I use this framework whenever a colleague asks for advice or a quick training on the use of AI at work. I realized I need something that will structure my advice and give me guidance what kind of help a person needs depending on their level of knowledge. When you go through it you will clearly understand where you currently are and what you need to advance to another level.

Level 1: You use Copilot chat extensively

This is where most people currently are (and many will stay). Using Copilot chat to summarize/write emails, brainstorm ideas or prep for meetings is the easiest way to boost personal productivity and quickly improve some outputs of your work. Depending on how the organization decides to set it up – using it to search through company resources (internal policies, accounting rules, process descriptions) can really cut time and unnecessary questions you ask your colleagues, other teams or HR.

More advanced Level 1 users will make a good use of not only Copilot chat but also Copilot assistant in Excel or PowerPoint, in order to find a formula error, prepare some slides or add a more complex formatting that is outside their skillset in that tool.

All that improves personal effectiveness that does benefit the company indirectly, but in my view cannot yet be used to show tangible savings or result in a reorganization.

Level 2: You use AI to improve processes

Here is where AI impact starts to be noticed by others. You were never really into Excel so your files looked messy, but now you use AI to improve/automate some of your processes. A file that took hours to refresh is now almost fully automated and the time you saved on maintenance is redirected to more value-adding activities. Repetitive tasks become smooth, quick, and easy to hand over to another person because it doesn’t take much time to understand what is happening inside of the files (not because they are less complex, but because they are optimized and set up the right way).

Two elements not everyone thinks about at this level:

  • Use AI to make the files simpler and more clear. Don’t stop at adding complex formulas or automating the flow, spend some time on better design as well.
  • Have AI write “refresh manual” or “handover notes” so that you spend less time teaching the new person when you already want to focus on your next career step.

At this level the impact of AI on processes is already strong and visible. Teams can allocate resources more effectively and spend more time on adding value.

Level 3: You use AI to run processes

In my current view, very few will ever reach this level. Here’s where you understand AI so well, that you can actually trust it to run processes for you. Of course, there is a spectrum of solutions (from having short recurring prompts to designing large, multistep agentic workflows), but here you need to be much more comfortable with AI than on previous levels.

Differences between models, optimizing token use, context engineering, hallucinations and evals, RAG architecture and many more are among the elements you need to understand to orchestrate AI to deliver meaningful and reliable outputs.

Here’s where the significant productivity from AI lies. Organizations seem to be still deciding how to unlock it, whether it is through internal AI experts that grow naturally, their Data & Technology structures or external consultants. One thing I know for sure is that we will see more and more dedicated effort spent on designing and implementing AI solutions.