The five levels of working with AI, and why most teams are stuck at level two

Forbes contributor TerDawn DeBoe laid out five distinct levels of working with AI. Worth knowing about.

Most people assume there's really only one level. You type something, you get something back and getting better just means writing clearer questions.

That's the first rung and when done well, feels pretty good. However, the biggest gains sit several rungs higher and most people never look up.

The five levels that TerDawn shared.

Level one is casual conversation. You use AI the way you'd use a search engine or talk to an expert or even to brainstorm out loud. No real structure, and likely no expectation of a repeatable result.

Level two is a better prompt. You add examples, more context, clearer instructions. This is where a lot of people plateau and where most business use of AI actually lives today. Why? Because it feels like real progress.

Level three is a saved template. You've written something good enough to reuse, so you keep it in a doc or a notepad and paste it in each time. This is definitely a real improvement over level two, but it's fragile. The template only works as well as whoever's pasting it and never seems to work the same for others "trying it out." Usually leading to further edits that take the results in a completely different direction.

Level four is a structured skill. Things like defined inputs, a fixed methodology, a defined output. This is the level where the format actually changes and not just the phrasing. Instead of a paragraph someone has to interpret and adapt correctly, there are specific fields and the instructions that made the original prompt good are built in rather than left for the next person to reconstruct.

Level five is autonomous agents that are capable of multi-step workflows without a person directing each step. Genuinely useful for a narrower set of problems than most of the hype suggests. Frankly, these are still early and perhaps not where most teams should be spending their first effort.

Why the gap between two and four matters more than it sounds

The distance between two and four isn't about AI sophistication. It's about what happens when you're not the one running it.

A great level-two prompt still only works well for whoever wrote it. Remembering the right phrasing. Or the right context. The right follow-up if the answer's off. None of that transfers easily.

A level-four skill removes the dependency. The fields ask for exactly what's needed. Whoever runs it gets the same quality result, whether or not they know how to prompt well themselves.

That's the actual gap. Not "we need better prompts." It's we need what made our prompts good to stop living only in the head of whoever wrote them.

Why most organizations never make the jump

A McKinsey survey found that 88% of organizations now use AI in at least one part of the business, but only around 6% see real, significant bottom-line impact from it. That gap isn't a sign AI doesn't work. It's a sign that most organizations are using it at level two, where the ceiling is real but quite low, while a small number have actually restructured how the work gets done, which is where level four and five start to pay off.

Using AI more isn't the same as using it differently. Adoption without a change in format plateaus. The organizations seeing real impact aren't necessarily using AI more often, they're using it in a shape that doesn't depend on one person's phrasing every single time.

Where to actually start

Don't try to jump straight to level five. Pick the one task you or your team repeats most often at level two or three right now, the prompt you keep rewriting slightly differently, or the template that only really works when you're the one pasting it in. That's the candidate for level four: write down what's fixed about it and what changes each time, and you've got the beginning of a real skill instead of another prompt someone else will have to guess at.

And don't fall into what TerDawn calls the "Lazy Tax": rewriting the same question from scratch every time, forgetting by next week how you got a good answer and never actually reusing anything. The cost isn't the time to write one question. It's paying that cost forever, on the same task, again and again.

Skillyo is built specifically for the jump from level two or three to level four: turning a prompt you keep rewriting into a skill with defined fields and a consistent result, runnable by anyone on your team. Worth five minutes to see what that looks like for the thing you repeat most.