Let's get into it. So let's talk about the 4 levels of AI proficiency. So at my company, we have unacceptable, we have capable, we have adaptive, and we have transformative. First level, unacceptable. People are just using ChatGPT, they're just using Claude, they're using it for research. That's not good enough. Then you have maybe the next level for capable. You have people that are using Claude Cowork, they're using Claude Skills, so maybe they're venturing a little deeper in, but it starts to get really interesting once people are adaptive. Maybe they're using Claude Code because Claude Code allows you to build whatever you want. Once you're able to inject yourself with that, then the next level is transformative where you're using agents to build end-to-end solutions. You can use a lot of these solutions out there like OpenCLAW, or you can use NemoCLAW for enterprise security. And a lot of these AI tools are getting better with their own versions of it. If you don't define your levels of AI proficiency clearly, what's going to happen is you're going to end up hiring bozos and you don't want that to happen. So if you want more like this on AI, business, and marketing, just comment newsletter and I'll send you more stuff.