Where should a beginner start with AI?

The beginner problem is usually too many choices, not too little information. Start with one assistant and a few real tasks.

What the data says about beginners

• A Gallup survey lists “unclear use case or value proposition” as the most common AI adoption challenge.
• Only 30% of employees say their organization has either general guidelines or formal policies for using AI at work, according to Gallup.
• An arXiv audit of generative search engines found that on average only 51.5% of generated sentences are fully supported by citations and only 74.5% of citations support their associated sentence

Step 1: Pick one assistant

Choose one chat assistant and stick with it for a couple of weeks. If your company has provided Microsoft Copilot, use it first because it is likely covered by your company’s data protections. Otherwise Claude and ChatGPT are common general-purpose choices. See which AI tool to use at work for a fuller decision guide.

Step 2: Check your company’s rules

Before pasting anything from work, find out whether your employer has an AI policy or approved tools. If not, keep to non-confidential material until you know. See whether your company needs an AI policy and whether it is safe to put company data in ChatGPT.

Step 3: Try five tasks

Try each of these this week, using real but non-sensitive material:

1. Ask it to summarize a long article or report you already read, so you can judge accuracy.
2. Paste a rough email draft and ask for a shorter version in your tone.
3. Ask it to explain a spreadsheet formula you inherited.
4. Ask for a checklist for a process you know well and see what it misses.
5. Give it messy meeting notes and ask for decisions, owners and open questions.

Notice what it gets right and where it invents or omits.

Step 4: Learn the prompt basics

A useful structure has four parts: who you are and the situation, the task, the format you want, and what to avoid. The Anthropic Claude prompting guide says the more precisely you explain what you want, the better the result. For example: “I am a finance manager writing to a vendor who missed a deadline. Draft a firm but polite email under 120 words. Do not threaten to cancel.” Details are in how to write better prompts.

Step 5: Build the checking habit

Microsoft’s own documentation says Copilot responses are not guaranteed to be 100% factual. Verify names, numbers, dates and quotations against the original. If the assistant cites a source, open it. The rule is simple: the more it matters, the more you check.

Step 6: Only then, go further

After a few weeks, look at features like uploading files, saving reusable instructions, and eventually tools that act on your files or email. The guide to Claude Code for non-programmers is a reasonable next step.

What beginners often get wrong

• Trying to learn everything before using anything.
• Using it for a big, high-stakes task first.
• Deciding it does not work after a vague prompt.

If you prefer to learn with an instructor, Learn AI Fast begins with foundations on the first morning and assumes no prior AI experience.

Sources

• Gallup, AI use at work has nearly doubled in two years: “unclear use case or value proposition”
• Gallup, AI use at work has nearly doubled in two years: “30% of employees say their organization has either general guidelines or formal policies for using AI at work”
• arXiv, Evaluating Verifiability in Generative Search Engines (Liu et al.): “on average, a mere 51.5% of generated sentences are fully supported by citations and only 74.5% of citations support their associated sentence”
• Anthropic, Claude prompting best practices: “The more precisely you explain what you want, the better the result”
• Microsoft Learn, Data, Privacy, and Security for Microsoft Copilot: “aren’t guaranteed to be 100% factual”