From zero to daily AI user in 30 days
Four themed weeks, one concrete task per weekday, about 20 minutes each. No signup, no course fees, no weekend homework.
Most people never build the habit because they try to learn everything at once. This plan does the opposite: one small task per day, each week with a single theme. Week 1 is your tool, Week 2 is prompting, Week 3 is your real work, Week 4 is small agents. Work weekdays; rest on weekends. Copy the 20 tasks into a note file and check each one off.
Week 1 — Pick your tool and feel safe
- Day 1: Open one free browser-based AI chat tool and ask it to explain your own job in ten lines.
- Day 2: Learn your rules — which of your company's data may go into the tool and which may not.
- Day 3: Paste a real, non-sensitive document and ask for a five-bullet summary.
- Day 4: Ask it to rewrite one of your own emails in two tones: formal and warm.
- Day 5: Save your three best prompts in a note file called my-prompts.md.
Week 2 — Prompting muscle
- Day 6: Give the AI a role first (you are an experienced HR business partner), then the task.
- Day 7: Ask for output in a specific format — a table, a bullet list, a subject line plus body.
- Day 8: Deliberately write a bad prompt, then fix it and note what changed.
- Day 9: Use the AI to critique your own draft, then argue with it.
- Day 10: Build your first reusable prompt template with placeholders in [brackets].
Week 3 — Your real work
- Day 11: Pick one recurring task (meeting notes, weekly report, inbox triage) and run it with AI.
- Day 12: Do the same task without AI, time both, and record the minutes saved.
- Day 13: Extend AI support to a second recurring task.
- Day 14: Find one spreadsheet chore and solve it with a formula the AI suggests.
- Day 15: Write down your before/after numbers — this becomes your performance-review evidence.
Week 4 — Workflows and agents
- Day 16: Chain two prompts: summarize a meeting, draft the action list, write follow-up emails.
- Day 17: Try a no-code agent builder on one tiny workflow, such as triaging incoming emails.
- Day 18: Connect the agent to your notes app, manually if needed.
- Day 19: Test the agent with an edge case and fix the prompt until it survives.
- Day 20: Present your results to one teammate, then repeat the plan at a harder level next month.
How it works
- Copy the 20 weekday tasks into a note file and block 20 minutes on your calendar each workday.
- Do exactly one task per day — no ahead-of-time bingeing, the habit is the point.
- At day 30, record how many minutes per week the AI-assisted version of your work now saves.
No. Every task uses the free tier of a browser-based chat tool. Week 4 mentions agent builders, but they also have free tiers — you only build the smallest version of one workflow, which costs nothing to try.
Skip it and continue tomorrow. The plan is designed to be 80% complete — missing 4 days out of 20 still gets you the habit and the measurable time saved. Do not restart from day 1.
Any mainstream browser-based chat tool with a free tier. The skills transfer between tools: role prompting, format control, fact-checking and chaining prompts work the same way everywhere. If your company already approved one, use that.
Yes — managers benefit even more, because weeks 3 and 4 turn into delegation and review workflows. Swap the example tasks for your own: status summaries, interview note synthesis, board-pack summarization.
The generator builds a custom plan from your minutes per day and goal. This page is a fixed, battle-tested schedule you can copy right now. Use this if you want to start today; use the generator if you need a plan tuned to 10 or 90 minutes a day.
Who this tool is for
This schedule is built for jobs made of recurring paperwork rather than code: an HR coordinator who runs onboarding every other Monday, an office manager who owns the supply order, a marketing assistant who rebuilds the same performance deck each month. Those roles share one useful property — a predictable weekly rhythm. A task that will happen again next Tuesday is a task you can change next Tuesday, which is exactly what twenty small assignments need in order to stick.
It is a poor fit in three cases. If your week is a series of one-off projects with new scope each time, the daily tasks will keep colliding with real deadlines. If you already write scripts or build automations, weeks 1 and 2 amount to revision. And if you cannot name a single task you perform more than twice a month, find that task before you start; the plan sharpens an existing routine but cannot invent one for you.
How to read your result
What this page produces is not a score. It is a four-week structure plus a short log you keep yourself, so read it in two passes. First, check that each week has a theme and five tasks you could actually perform at your own desk, with the software you already have. A week where day 11 says "automate reporting" without naming the report is a week you will skip.
Second, read your own log at day 20. A usable result looks like this:
- Week 1 ends with a note file containing at least three prompts you rewrote yourself, not three prompts you merely read.
- Days 11 to 15 produce at least two pairs of numbers: the same task timed with and without assistance.
- Day 19 leaves a written edge case and the instruction you changed because of it.
- Day 20 has one colleague who has seen the output and asked a question about it.
If a week produced conversation but no artefact, treat it as unfinished and repeat it rather than moving on to the next theme.
A worked example
A payroll and benefits coordinator at a 70-person logistics company followed the plan in September. Her three recurring items were monthly payroll variance notes, a benefits question inbox, and a Friday status email to the finance lead. She booked 08:40 to 09:00 as a standing calendar block, used day 3 to summarise a six-page policy PDF, day 11 to draft the variance notes, day 14 to solve an employee-ID lookup with =XLOOKUP(A2,Sheet2!A:A,Sheet2!D:D,"not found"), and day 17 to sketch a triage agent for the benefits inbox.
| Day | Task | By hand | With AI | Saved |
|---|---|---|---|---|
| 11 | Monthly payroll variance notes | 38 min | 14 min | 24 min |
| 12 | Same notes, timed a second time | 36 min | 15 min | 21 min |
| 13 | Friday status email to finance | 25 min | 9 min | 16 min |
| 14 | Employee-ID lookup across two sheets | 31 min | 6 min | 25 min |
The numbers are hers, not a benchmark; the point is that by day 15 she had four rows of evidence for a review conversation instead of a vague claim that she "uses AI now". The XLOOKUP row is the one she repeated in October without opening a chat window, because a formula she understands survives a tool change and a lapsed subscription.
Common mistakes
- Batching the week on Sunday. Seven tasks in one sitting removes the spacing that makes a habit, and it breaks the day 12 measurement.
- Jumping to week 4. Agents are the interesting part and the worst place to begin; without weeks 2 and 3 you have nothing reliable to hand an agent as instructions.
- Skipping the timing day. Working slowly on purpose feels awkward, but day 30 is empty without a "before" number.
- Treating the prompt file as optional. Prompts left in chat history vanish once a conversation is cleared.
- Pasting real employee or client data into a tool your company has not approved. Day 2 exists for this; write the rule down before day 3.
- Restarting from day 1 after a missed day. The plan assumes roughly 80 percent completion, so continue with the next task instead.