A one-page map

Where are you with AI at work?

Four levels, the signs of each, and a free lesson to start from

Everyone at work seems to be using AI, and it's hard to tell where you stand. Some colleagues bring it up in every meeting, others never mention it, and most of the advice going around boils down to using it more, which doesn't tell you what to do on Monday morning. This page splits that advice into four levels you can spot in your own week, each with an ordinary work example and a free lesson to start from.

The map

Four levels of using AI at work

  1. Level 1SpeedThe work you already do, done faster.See the signs
  2. Level 2AltitudeAsk for the result you need and let the AI work out the steps.See the signs
  3. Level 3DepthDo the task the way the best person in the field would, and check the result against something other than the AI.See the signs
  4. Level 4RangeBorrow methods from fields your task doesn't obviously need.See the signs

Read the signs under each level and notice which ones sound like your week. Most people sit in more than one row, and there's no score to add up. A good place to start is the level just past the one that feels familiar.

Where the four levels come from

The levels come from a post by Felix Rieseberg, an engineering lead at Anthropic. His point is that being told to be more ambitious with AI is common advice and hard to act on, so he breaks ambition into four kinds: speed, altitude, depth and range. He writes that the approach works with any AI model. The descriptions on this page are our paraphrase, and the office examples are ours.

Read Felix Rieseberg's post

The free lessons linked below use Anthropic's Claude. The habits they teach carry over to other AI assistants.

Level 1

Speed

The work you already do, done faster.

You're here if

  • You use AI to draft emails, recaps or summaries you'd otherwise write from scratch.
  • You give it one small task at a time and fix the result yourself.
  • You've barely tried it and want a low-risk first step.

At work it looks like

After a meeting, you paste your rough notes and ask for a short recap for the team, with owners and dates. You spend your time reading and correcting the draft instead of typing it.

Level 2

Altitude

Ask for the result you need and let the AI work out the steps.

You're here if

  • You split a job into small requests and stitch the answers together yourself.
  • The work will be discussed in a meeting, which usually means someone needs a decision out of it.
  • You know what the deliverable is, and you're less sure what the people reading it need to decide.

At work it looks like

Your quarterly review. At the speed level, the AI fills in last quarter's template. At this level you ask what changed, why it changed and what to do about it, then ask for the three decisions that matter most, with the case for and against each one.

Level 3

Depth

Do the task the way the best person in the field would, and check the result against something other than the AI.

You're here if

  • AI answers look polished, but you couldn't say whether an expert would sign off on them.
  • You've had an AI summary that sounded right and got a number wrong.
  • Your team holds a belief about the business that nobody has tested against the actual data.

At work it looks like

A spreadsheet of monthly sales by region. Ask how an experienced analyst would approach it and which raw figures they'd start from. Then ask for the exact numbers behind every conclusion and check them against the spreadsheet yourself. The AI checking its own answer doesn't count as a check.

Level 4

Range

Borrow methods from fields your task doesn't obviously need.

You're here if

  • Every problem you work on ends up in the same format, the same deck or the same memo.
  • You want someone to find the holes in a plan before your manager does.
  • You've wondered how an editor, an auditor or an investor would handle the task in front of you.

At work it looks like

A long email thread is heading toward a decision. Before it lands, ask the AI to run a pre-mortem on the plan, the way a project team assumes it has already failed and asks why, and then to look for weak spots the way a skeptical investor reads a company's numbers. Ask it to do what that person does. Asking it to sound like them only changes the tone.

An honest limit

The higher levels work best when the AI can reach the files and data the work depends on, and many workplace setups don't allow that yet. Check what your company permits before you upload anything, and keep the final say on anything that goes out under your name.

What's free

Start free, no card needed

  • Try the first 2 lessons of a course without an account.
  • Create a free account and your first 3 finished lessons are free.
Start the free first lesson

Find your level, then try one lesson

The free first lesson takes about 6 minutes.

Start the free first lesson

If a colleague is asking the same question, send them this page.

Nightschool AI is independent and is not affiliated with or endorsed by Anthropic, Microsoft, OpenAI, Google or Meta. Product names on this page only identify the tools our lessons teach. Felix Rieseberg's post is credited as the source of the four levels; he did not review this page.

Nightschool AI is an independent learning platform and is not affiliated with, endorsed by, or sponsored by Anthropic. Claude is a trademark of Anthropic, PBC. Looking for Anthropic's official Claude Academy? It's at academy.claude.com.