What is the fastest way to learn AI for work?

People who get good at AI quickly tend to do the same few things. They use it on work that matters to them, they pay attention to why an answer was bad, and they keep a record of prompts that worked. Nothing about that requires a technical background.

A two-week plan

Days 1 to 3: pick tasks, not tools. List five tasks you repeat each week that involve reading, writing or spreadsheets. Choose the two that annoy you most. These are your practice material; generic tutorials are a poor substitute.

Days 4 to 7: do the tasks with one assistant. Choose one mainstream assistant (Claude, ChatGPT or Microsoft Copilot if your company provides it) and stay with it. Switching tools every day is the most common way to learn slowly. For each task, give it the input, say what a good output looks like, and say what to avoid.

Week 2: fix what went wrong. For every bad result, work out why. Was the instruction ambiguous? Did it lack context? Did it need an example? Add that to your instruction and try again. Our note on writing better prompts covers a simple structure.

End of week 2: save what works. Keep a document of instructions that produced good results. This becomes your personal prompt library and is worth more than any certificate.

What the research suggests

• In a study of 5,179 customer support agents using a generative AI assistant, productivity measured as issues resolved per hour rose by 14% on average, including a 34% improvement for novice and low-skilled workers. The less experienced workers gained the most, which is encouraging if you are starting out.
• Anthropic prompting guidance says examples are one of the most reliable ways to steer Claude output format, tone, and structure. This is why the plan above has you paste in an example of a good output.

What to skip at first

• Model comparisons and benchmarks. They change monthly and rarely affect your daily work.
• Prompt “hacks”. Clear writing beats tricks.
• Building anything complex. Learn to get good answers before you try automation.

Where to go after the basics

Once you can reliably get good drafts and summaries, the next jump in value comes from automation: having the AI perform steps for you rather than just answering. That is where tools such as Claude Code, Codex and Microsoft Copilot in your Office apps come in. The guide on automating Gmail with Claude shows an example.

Structured help versus going alone

You can absolutely learn this alone, and many people do. The cost is time spent on dead ends, and not knowing what you have not tried. A structured course compresses that. Options include online video courses, live online workshops and in-person classes; the trade-offs are laid out here. What matters is that whichever you choose has you working on your own material and gets feedback on your output.

A caution about speed

Learning to use AI fast does not mean trusting it fast. Verify numbers, names and citations before they go anywhere important, particularly in the first month when you do not yet have a feel for where the tool fails.

If you would rather compress the learning into two days, Learn AI Fast is a two-day in-person class in Boston where you build automations on provided case files for email, spreadsheets and documents with an instructor in the room.

Sources

• NBER, Generative AI at Work (Brynjolfsson, Li and Raymond), Working Paper 31161: “issues resolved per hour, by 14% on average, including a 34% improvement for novice and low-skilled workers”
• Anthropic, Prompting best practices: use examples effectively: “Examples are one of the most reliable ways to steer”