Week 5: Claude for Data & Repeatable Workflows

1. Analyzing Data & Generating Reports

Claude can take a spreadsheet or CSV directly and produce computed summaries — but language models predicting plausible text (Week 1) are not calculators, so for anything numeric, ask Claude to show its work rather than just state a final number.

"Using the attached sales data, calculate:
1. Total revenue by region
2. Month-over-month growth rate for each region
3. Which region had the most volatile month-to-month performance

Show the calculation for each number, not just the final figures, so I
can spot-check them."

Where your interface supports it, prefer letting Claude write and run actual code against the data rather than reason about numbers in prose — exact computation instead of estimation. Either way, treat any single important figure the way you'd treat a number from a junior analyst's first draft: worth a quick manual spot-check before it goes in front of anyone else.

2. Building Reusable Prompt Templates

If you've refined a prompt through a few rounds of trial and error and it now reliably does what you want, save it — don't reconstruct it from memory next time. A template with clearly marked placeholders turns a one-off good prompt into a standing tool.

WEEKLY STATUS SUMMARY TEMPLATE:

"Summarize these raw notes into a status update for {{AUDIENCE}}.

Format:
- One-sentence headline: overall status (on track / at risk / blocked)
- 3 bullet points: what shipped this week
- 1-2 bullets: what's blocked, with who's needed to unblock it
- Next week's top priority, one sentence

Raw notes:
{{NOTES}}

Tone: factual, no filler phrases like 'great progress' or 'exciting
update.'"

Store templates somewhere you'll actually reuse them — a Project's custom instructions (Week 4), a personal notes doc, or a snippet manager. When a template stops producing good results because the underlying task changed, revise it and note what changed and why, the same way you'd version any other piece of reusable tooling.

3. Where It Helps (and Where It Doesn't)

The biggest efficiency mistake isn't under-using Claude, it's applying it uniformly without noticing which parts of a task it's actually good at.

CLAUDE TENDS TO HELP A LOT WITH:
- First drafts of anything (emails, code, docs) — starting from zero is
  the most expensive part of many tasks
- Boilerplate & repetitive transformations (reformatting, translating
  between formats, generating variations)
- Summarizing large volumes of text you'd otherwise skim
- Brainstorming breadth — generating options for YOU to judge

CLAUDE HELPS LESS, OR NEEDS EXTRA CARE, WITH:
- Facts that must be current/verified right now (pair with a real source)
- High-stakes numeric or factual claims (verify before trusting)
- Decisions that hinge on judgment only you have the context to make
- Deeply novel creative direction — Claude extends a direction well,
  it doesn't reliably originate the ONE direction that's right for you

The practical version of this: use Claude to take the expensive first 80% off your plate — the blank page, the boilerplate, the first pass — and spend your own judgment on the last 20%, the decisions and verification that actually needed a human. Applying that same 80/20 split unevenly (leaning on Claude for the 20% that needed you, or doing the 80% by hand out of habit) is where the time savings leak away.

4. A Practical Efficiency Checklist

A short list to run through before a task, pulling together Weeks 1–5 into habits rather than a pile of separate techniques:

[ ] Is this recurring? → Build/reuse a template, don't freestyle it again
[ ] Does this need standing context? → Use a Project instead of restating it
[ ] Is this multi-part or long? → Structure it with XML tags (Week 2)
[ ] Does correctness really matter here? → Ground it in real sources,
    require citations, and plan to verify the output yourself
[ ] Am I about to say "make it better"? → Stop, give targeted feedback
    on specific parts instead
[ ] Is this actually a Claude-shaped task at all? → Check Section 3

None of these are new ideas at this point — the checklist's only job is making sure the habits from the last four weeks are the ones you actually reach for under time pressure, when it's tempting to just type a quick vague ask and hope.

5. Hands-on Exercise

6. Knowledge Check

Four quick questions. Expand each to check your answer.