Phase 3 · Claude · Level 1 · Foundations
What Claude is great for
By the end, you'll be able to…
- Recognise five everyday work tasks Claude does well and spot them in your own week
- Write a first prompt for each: document Q&A, careful writing, analysis, learning and beginner coding help
- Judge which tasks are safe to trust as-is and which need checking against a source
Why it matters
A blank chat box is intimidating precisely because it can do so much. This lesson gives you five concrete, real-world jobs Claude does well, with a worked prompt for each, so you can look at your own week and think, that one, Claude could help with, and know how far to trust the result.
What Claude is good at
Claude is a large language model: a system trained on an enormous amount of text so it can read, write, and reason in plain language. That gives it a particular set of strengths. It is comfortable with long documents and large amounts of pasted text; it tends to write carefully and naturally, with a good ear for tone; and it will think a problem through with you rather than just fire back a one-liner. Those three strengths shape everything below.
The trick to getting value isn't learning clever tricks; it's recognising, in your own week, a task that plays to those strengths. Here are five that come up constantly at work, each with a Fernway example you could paste in today and a worked prompt to copy the shape of.
1. Asking questions of a long document
This is where Claude shines brightest. Hand it a long piece of text (a report, a contract, a policy, a set of meeting notes, a sprawling email thread) and then interrogate it. You're not skim-reading twenty pages hoping to spot the clause you need; you're asking a colleague who has just read the whole thing.
Say Priya at Fernway has sent round the hybrid-working policy and you need one fact out of it before a meeting.
I've pasted our hybrid-working policy below. What does it say about using personal devices for company work? Answer in one sentence, then quote the exact line it comes from. [paste the policy]
Why this works: It names the document, asks a single specific question, and demands the exact line, so you get an answer you can verify in seconds rather than a vague paraphrase.
Claude replies that personal, unsecured devices shouldn't be used unless explicitly approved, and quotes the clause. You glance at the quote against the original, and you're done. Because Claude answers from the text in front of it rather than from general memory, this is one of its most reliable uses. The one caveat: it can still misread or overstate, so for anything load-bearing, ask for the quote and check it.
2. Careful, sensitive writing
When a message needs a delicate touch, Claude is a strong drafting partner. It's good at tone: warm without being gushing, firm without being rude. Think of the awkward messages you put off writing: a reply to an upset customer, a polite chase, a piece of bad news, a note declining something without causing offence.
A customer logged a support issue with us two weeks ago and never heard back. They've emailed, annoyed but polite. I want to apologise sincerely without making excuses, explain we're fixing how we track feedback, and offer to sort it this week. Write a warm, professional reply of about 120 words. Don't invent any details I haven't given you.
Why this works: Giving the situation, the tone you want and a length turns a vague 'write an email' into a describable target. The 'don't invent details' line stops it filling gaps with things that never happened.
You'll get a draft you can actually send after a nudge or two ("a little warmer", "shorter", "drop the second paragraph"). It's a drafting partner, not a vending machine: expect to shape it. And always read the final version back to check it sounds like you before it goes out.
3. Making sense of messy analysis
Claude can read data and notes and help you see the shape of them. It won't replace a spreadsheet for exact sums, but it's excellent at "what's going on here?". It spots patterns, flags oddities, and turns a mess into a plain summary.
Fernway's sales export is a good test: inconsistent dates, a region misspelled "Sotuh", blank totals, a missing rep name.
Here's our April sales data as a CSV. Give me: (1) three plain-English observations about how the regions compare, and (2) a list of any data-quality problems you notice (missing values, typos, inconsistent formats) that I should fix before trusting the totals. [paste the CSV]
Why this works: Asking for observations plus data-quality problems separately gets you both the story and the caveats, and naming the checkable columns invites Claude to show its working rather than assert a single number.
Notice the second half. By asking Claude to flag the data problems, you turn its "helpfulness" into a safety check. It'll point out the blank totals and the "Sotuh" typo rather than quietly averaging around them. For any figure you'll actually quote, still confirm it in the spreadsheet itself; Claude is better at the narrative than the arithmetic.
4. Learning and explanations
Claude is patient and never makes you feel daft for asking, which makes it a good explainer. Ask it to explain a term, a process, or a baffling letter, and to pitch it at your level.
Explain what a company's "core hours" policy means and why an employer might set one, as if I've never come across the idea. Keep it to a short paragraph, then offer me one follow-up question I might want to ask next.
Why this works: Setting the level ('never dealt with one') and inviting a follow-up turns a one-shot definition into a back-and-forth you can steer until it clicks.
The useful habit is the follow-up: read the explanation, then ask about the exact bit that didn't land. Keep going until it clicks. Just remember that for anything high-stakes (health, money, the law, employment rights) a clear explanation is a starting point, not a ruling. Claude can sound completely sure and still be wrong; the tool producing plausible-but-false information is called hallucination, and it's why you confirm the things you'll act on.
5. Coding and spreadsheet help for beginners
You do not need to be technical to get value here. For a total beginner, Claude can explain what a confusing spreadsheet formula does, help you write a simple one, or walk you through a fiddly computer task one step at a time.
In Excel, I have amounts in column C and a status word ('paid' or 'unpaid') in column D. I want to add up only the amounts where column D says 'paid'. Give me the formula, explain each part in plain English, and tell me what to change if my columns are different.Why this works: Describing the columns and the goal in everyday words lets Claude produce the formula and explain it, so you learn what it does rather than pasting something you can't check.
When Claude produces something substantial (a formula, a table, a checklist, a small piece of code) it often opens in a side panel called an artifact, which you can read and edit in place. We cover that properly later in the phase. What Claude can't reliably do is build a finished, error-free program you don't understand; it makes mistakes in code just as in prose. For small, checkable tasks, though, it's a patient guide.
Try it now
Common mistakes
- Treating it as a search engine. Claude predicts likely text; it doesn't look up guaranteed facts. Fluent is not the same as correct.
- Trusting a document summary without a spot-check. Ask it to quote the source lines and check the ones that matter.
- Asking a spreadsheet question and quoting the number blind. Claude is stronger on the story than the sum. Confirm any figure you'll actually use in the sheet itself.
- Expecting one perfect answer. Its real strength is the back-and-forth; the first reply is a draft to improve.
- Over-trusting it on high-stakes facts because the writing is so assured. Medical, legal, financial, or anything you'll act on: a confident, well-written answer can still be wrong. Verify elsewhere before you rely on it. The quality of the prose tells you nothing about the truth of the claim.
Keeping current
What Claude can take in and produce keeps expanding: longer documents, more file types, richer analysis. The five jobs here are durable, but the edges (how much it can read at once, what it can open) shift with each release. For the current picture, see the official Claude release notes. Accurate as of 13 July 2026.