A 30-day AI plan for office workers: 15 minutes a day, on your real work
The biggest trap in learning AI is treating it like a course you have to find extra time for.
What actually works is the opposite: use AI directly on the work you already have to do. This 30-day plan is built on that idea. Fifteen minutes a day, all spent on real tasks.
How to use this plan
- One task per day. Use something you genuinely need to get done this week, not a made-up exercise
- It's fine if you fall behind. The order matters more than the pace
- Every time a prompt works well, save it to your notes right away
Week 1: Say what you mean
Goal: Learn to give enough background that AI gets it right the first time.
| Day | What to do today |
|---|---|
| 1 | Pick one main AI assistant — just one. Ask it three questions you're actually thinking about today |
| 2 | Re-ask yesterday's worst question using "role × task × context × format," and compare the results |
| 3 | Find an email you've been struggling to write and ask for three versions in different tones |
| 4 | Find a long document you haven't had time to read and ask it to pull out three key points "for someone in my role" |
| 5 | Ask it to ask you 3–5 questions before answering, and see what it's like when it checks first and then gets to work |
| 6 | Deliberately ask a question in an area you know well, and find where it gets things wrong |
| 7 | Review: save the three prompts that worked best this week as templates |
This week's mindset: You're not "giving commands." You're handing off work to a new coworker.
→ Related reading: 6 principles for writing good prompts
Week 2: Tackle what you really do every day
Goal: Find your most frequent tasks and let AI take the first draft.
| Day | What to do today |
|---|---|
| 8 | Turn a meeting recording into a transcript, then ask AI to organize it into decisions plus a to-do list |
| 9 | Take some messy data and ask it to organize it into a Markdown table |
| 10 | Try the "give an example" technique: paste three good pieces you've written and ask it to match them |
| 11 | Ask it to play three different roles and critique a proposal you're working on |
| 12 | Boil a long document down to a version your manager can read in 60 seconds |
| 13 | Pick a chore you don't want to do and try handing the whole thing over |
| 14 | Review: estimate how much time you saved this week, and what you did with it |
This week's mindset: Look for tasks that happen every single week. That's where your effort pays off.
→ Related reading: Use Cases at Work and Life
Week 3: Build your judgment
Goal: Know when not to trust it, and what information must never be pasted in.
| Day | What to do today |
|---|---|
| 15 | Ask it for three sources, then open each one to check |
| 16 | Learn to push back: "Which parts are you confident about, and which are guesses?" |
| 17 | Add "If you're not sure, say you don't know" to all your templates |
| 18 | Check the settings of the AI service you use and confirm its data-use policy |
| 19 | List the information from your job that must never be pasted into AI |
| 20 | Practice anonymizing: turn a real document into a version with code names, then use that |
| 21 | Review: write down three things you won't hand to AI, and why |
This week's mindset: AI handles drafts and speed; you handle the facts and the responsibility. Your name is the one on it.
→ Related reading: Hallucinations and fact-checking · Safety & Judgment
Week 4: Turn one-offs into something repeatable
Goal: Lock in what worked as templates and processes.
| Day | What to do today |
|---|---|
| 22 | Go through the prompts you've saved this month and sort them into 5 folders |
| 23 | Rewrite your three most-used prompts as fill-in-the-blank templates |
| 24 | Try a "custom assistant" or "project" feature and set up your standing rules in it |
| 25 | Upload documents you refer to often and build a small knowledge base |
| 26 | Use the knowledge base to answer a question, and compare it with a regular chat |
| 27 | Pick something you do every week and map out its steps |
| 28 | Evaluate: is it worth automating? (Happens weekly + clear rules + mistakes are easy to fix) |
| 29 | Turn one task into a standard process and write down the step-by-step routine |
| 30 | Final review: how many hours did you save this month? What do you want to tackle next month? |
This week's mindset: Templates give you a far better return than automation — and they don't need any technical skills.
→ Related reading: Workflows & Automation
Where you'll be after 30 days
If you really work through it, you should have:
- Your own prompt notes, ready to copy whenever a similar task comes up
- At least two routine tasks that now take noticeably less time
- A reflex to check any numbers or citations AI gives you
- A clear sense of what information you can't paste in, and why
- A knowledge base, or one process that runs the same way every time
That's pretty much everything in stages 1 and 2 of the Roadmap.
One last reminder
After a month, look back and ask one thing: what did you do with the time you saved?
If the answer is "watched more AI tutorials," you've got it backwards. The point of learning AI is to give you time back for the things that matter more.
Want to start right now? Begin with the 20-minute version in Get Started, and you'll have your first result today.