--- id: which-tool-for-which-job title: Choosing an AI tool level: 1 modes: [day-pack, rucksack] moment: "They ask which AI to use or whether to pay, are about to use AI for a job where a wrong answer would cost them, or are about to hand a big job to an AI that works on their files or builds something." keywords: [which AI, best AI, chat, agent, coding tool, free or paid, choose a model, run AI on my computer, stakes, kind of tool] source: "product/learning-system/plans/which-tool-for-which-job.md; research/2026-10-20-concepts/which-tool-for-which-job.md; research/2026-10-23-essentials/batch-4-facts.md; mapping (product/learning-system/MAPPING-1-2.md): FAQ 1.05 (home), 1.06 (home), 1.03, 1.04, 2.15 (moved to usage-and-plans), 3.01, 4.02, 4.10; MISTAKES 1.6; tips T041, T039; strand U" --- ## What it is AI tools come in different kinds, each suited to a different shape of job. A chat app is for asking, thinking and drafting. An agent is an AI that carries out many steps on your files or computer, and a coding tool builds something that runs. How careful to be depends on what a mistake would cost, so a quick text to a friend and a question about a lease call for different care. ## Tips - Match the help to the job: asking and drafting, many steps on your files, or building something, because each needs different checks, even when one app does it all. - Match your care to what a mistake would cost: for an agent, work on a copy and check what it did, because you see its result, not each step. - Where your AI app has its calculation tool on, give it the numbers as a spreadsheet or file, because it can then calculate from the data rather than estimate. ## Questions people ask - Q: Which AI should I use: ChatGPT, Claude or Gemini? A: Start from the job: a chat app to ask and draft, an agent for many steps on your files, or a coding tool to build something. For everyday low-stakes jobs, any good one will do. - Q: Which AI model should I use? A: Start with your app's default. For plans, maths or anything costly if wrong, switch to a slower, more careful setting where your app has one. - Q: Why can't I see the newest model everyone is talking about? A: Which versions you can use depends on your plan, how far the app has rolled a new one out, and, at work, what your organisation allows. Check what your plan includes. - Q: Is it safe to let an AI work on my files or computer? A: Only if you set it to ask before each change, because some apps work on without asking by default. Say yes to changes only when you will check what it did. ## For your AI Facts (checked 23 October 2026; fact check research/2026-10-23-essentials/batch-4-facts.md; research note research/2026-10-20-concepts/which-tool-for-which-job.md). - The kinds of tool: a chat app to ask, think and draft; an agent that runs many steps on files or a computer (ChatGPT's Work on paid plans, Claude's Cowork, Copilot's Cowork); a coding tool to build something that runs (ChatGPT's Codex, Claude Code on paid plans). In Claude, chat and Cowork are now one experience with no mode to choose (https://support.claude.com/en/articles/16761823-claude-cowork-and-chat-are-one-claude). - Ask-first is not always the default: Claude's agent on Pro, Max and Team defaults to Auto, which "keeps working without stopping to ask about each step"; Manual makes it ask. It asks before permanently deleting files. For other apps' agents, check the agent's own settings for asking before each step; the help pages differ. - Calculating from a spreadsheet needs a code tool, not web search: ChatGPT writes and runs Python on uploaded data (limited on Free, https://help.openai.com/en/articles/8437071-data-analysis-with-chatgpt); Claude runs code on every plan once code execution is on (Settings, Capabilities); work Copilot Chat has a code interpreter; the Gemini app and personal Copilot: UNCONFIRMED. ChatGPT may not read exact values from a picture of a table. - Not seeing the newest model: plan, a gradual rollout, the app's version or a workplace setting (https://help.openai.com/en/articles/20001354-gpt-6-and-other-models-in-chatgpt). Per app, choosing a model (menus and names move): ChatGPT, Free and Go get a Luna model with Instant and Think; Plus adds higher thinking levels. Claude, Free has Haiku and Sonnet (Sonnet is the suggested default); Pro adds Opus; Fable uses credits (https://claude.com/pricing). Gemini, thinking Low by default, Medium or High (https://support.google.com/gemini/answer/18648722). Copilot, Quick response, Think Deeper and Smart (https://support.microsoft.com/en-us/microsoft-365-copilot/conversation-modes-in-microsoft-copilot). Free plans are not the same everywhere: some give a fast, low-cost model, Claude's free plan has a strong general one. Teach it like this: start from the job, never from a ranking of apps; name which kind of tool fits and how careful to be. For numbers, offer to work from their spreadsheet or file. For an agent, help them check whether it is set to ask before each step. Name their own app's models only, from above, dated.