The "ADAPT" framework is very interesting, professor Mosis. I am curious to your opinion about what Jen-Hsun Huang has explained recently: The amount of AI-credits on the vacancy as a promotional plus on the secundary conditions (next to days of holiday, vitality budget etc.). How does the amount of AI-credits play a role in the "ADAPT" framework? Personally, I am sometimes already out of credits in the exploration phase to see what AI can and what it can't before I embed it in my daily work.
Biem, you've spotted something most leadership teams are still missing, and it reveals a structural mismatch in how compute is being allocated.
Huang's insight holds: treating AI-credits as a secondary employment condition is smart economics. It puts a price on scarcity and forces deliberation about use. But there's a catch that only shows up if you track the ADAPT stages.
Autodidactic stage learning can't be squeezed. It's the stage where intrinsic motivation gets built. Ration it too heavily, and you don't get cost savings. You get a 95th percentile organization that never figures out how to move past casual dabbling.
Your specific experience matters: if you're running out of credits during exploration, the credit system has been designed for production efficiency, not learning velocity. Most organizations I work with show the same pattern. They allocate something like 10% of compute budget to discovery and learning, 90% to production. It should be flipped for the first 18-24 months of adoption.
The research data is clear. KPMG found only 5% of users hit high-impact use. In every audit I've done, that 5% correlates with teams that had enough exploration budget to move through Autodidactic and Didactic without rationing.
I'd tell Huang this: if the goal is real adoption, not just license penetration, then AI-credits allocated to learning aren't a cost. They're the foundation of capability. The organization that lets you exhaust your exploration budget is the one that reaches transformation.
The "ADAPT" framework is very interesting, professor Mosis. I am curious to your opinion about what Jen-Hsun Huang has explained recently: The amount of AI-credits on the vacancy as a promotional plus on the secundary conditions (next to days of holiday, vitality budget etc.). How does the amount of AI-credits play a role in the "ADAPT" framework? Personally, I am sometimes already out of credits in the exploration phase to see what AI can and what it can't before I embed it in my daily work.
Warm regards, Biem
Biem, you've spotted something most leadership teams are still missing, and it reveals a structural mismatch in how compute is being allocated.
Huang's insight holds: treating AI-credits as a secondary employment condition is smart economics. It puts a price on scarcity and forces deliberation about use. But there's a catch that only shows up if you track the ADAPT stages.
Autodidactic stage learning can't be squeezed. It's the stage where intrinsic motivation gets built. Ration it too heavily, and you don't get cost savings. You get a 95th percentile organization that never figures out how to move past casual dabbling.
Your specific experience matters: if you're running out of credits during exploration, the credit system has been designed for production efficiency, not learning velocity. Most organizations I work with show the same pattern. They allocate something like 10% of compute budget to discovery and learning, 90% to production. It should be flipped for the first 18-24 months of adoption.
The research data is clear. KPMG found only 5% of users hit high-impact use. In every audit I've done, that 5% correlates with teams that had enough exploration budget to move through Autodidactic and Didactic without rationing.
I'd tell Huang this: if the goal is real adoption, not just license penetration, then AI-credits allocated to learning aren't a cost. They're the foundation of capability. The organization that lets you exhaust your exploration budget is the one that reaches transformation.