Which AI product should I try first

By Friday the same Slack channel had Dana’s four thank-you screenshots, Priya’s two coding agents, Omar’s $4,250 draft credit, Leah’s cafeteria CSV, and a meetup flyer that spelled Analytics with a y. Five people. Five jobs. One sentence they all used: “we should standardize on a model.” They did not have a job problem. They had a door problem. A decision tree is a hallway map, not a trophy ranking.

This is Part 6 of Which AI product should I use? It folds Parts 1 to 5 into one breath. Defaults below are try-this-first, dated, and allowed to be wrong for your house. Re-check vendors the week you write the team note. Brand-deep tutorials stay in Claude, ChatGPT, Gemini, and Grok. If you only remember one row, remember privacy wins ties. The thank-you can wait. The patient CSV cannot go back in the bottle. Write the 10-line note before you buy another seat. Then use the door you wrote down, even when a new demo lands in the group chat. Demos are optional. Unread thank-yous and unreviewed diffs are not. If the note still has blank brackets on Friday, you do not have a stack. You have a thread. Fill the brackets, send the note, then close the extra tabs. One door per job is enough for a month.

What you’ll learn

• A one-breath chooser you can run before you open an app
• Try-this-first defaults for writing, coding, office files, privacy, and pixels
• When to keep one tool for 30 days instead of A/B testing your life
• A Friday checklist for a team that already has too many logos
• Where this chooser stops and the open-models series starts

One-breath chooser

• Writing or a question: chat box. Part 1.
• Code in a real folder: one coding agent. Part 2.
• Mail, Docs, Sheets, local folders: the door next to the files. Part 3.
• Must not leave the machine, or you would not email the vendor: do not paste; local or a work contract. Part 4.
• Picture, voice, live claim: different surface, inspect pixels and links. Part 5.

If two bullets apply, do the privacy one first. Leah’s CSV does not become safer because Gemini is good at Docs.

Try this first

Rule of thumb: One default per job family for 30 days. Four logos on a Friday is not a strategy.

A 10-line team note

Paste this, then fill the brackets. That is the whole “standardize” meeting.

AI doors (review quarterly; names move)
Writing: [ChatGPT | Claude] in the work account. One version. We Send.
Coding: [Claude Code] in a clone. One agent per branch. We read diffs.
Office: [Gemini in Workspace | Outlook add-in | Cowork]. No auto-send. Money from the Sheet.
Private: no patient / payroll / HR files in personal chats. Toy data or local.
Images: no on-image type from the model. Overlay. No fake faces.
If two doors apply, privacy wins.
Owners: [Dana writing, Priya coding, Omar office].

What that note does: it names doors and owners. It does not rank GPT-5.5 against Opus 4.8. Leaderboards rot. Jobs do not.

Worked example: one Friday, five jobs

None of those people needed a new subscription. They needed the hallway. If your house already paid for one family, use it until it fails twice on the same kind of task. Then switch that job, not your identity.

Where this series stops

Setup (accounts, Free vs paid, 2FA) is the next Phase P series, not this chooser. Prompting habits are after that. Agents as a concept get their own path. Open models, licenses, and “what open means” start at OS1. Do not wait for that page to pick a chat box for Dana’s email. Do wait for it before you tell legal “we run Llama, so we’re open source now.”

What “fails twice” means

Switching tools every afternoon is how Dana got four screenshots. A default earns 30 days. “Fails twice” means the same kind of task, not one ugly sentence. If Claude’s drafts are stiff after you named the audience twice, try ChatGPT for writing. If Claude Code keeps touching seven files for a one-line check after you said “one file,” maybe Codex or Grok Build fits your repo better, or maybe your brief is still “just let it run.” Change one variable.

House overrides beat internet defaults. If IT already paid for Copilot and locked the tenant, Omar does not need Grok in Outlook this month. If the design team already lives in a specific image tool, the meetup flyer should go there, with the same inspect-the-letters rule. This chooser is for people who do not have a house yet, and for houses that collected logos by accident.

How this sits next to Practical AI

Practical AI teaches judgment for analysts: tokens, evals, RAG in plain English, what not to paste. This series picks a door so you are not running Practical AI inside four products at once. You still need both. A perfect Claude Code setup will still invent a metric if you ask it to “make revenue look healthy.” Grain, definitions, and evals do not retire because the logo is pretty. Link out. Do not redo those lessons here.

If you are choosing among models inside one product (Haiku vs Sonnet vs Opus, Grok 4.5 vs 4.6), that is a brand-series job, not this hallway. Pick the door first. Then pick the SKU. Mixing those questions is how Friday becomes a benchmark argument and the thank-you still does not send.

Common mistakes

• Standardizing on a model name instead of a door.
• A/B testing four apps on one thank-you.
• Letting privacy lose to convenience because Gemini was already open.
• Installing a coding agent to win an email.
• Never writing the 10-line note, then arguing in Slack forever.

A 30-day calendar that is not a content plan

Week 1: writing chat only. Every person who thinks they need AI sends one real draft through one box. Week 2: anyone who actually edits a repo may add one agent, in a clone. Week 3: office add-ins for people whose IT already allowed them, summaries only. Week 4: image stills with overlay, no video binge. Privacy is every week. If someone wants Llama on a laptop, they read OS1 before they buy RAM, and they do not do it with patient files as the first prompt.

That month will feel slow if your feed is full of agents. It will feel fast if you remember Dana’s 47 minutes. The cadence on this site for publishing is one post per series per day; your personal cadence for tools can be even slower. Empty lanes stay empty. Do not pad Friday with a second ChatGPT just because Gemini had nothing to do.

When a vendor rebrands (Bard to Gemini, a CLI to a new animal name, Cowork on web), update the 10-line note. Do not restart the 30 days unless the door actually changed. A new model SKU inside the same chat box is not a new door. Haiku vs Sonnet vs Opus is a later click, inside the Claude series, after you already opened Claude. The same is true of Grok 4.5 vs 4.6, and of whatever Gemini is calling the coding CLI the week you read this. Door first. Nickname second. If you cannot say the job in one clause, you are not ready to pick a SKU, and you are definitely not ready to buy four of them.

How to practice this week

Fill the 10-line note. Pick one default per job you actually have. Ignore the jobs you do not have. Then go use the door. Learn cards: Learn.

Quick recap

• Pick a door by the job. Privacy wins ties.
• One default per family for 30 days.
• Write the 10-line note. Stop collecting logos.

Sources

• Parts 1 to 5 of this series (job maps)
• ChatGPT, Claude, Gemini, Grok (confirm live surfaces)
• AMS Learn
• OS1: What open means

More definitions in Key Terms →

Merch on Etsy

Open store →

Mute Mouse Pad, Data Science Mouse Pad, Analytics Mouse Pad, You're On Mute 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 →