Chat and insights

Asking it something

Open the assistant and ask in plain language. Replies stream as they are written rather than appearing all at once, so you can start reading immediately.

It has context on your tasks, notes and events, which is what makes it different from a general chatbot: “what’s left before Friday” is a question it can actually answer.

Every reply carries a reminder that AI-generated content may be incorrect and should be checked. Treat it as a fast reader of your own data rather than an authority on it.

Suggested questions

A new chat opens with a few suggestions, and they are built from your actual data rather than being a fixed list. You will see the number of overdue tasks you really have, or the name of an event with replies outstanding, phrased as a question you can tap.

They are worth reading even when you came in with your own question, because they surface things you were not looking at. A suggestion about guests who have not replied is often the first you hear of it.

It can change things, not just answer

Marvelist AI can act on your lists as well as read them, so “move everything unfinished to tomorrow” is an instruction rather than a request for advice.

Anything it changes is recorded in change history attributed to Marvelist AI, so a change you do not recognise can be traced rather than guessed at.

Previous conversations

Conversations are kept, listed down the side, so you can go back to one rather than re-explaining. Each can be renamed, which is worth doing for one you will return to, and deleted when you are done with it.

New chat starts a fresh one. Do that when you change subject: a long conversation carries its earlier context into every later answer, which is helpful until it is not.

Insights

Alongside chat, Marvelist AI surfaces insights about your tasks: patterns, things slipping, what is worth attention. They are generated from your own data rather than being generic productivity advice.

Usage

There is a monthly allowance, with a meter on the assistant screen showing how much you have used, so you are not guessing.

It resets at the start of each calendar month. If you reach the limit, the meter says so and chat stops until the reset. Nothing else about Marvelist is affected, and nothing is lost: your conversations, memories and lists are all still there.

The allowance is measured by how much text goes back and forth, not by the number of messages. A long thread costs more than a short one, and a question asked in a fresh chat costs less than the same question at the end of a long conversation, because the earlier messages are sent along as context. Starting a New chat when you change subject is the easy way to make it last.

Memory

The assistant remembers durable facts and preferences about you from your chats, and uses them to personalise later replies.

Settings → Marvelist AI → AI Memory is where you see everything it has remembered. You can turn the whole feature on or off, switch individual memories on or off, edit them, delete them, or clear the lot and start again.

Each memory records which conversation it came from, so one you do not recognise can be traced back to what you said rather than being taken on trust. Switching a memory off is the gentler option: it stops being used without being destroyed, which is what you want when you are not sure whether it is the one skewing an answer.

Worth opening occasionally. Memory is built from what you happened to mention, so it accumulates the odd thing that was true once and is not any more, and an out-of-date memory quietly skews answers until you notice it.

Anything on a Vault-protected list is never read by the assistant and so never ends up here.

On iOS

There is an Ask Marvelist AI Siri shortcut, so you can ask without opening the app. See Siri and Shortcuts.

Getting good answers

Be specific about scope. “What’s due this week on the launch list” beats “what should I do”, because the second makes it guess what you meant by important.