“Best AI assistant for work” and “best AI assistant for personal use” are the two most searched versions of the question, and they have different answers. Not because the models differ, but because the job does.
At work the assistant has to act inside accounts: the inbox, the repo, the CRM, the ad manager, the CMS. What matters is reach and permissions. At home it has to be there when you think of it: on the phone, by voice, in the chat app you already have open. What matters is presence and price.
Both lists below. Prices checked 2026-09-21.
Disclosure: TODO for AI is ours. It appears on one of the two lists and not at the top of the other, which is the point of the post.
For work
The test: hand it “reply to the three unanswered support emails, then update the CRM, then fix the failing test and open the PR” and count how much of that it finishes without you.
1. TODO for AI
A to-do list where each item is done by an agent with a shell on your PC, your logged-in Chrome via an extension, and a cloud VM for anything that should keep running. Any model on one plan. Allow, ask first or block per tool. Finishes all three items if the tools are connected. Weak spots: setup is not zero, browser items wait for Chrome, free tier is Sonnet and Haiku only. From free; every model at $20. How it does the tasks.
2. Claude, with Cowork
The best document producer on either list. Cowork takes a folder and an outcome and returns spreadsheets with formulas, decks, reports, organised files, with scheduled runs in Anthropic’s cloud. Claude in Chrome reaches logged-in sites on Max and Team. Nothing runs on your shell; one vendor. From $20. Comparison.
3. Claude Code or Codex
If the work is the repo, these are the mature choice. Fix, test, PR, review, with cloud sessions and scheduled routines. Both are repo-first: Claude Code reaches inbox and CRM work through MCP connectors and Claude in Chrome while you are there, Codex stays in the repo, and the unattended cloud runs of both are repo-shaped. From $20, or $8 with ChatGPT Go for Codex. Claude Code comparison, Codex comparison.
4. Lindy
For repeatable inbox-and-CRM workflows an ops team designs once. Credit priced from $30. Strong where the process is fixed, wrong where it varies. Comparison.
5. ChatGPT or Gemini
Drafting, research, summaries, and Gemini for anything inside Google Workspace. Both are answerers with agent features arriving. Excellent second tool, weak first tool for work that has to be shipped. Free; $20.
For personal use
The test: it is 22:40, you are on your phone, and you want to say “remind me to cancel the trial before Thursday and find a cheaper flight for the 14th” and have it happen.
1. ChatGPT
The best phone app, voice mode, memory of your preferences, and everyone you know can use it. Agent mode handles public-web tasks like the flight search. Free tier is generous; Plus $20. Weak spot: it works in OpenAI’s browser, not yours, so “cancel the trial” means logging in there or wiring a connector first.
2. Claude
Same shape with a better writer and Cowork for the occasional bigger personal job: tax folder, trip plan, the spreadsheet for the renovation. From $20.
3. Gemini
If your life is in Gmail, Calendar and Drive, Gemini is already inside them and free. The assistant on Android. Weak outside Google.
4. Hermes Agent
For the technical person who wants an assistant in Telegram or WhatsApp that runs on their own box, remembers them, and writes its own skills. Free and MIT; you host it. Not for anyone who does not want to run a server. Comparison.
5. TODO for AI
Fifth, honestly. It has voice, a mobile app in progress, and it will do the trial cancellation in your real browser, but it is built for a list of work and the setup is aimed at people with a machine to attach. If your list is personal chores, ChatGPT is the better first pick and this is the upgrade when the chores start needing your accounts.
Why the lists differ
| Work | Personal | |
|---|---|---|
| Deciding factor | Reach into your accounts and machine | Presence on your phone and voice |
| Best pick | TODO for AI | ChatGPT |
| Runner up | Claude Cowork | Claude |
| Common mistake | Buying a chat app and pasting outputs into tools by hand | Buying an agent platform for grocery reminders |
| When to switch | The list is right and still not getting done | The chores start needing logins |
One assistant does not win both, and the roundups that say otherwise are ranking on writing quality. Pick by the job.