All guides

A 90-Day Plan to Learn AI Skills for Work

By James Finley. Published Oct 14, 2026.

This 90-day plan is for people with full-time jobs and no technical background. You need a few hours a week and real tasks to practice on.

The plan at a glance

Plan on 3 to 4 hours a week. For example, two 45-minute sessions on weekdays and one longer session on the weekend. If a week falls apart, do the short sessions and keep going. Don't start over.

Weeks 1 to 4: learn the basics first

Four skills come before anything else, whatever your job:

  1. Clear prompts. Say what you want, who it's for, what the AI should use, and how the result should look. OpenAI's and Anthropic's prompting guides both start here: be specific and give the details the AI can't guess.
  2. Checking the answer. AI can sound sure and still be wrong. OpenAI's help center says to verify quotes, data, technical information and references before you rely on them.
  3. Knowing your tool. Learn where it keeps files, memory and settings, and what your plan includes.
  4. Keeping private data out. Don't paste customer records, pay, health details or passwords into an AI tool your company hasn't approved.

Pick one AI tool and stay with it for the whole 90 days. Switching tools every week means starting over every week.

Practice: each week, rewrite three real emails or documents with AI, then compare its version with yours.

Weeks 5 to 8: practice the tasks your job has

The basics are the same for everyone. What you practice next depends on your role.

Practice: start a prompt library. Save each prompt that worked, with a note on what you had to fix.

Weeks 9 to 12: use it on real work

Each Monday, pick three tasks already on your calendar: an email to write, a report to summarize, a meeting to prepare for. Do each one with AI, then check the result the way you'd check a new coworker's first draft.

Try this with a long email thread: "Summarize the decisions made, list the open questions, and draft a reply confirming next steps." Compare it with what you would have written, and fix the prompt wherever it guessed.

Two habits make answers more reliable. Anthropic's guide on reducing hallucinations suggests telling the AI it's fine to say "I don't know," and asking it to quote the part of your document each claim came from. Anthropic also says these steps reduce mistakes but don't remove them, so you still check.

How to know a skill is ready for work

Before you rely on AI for a task, ask yourself four questions:

  1. Speed: can I do the task with AI in about the same time as without it, or faster?
  2. Checking: can I spot when the answer is wrong, without asking someone else?
  3. Variation: can I write a good prompt for a similar task I haven't practiced?
  4. Explaining: could I show a coworker my steps and say why I trusted the result?

If the answer to question 2 is no, keep practicing before you use AI on work that leaves your hands.

Doing it as a team

For a team of 3 to 10 people, give each person one role track and add two habits. Hold a 30-minute weekly show-and-tell where each person shares one prompt that worked, and keep a shared prompt library. Ask one person to own a one-page AI policy that says which tools are approved and what data stays out.

Where a course helps

You can follow this plan on your own. A course adds structure and feedback: exercises with a checklist and a model answer, so you can see what good looks like. Our AI Starter Pack path covers weeks 1 to 4 with AI Foundations and Prompting Fundamentals. For weeks 5 to 8, see AI for Marketing Foundations, AI for HR and Operations or AI for Managers and Team Leaders. Each of these courses has a free preview lesson, with no account needed.

Related courses

Sources