First week of useful tasks (write, plan, learn)

Monday morning, 9:14. Your calendar has a stakeholder update at 11, a half-built project plan you promised last week, and a 40-page PDF someone called “light reading.” You open ChatGPT, type “help me with work,” and get a cheerful essay about productivity. You close the tab. Nothing useful left the building.

This is Part 1 of the ChatGPT everyday tutorial. The series assumes you already know what ChatGPT is, which plan you are on in broad strokes, and how to stay in ordinary Chat without wandering into Work, Codex, or the API. If those foundations are still fuzzy, start with Learn ChatGPT from scratch, then skim ChatGPT product map so Chat, Work, and Codex do not blur. Here we stay in Chat on purpose: five weekdays, three job buckets (write, plan, learn), one real output per day.

The grain is simple. By Friday you should have five artifacts you can actually use: a draft you can ship after editing, a plan you can run, a plain-language teach-back, a tighter rewrite of something you already wrote, and a short review note that decides what worked. Ten clever toy prompts do less for you than five finished objects.

What you’ll learn

• How to run a Mon-Fri Chat week with one real deliverable each day
• Three job buckets that cover most everyday Chat work: write, plan, learn
• Worked prompts you can paste, with a sample output shape for each day
• Paste and privacy habits that keep the week useful instead of scary
• When to stop generating and start editing so Chat does not own your voice
• What to skip this week (Work, Codex, Custom GPTs, API keys) so you build muscle, not a product museum

Stay in Chat for this series

OpenAI’s product surface keeps growing. Desktop can show Chat, Work, and Codex side by side. That is useful later. It is noise this week. Everyday tutorial means conversation: you type, the model answers, you edit, you ship. Work-style multi-step agents and Codex-style coding agents have their own tutorial tracks after this series. If a button looks exciting and agentic, leave it alone until Friday’s review says you are bored of plain Chat (you will not be bored after five real outputs).

Use chatgpt.com on a laptop if you can. Full keyboard, easy copy into email or Docs, room to rename threads. Phone is fine for a rewrite on the train after the draft exists. Same account, same history habits from Learn ChatGPT. Rename each day’s chat with the deliverable name so next week’s you can find “Q3 update draft” instead of “New chat.”

Three job buckets

Most useful Chat sessions fall into one of three buckets. Name the bucket before you type. Naming stops the “help me with work” spiral.

You will mix buckets later (plan a study week, then write a summary). For days 1-5, pick one primary bucket each morning so the success test stays clear: did a usable artifact exist when you closed the tab?

The week map

Here is the schedule this post walks through. Swap topics to match your job, keep the shape.

If a day slips, do not stack two full sessions “to catch up.” Skip to the next day’s shape or shrink the deliverable. The habit is finishing, not filling a bingo card.

Rules for the whole week

Paste less than you think

Use redacted samples, fake customer names, and synthetic numbers when you are practicing. If you would not put it on a shared slide, do not paste it into a personal Free or Plus chat without checking your company policy. Business and Enterprise workspaces have different controls; still treat chat as a place other people or processes might see content later. Learn ChatGPT Part 8 covered judgment; this week is practice, not a privacy seminar. The short version: no payroll files, no full CRM dumps, no unreleased strategy decks.

Facts you did not provide are suspects

ChatGPT will invent plausible metrics, quote sources it never read, and fill gaps so the prose looks complete. Your job is to supply the facts (or mark them as unknown) and delete anything you cannot defend. If Monday’s draft invents “revenue up 12%,” and you never said that, the draft is wrong even if it sounds executive.

Edit out loud

Read the final version as if your name is on it. Change idioms that are not yours. Swap corporate mush for the words your team actually uses. Chat is a first draft machine, not a spokesperson. The week fails if you ship unedited AI tone into Slack and then spend Friday explaining yourself.

One chat per deliverable

Do not run Monday’s update, Tuesday’s plan, and Wednesday’s study thread in one endless conversation. Context gets crowded, and you will lose the artifact. New chat, rename it, archive when done. Projects (if you use them from Learn ChatGPT) can hold related files later; this week you can keep it simple with five named threads.

Monday: write a real update

Pick something you must send this week. A project update, a status note to a VP, a client progress email. Gather 4-8 true bullets only you know: what shipped, what blocked, what you need, dates that are real. No vibes. No “we’re crushing it.”

Worked prompt (copy and fill)

You are helping me draft a work update I will edit and send under my name.

Audience: non-finance managers on the ops leadership channel.
Length: 350-450 words.
Tone: plain, calm, specific. No hype words. No “excited to share.”

Use ONLY these facts. If something is missing, write [UNKNOWN] instead of inventing:
- Shipped: weekly dashboard v2 live for East region on Mon 7/14
- Open issue: West region still missing store #118 in the feed (owner: data eng, ETA Wed)
- Decision needed: freeze new KPI requests until Fri so we can fix grain on “active store”
- Next milestone: pilot South region, target date next Tue if West feed is clean
- Risk: if #118 is not fixed by Wed, South pilot slips one week

Structure:
1) What changed this week (5-7 sentences)
2) Blockers and owners (bullets)
3) Decision I need from readers (one clear ask)
4) Next seven days (bullets)

End with two short questions I can paste into the channel to start discussion.
Do not invent numbers, quotes, or customer names.

What a good Monday output looks like

You should leave with a draft in your notes app that still has your facts, no surprise percentages, and a clear ask. If the model added “NPS jumped to 72,” delete it. If the tone sounds like a press release, run one follow-up: “Rewrite at 12th-grade reading level. Shorter sentences. Keep structure.”

Ship path: edit 10 minutes by hand, paste into Slack or email, keep the chat renamed 2026-07-mon-ops-update (or your date). That rename is part of the deliverable.

Tuesday: plan a week you can actually run

Plans fail when they list twenty tasks and zero time. Tuesday’s job is a one-week plan with time boxes, a refuse list, and one success metric that is observable (not “be more strategic”).

Worked prompt

Turn my messy list into a one-week plan I can follow.

Constraints:
- I have about 18 focused hours this week for deep work (not meetings).
- Standing meetings already eat Mon/Wed 10-11 and Thu 2-3.
- I will not work past 6pm on weekdays this week.
- Prefer finishing three meaningful outcomes over starting eight.

My messy list (unordered):
- Finish South region pilot brief
- Reply to 12 open tickets in the metrics channel
- Prep 20-min teach-back on active store grain
- Clean personal notes from last retro
- Book dentist
- Draft QBR outline for August
- Review PR on dashboard filters
- Write hiring scorecard notes for analyst role

Output format:
1) Three outcomes for the week (done means what?)
2) Day-by-day schedule in 60-90 min blocks inside my available hours
3) Parking lot (explicitly deferred)
4) Refuse list (things I will say no to if they appear midweek)
5) Friday 15-min review checklist

If the list does not fit 18 hours, cut items and say what you cut and why.
Do not invent meetings or owners I did not name.

Sample shape of a strong plan

Notice dentist and ticket pile may land in parking lot or refuse list. Chat cannot know your real life; you still decide. The value is sequencing and honest cuts, not a perfect Gantt chart.

Wednesday: learn one hard thing simply

Pick one topic that shows up at work: contribution margin, cohort retention, primary keys, “active customer,” prompt injection, whatever you keep nodding about in meetings. Goal is not a PhD. Goal is a teach-back you could give a smart new hire in ten minutes.

Worked prompt

Teach me like a smart colleague, not a textbook.

Topic: what “active store” should mean for a weekly retail ops dashboard.
Audience for my teach-back: new ops analysts, first month on the job.
I already know: we have a store dimension table and a daily sales fact.
I am confused about: whether temporary closures count, and whether online-only locations count.

Please:
1) One-sentence definition options (give me 2-3 grains, not one true religion)
2) Everyday analogy that is not cute for its own sake
3) Tiny worked example with 5 fake stores and which would count under each definition
4) Common mistake teams make
5) How I would check understanding (5 short quiz questions with answers in a separate section labeled ANSWERS)

Do not invent company policies. Mark unknowns as decisions our team must make.
No footnotes to papers I did not provide.

How you finish Wednesday

Answer the quiz yourself without peeking. Fix wrong answers by asking Chat to re-explain only that slice. Then write a five-sentence teach-back in your own words in a note. If you cannot write those five sentences without looking, the session was reading, not learning. Part 4 of this series goes deeper on studying hard topics; Wednesday is the lightweight version that still produces an artifact.

Thursday: rewrite for length and voice

Take Monday’s update (or a long email you already sent that felt like molasses). Goal: half the length, same facts, more you.

Worked prompt

Rewrite the draft below.

Goals:
- About 50% of the original word count
- Keep every concrete fact, date, owner, and ask
- Delete filler and repeated ideas
- Match a direct workplace voice (I will edit further)
- Keep the four-part structure if it still fits; otherwise use: status, blocker, ask, next

Refuse:
- New numbers
- Soften language that hides a real risk
- “Hope this helps” closings

Draft:
[PASTE YOUR MONDAY DRAFT OR REAL EMAIL HERE]

Compare side by side. If the short version lost the decision ask, you cut too hard. If it still has “synergize across stakeholders,” that word never came from your bullets: delete it. Thursday trains the muscle Part 2 will sharpen: structure and examples make rewrites better than “make it better.”

Friday: review without ceremony

Friday is not a productivity TED talk to yourself. It is a half-page note so next week starts from evidence.

Worked prompt

Help me write a Friday review note from my bullet log.

My log:
- Mon: sent ops update after 12 min of hand edits; Chat invented one metric I removed
- Tue: plan was too dense; I only finished 2 of 3 outcomes
- Wed: teach-back on active store; still fuzzy on temp closures
- Thu: rewrite cut well; tone still slightly generic
- Fri: (today)

Produce:
1) What shipped (artifacts)
2) Where Chat helped vs where I had to fight it
3) One paste or privacy moment I want to be stricter about
4) One habit for next week (single sentence)
5) One topic to deepen (point me at practice, not hype)

Keep the whole note under 250 words. No pep talk.

Save the note where you keep work retros. If nothing shipped Mon-Thu, write that honestly. The review is still the deliverable. You cannot improve a week you refuse to describe.

A full mini example: one person’s week

Here is a compact story you can mirror. Names and numbers are teaching toys, not a case study of a real company.

Alex is an ops analyst. Monday’s bullets: East dashboard live, store #118 missing, freeze new KPI asks until Friday. Chat drafts 420 words; Alex kills a fake “12% lift” line and ships 380 words to the ops channel with two questions. Tuesday’s plan puts the South pilot brief and PR review as outcomes; ticket pile goes to a 45-minute batch Friday morning, not every day. Wednesday Alex studies “active store,” picks grain option B (open for sales at least one day in the week), and fails two quiz questions about pop-up stores. Thursday rewrite cuts Monday’s update to 190 words for a VP who hates walls of text. Friday note: “Chat is fast at structure; I still own numbers. Next week: put refuse list in every plan prompt.”

Five artifacts. Zero Custom GPTs. Zero API keys. Zero Work mode. That is a successful first week.

Common mistakes in week one

What not to do this week

• Do not build a Custom GPT “for productivity.” You do not know your repeating tasks yet.
• Do not open Platform API keys because a thread said serious people have keys. Map Part 4 covered that split.
• Do not run Work or Codex jobs for these five deliverables. You need Chat judgment first.
• Do not paste the whole company handbook “for context.” Paste the page you need, redacted if needed.
• Do not chase model names in the picker every hour. Use a solid default and focus on the prompt and the edit.

How this fits the AMS path

Foundation series: Learn ChatGPT from scratch (what it is, plans, first minutes, modes overview, memory/Projects, multimodal, connectors, privacy judgment). Orientation series: ChatGPT product map (Chat vs Work vs Codex, Custom GPTs, models, API vs app). This everyday tutorial is the default next track for most readers: useful Chat habits before agentic blast radius. Later: Work tutorial, Codex tutorial, Custom GPTs tutorial. Sibling stack on Anthropic starts at Learn Claude from scratch. All paths: Learn.

Practice for next week (and Part 2)

Part 2 is Better answers with structure and examples: role, goal, constraints, example, format; weak vs strong prompts; how to iterate when the reply is wrong without starting over from vibes. Bring one weak prompt from this week and we will rebuild it.

1. Run Mon-Fri as written (or compressed into three days if your calendar is brutal: write, plan, learn).
2. Save five renamed chats and five artifacts outside Chat (docs, notes, sent email).
3. Keep a “weak prompt graveyard” note with two lines that failed. You will need them in Part 2.
4. Re-check current plan and data controls on OpenAI’s Help Center if your company is writing policy this month. Product labels move.

Quick recap

• Stay in Chat. One real output per day beats ten toy prompts.
• Buckets: write, plan, learn. Name the bucket before you type.
• Mon write, Tue plan, Wed learn, Thu rewrite, Fri review.
• Supply facts or mark [UNKNOWN]. Delete invented numbers.
• Edit under your name. Rename every thread.
• Foundations: Learn ChatGPT and the product map. Next part: structured prompts.

Sources

Product surfaces and help articles change. Re-check official pages when you set team habits or policy.

• ChatGPT app (everyday Chat surface for this series)
• OpenAI: ChatGPT product home (product overview and plan entry points)
• OpenAI Help Center (plans, data controls, and feature articles; search current titles)
• OpenAI Help: Data Controls FAQ (consumer-oriented data control orientation)
• OpenAI: Usage policies (what you may not use the services for)
• Analytics Made Simple: Learn ChatGPT from scratch (foundation series before this tutorial)
• Analytics Made Simple: ChatGPT product map (Chat vs Work vs Codex and related orientation)
• Analytics Made Simple: Learn (broader learning paths)

More definitions in Key Terms →

Merch on Etsy

Open store →

You're On Mute White Ceramic Coffee Mug, Data Science Mug, Data Mug, Analytics Mug, Statistics Mug, Programming Mug, Gift Mug, Coffee Mug View on Etsy → Imbalanced Classes Mouse Pad, Data Science Mouse Pad, Analytics Mouse Pad, Imbalanced Mouse Pad View on Etsy → You're On Mute White Ceramic Coffee Mug, Data Science Mug, Data Mug, Analytics Mug, Statistics Mug, Programming Mug, Gift Mug, Coffee Mug View on Etsy → Good Cheap Fast White Ceramic Coffee Mug, Data Science Mug, Data Mug, Analytics Mug, Statistics Mug, Programming Mug, Gift Mug, Coffee Mug View on Etsy →