Knowledge Worker AI Onramp

This post is for my normie friends.

My guess at what your AI onramp will look like:

  1. Copy/Paste: You copy/paste with chat
  2. Copilot: You give AI access to read/write where you are
  3. Autopilot: You gain trust with outputs for X, flip switch from copilot to autopilot
  4. Systems Thinking: Add more tools, more access, flip more switches

1. Copy/Paste#

You've tried at least one of the 3 major AI chat bots from Anthropic, OpenAI, or Google.

You've copy-pasted some text in there to have it sanity check you on this and that.

You've even realized that they are each a lil or a lot different.

You've seen that different models within a provider have different qualities of intelligence, say Haiku, Sonnet, Opus for Anthropic.

You've had it do some research for you, again finding out that some do better than others.

You might even pay for the entry point pricing for on of these providers.

You have caught some errors, some mistakes, some good info.

You're annoyed about all the copy-pasting and wondering if there is a better way...

2. Copilot#

You've got the AI chat app on your phone.

You realize the desktop app can do things the web version can't, like touch files on your computer.

You do some research have it output a CSV/doc file.

You connect it to your gmail and have it summarize all emails since you checked yesterday.

You've heard of skills/mcp/openclaw and don't know what any of it means yet.

You are starting to understand what is meant by "context window".

You have saved 5min here and there, sometimes even 30min on some tasks as your copilot handles parts of a greater task.

You start wiring more connections and tools into the app so it has more information to work from without you needing to stuff everything into each chat.

You setup some daily tasks to run on your local computer via the desktop app.

You start to really like its voice in replying to very specific requests that happen frequently.

You're wonder why you need to approve send every time for the simple stuff that you trust to get right every time now...

3. Autopilot#

You're trusting it on tasks of this type and that complexity.

You start to look for ways to have it just do those things.

You have figure out how to get the agent to close the full loop on that task - that one task is just on autopilot now.

You watch a youtube video about running your agents tasks in the cloud.

You have the desktop app drive your browser.

You're thinking about the necessary context required for any agent to complete a given task.

You're wondering how to get that context into the agent, you look for the right connectors to get the organizations info into your agent so it has everything it needs.

You start having AI play some copilot role in almost every task you have.

You have a few tasks on autopilot now.

You start to see what might be required to automate more things...

4. Systems Thinking#

You are starting to see how you need to gather up certain context across the org in new ways in order for the agent to have the information it needs. Things that were just institutional knowledge before now needs to be exposed to the agent.

You've deliberately chosen not to automate x, y and z. Cause friction is sometimes a feature. And there are things you should hold in your head to feed your strategic thinking.

You thoughtfully re-arrange your workday so that you can have thoughtful focus time for the things that require your best thinking.

Where are you on this path?