Is AI hard to learn? No. Here’s your first 90 days

It’s not hard to start learning AI. You don’t need a technical background to ask an assistant to draft an email or organize your notes. You type what you want in plain English.

Getting useful results is what’s challenging. If you’ve ever opened a chatbot, gotten a bland answer, and wondered what to try next, another list of tools won’t help, but a structured approach just might.

Here’s how we’d structure your first 90 days in six steps, from writing one prompt that works to a repeatable method for work you already do. Every step has a checkpoint, so progress means more than feeling comfortable with a chatbot.

No, AI is not hard to learn

Start with a small task you already know how to do. Hand it to your tool of choice, look at what comes back, and judge it yourself. That’s the easy part. What takes practice is explaining what you need, spotting a weak answer, and knowing when the tool isn’t the best option.

An April 2026 report from the Federal Reserve Bank of New York offers some perspective. In its November 2025 survey, about 66% of employed respondents who had used AI at work in the previous year reported higher personal productivity, and 40% said it helped them finish tasks faster. But 19% said they were still learning and tasks took longer as a result.

Those are self-reported experiences, not a measure of how easy AI is to learn or how quickly you’ll improve. Give yourself time to practice, then check whether your work is improving and whether the time spent prompting and reviewing is worthwhile.

Three distinct paths to learning AI

Your learning journey depends entirely on your goals. Operating an AI assistant requires a vastly different skill set than developing a model from scratch. Clarify the type of work you intend to do before picking a starting point.

Comparing AI to traditional computer science degrees is only relevant if you plan to build models. For practical workplace tasks, like drafting emails or organizing updates, the steps below provide a solid foundation.

Why you don’t need to code to learn AI

Using AI and building AI are different jobs. To use a mainstream AI assistant, you don’t need to understand how the model was trained, write software, or manage the systems behind it. All you need to do is explain the work, provide the information the tool should use, and judge whether the response is suitable.

Coding only matters when you want to build an AI product, create a custom integration, or change how a system behaves. It is not a prerequisite for asking an assistant to draft a customer response, turn approved notes into a status update, or help you compare options. The technical work happens inside the tool; your work is deciding what should happen before and after it.

Danoosh Kapadia, an instructor in our AI Essentials course, describes prompting this way: “It has a lot more to do with communication skills than technical skills.”

Why AI can feel hard at first

Knowing you can start without coding doesn’t tell you what to do on Monday morning. If you’re collecting tutorials but never getting to practice, make the next decision smaller: one task, one assistant, one result to review.

Give your learning a sequence

A tool tutorial and a prompting video can each be useful, but neither helps you make progress. Choose material that helps with the task in front of you, test the process, then decide what you need next. You don’t need to research AI before you use it.

For example, if you want help with a weekly update, begin by learning how to turn notes into a draft. Save the tutorial about connecting several apps until you have a reason to connect them.

Check what support your workplace provides

In the same New York Fed survey, 15.9% of employed respondents said their employer offered AI training. That doesn’t tell you what your own organization provides, so check its approved tools and ask about training before choosing a personal account for work.

If workplace use isn’t allowed, you can still practice with fictional tasks or public information on your own time, and you can still make progress without uploading company material.

Your first 90 days: how (and where) to start learning AI

We recommend starting with one familiar task, then carefully rolling out what you learned to other tasks. Later, you’ll map a process you know and document a method someone else can follow.

The schedule below is our suggested practice plan. The 90 days and time budgets are planning estimates, not measured learning times or promises of competence. Move at a pace that lets you meet each checkpoint.

Want something more structured? Try our AI Essentials or AI Automation courses.

Where possible, use these sessions for work already on your calendar. Count the time you spend checking and correcting the output as part of the practice.

Step 1: Pick one AI assistant and set it up properly

In week 1, choose an AI assistant that your workplace permits and make a habit of using it regularly. You can compare alternatives later, once you know what you need.

For practice, use fictional or public material. Keep confidential business information and personal data out of an unapproved tool; the UK National Cyber Security Centre advises against putting sensitive information into public AI prompts. A paid account alone doesn’t establish permission to use workplace data.

Write a short background note describing your role in general terms and how you prefer responses to be written. If your assistant offers personalization settings, you can enter relevant details there; otherwise, save the note to reuse it in a new conversation. Add task-specific details each time, even when you’ve saved your preferences.

Milestone: you have an approved tool, know what information you can use, and have a background note ready for your next task.

Step 2: Practice structured prompting on real work

During weeks 1–4, attach your practice to a recurring task, such as a weekly status update or analysis.

Keep a simple record alongside the task:

• The prompt you used and what you wanted back.
• What you had to correct, including omissions or invented details.
• Total time spent, compared with doing the task yourself.

When the output misses something, identify the specific problem before trying again. If a status update leaves out information, ask the assistant to include it and explain why it matters. If it invents a deadline, correct the draft and add an instruction to flag missing information rather than fill it in.

Milestone at day 30: you can produce a useful draft for one recurring task and explain how you checked it. If you’re still rewriting most of the result, try a smaller task or ask a colleague to review the prompt with you.

Step 3: Decide which of your tasks AI should touch

In weeks 5–6, review your typical workweek and decide where to expand. Repetition alone doesn’t make a task suitable: you also need to understand the consequences of a mistake and be able to review the work.

In his guidance on when to use AI, Ethan Mollick recommends using it in work where your expertise allows you to assess the output and cautions against uses that require very high accuracy. For this plan, use these four questions to sort your tasks:

• What would I ask AI to do, and can I separate that part from the rest?
• What would happen if the answer was wrong?
• Who can check the result against reliable information?
• Does it improve the work or save time after review and corrections?

Drafting a routine update from approved notes is a reasonable candidate if you can check every fact. A decision about an employee’s performance or a contract’s legal meaning needs the responsible person’s judgment and the appropriate review process. Keep those decisions out of your beginner practice list.

Milestone: name two or three tasks to try, plus at least one to leave out and the reason why. Apply the same questions whenever a new use comes up.

Step 4: Move from first drafts to thinking work

In weeks 7–9, try asking the assistant to help you examine an idea. You can use it to compare options or identify assumptions in a plan, while keeping responsibility for the conclusion.

Suppose you’re deciding whether a routine team update needs a meeting. Describe the purpose using non-sensitive details, then ask:

Compare a written weekly update with a 20-minute meeting for this purpose. Explain what each option makes easier and harder. Identify assumptions in my description and questions I should ask the team before deciding.

Review the comparison against what you know about your team. A plausible objection is something to investigate, not evidence of what your colleagues think. Check factual claims against their original sources, and use your established calculations to verify any numbers.

Milestone at day 60: you have reviewed the comparison or critique and can explain what you accepted, changed, or rejected. Continue practicing through week 9 if you need more experience before moving on.

Step 5: Put your own expertise to work

During weeks 10–11, choose one process you know well and map it from start to finish. Your experience helps you notice missing context, exceptions, and approvals that a tidy AI-generated plan can overlook.

For a weekly project update, your map might look like this:

• You gather approved notes and confirm the current status.
• AI helps arrange the notes into a draft update.
• You check owners, dates, and unresolved issues against the source material.
• You decide what needs escalation and send the approved update.

Mark the information needed at each stage and who reviews the result. Try the process manually before considering any connections or automatic actions; mapping it doesn’t require building an automated system.

Milestone: you have one process map showing where AI assists, where you review, and which decisions remain with you.

Step 6: Make it useful to people other than you

In weeks 11–12, document the method you’ve found most useful. Choose one that has held up across repeated attempts, including its limitations.

Keep the instructions short enough for a colleague to try. Include:

• When to use the method and when to avoid it.
• The approved tool and permitted input material.
• A reusable prompt and a fictional or shareable example.
• The checks to run before using the output.
• What to do when the result is wrong or takes too long to fix.

Ask a willing colleague to try it and note where they need clarification. At day 90, review their feedback alongside your own practice record and update the method. Sharing something useful makes your progress visible without depending on a particular reaction from your team.

Milestone at day 90: someone else can follow your instructions, and you can explain the method’s benefits and limits. Choose your next task from there.

Where to start this week

You can follow this plan on your own or add support where you’d find it useful. Choose based on your schedule, budget, and how you prefer to learn.

Free self-directed learning

Choose one tutorial, test its advice, and keep what works. Sandeep Swadia’s beginner AI walkthrough introduces choosing an assistant and giving it structured instructions. Use those exercises as a starting point; our checkpoints above are a way to assess your own progress.

The tool’s own guides

Use the provider’s current help pages to understand a feature or setting. Check that the instructions match your account and the tool’s version, then return to your practice task. Documentation is useful alongside a plan you choose yourself.

Learning with a colleague

Agree on 15 minutes each week to compare a prompt that worked and one that didn’t. Bring examples you’re allowed to share and discuss what you changed after reviewing the answers. You can adapt the plan together around the work you actually do.

A structured cohort program

Our AI Essentials course pairs self-paced video coursework with live sessions and a cohort of fellow learners. It’s a paid option for people who want a set sequence and instructor access. Alternatively, self-guided study is a practical choice if you prefer to select your own materials and pace your practice independently.

Whichever route you choose, start with one small task this week. Write down what a useful result would look like, try the prompt, and check the answer against that standard before using it.

FAQs about learning AI

Is AI harder to learn than computer science?

Using an AI assistant for everyday tasks doesn’t require studying computer science. Building AI models is a technical pursuit that involves programming and mathematical foundations; if you’re choosing between academic programs, compare their actual coursework rather than treating AI as a single, fixed level of difficulty.

Do I need to know how to code to learn AI?

No, you can use AI assistants through ordinary language. Start by giving clear instructions and relevant context, then checking the result; coding becomes relevant if you choose work that requires it, such as developing models or custom integrations.

How long does it take to get comfortable using AI at work?

You may start to get comfortable within a matter of days. Use the 90-day plan as a suggested schedule, then adjust it to your tasks and experience. It isn’t a measured average or a deadline: a useful checkpoint is whether you can get a verifiable result without spending excessive time correcting it.

I’m in my 50s and not technical. Is it too late to start?

No. You can begin with a task you already understand, and your work experience can help you judge the answer. There’s no need to learn programming before trying the exercises here.

What’s the first AI tool I should use?

Start with an assistant that your workplace permits for the task and information involved. For personal practice with fictional or public material, a free account can be a starting point; check its current limits and stay with one tool long enough to learn how it responds.