# Suggested commands

Suggestions are opening phrasings, offered from the screen you are on and the state of your workspace. They are a way past the blank box, not a menu of everything the Assistant can do and not a plan it intends to carry out. Choosing one puts words in the box; anything that would change a record still arrives as a proposal you confirm.

- **Status:** Available
- **Audience:** both
- **In the app:** #/home, #/needs-you, #/inbox
- **Last verified:** 2026-09-10
- **Canonical:** https://connectbyjbrh.com/docs/assistant/suggested-commands/

## What they are drawn from

Two things. The first is where you are: the route and, if the page is showing one subject, that subject — the same verified screen context described in [Screen awareness](/docs/assistant/screen-context/). The second is the state of your workspace, which is what makes a suggestion on a quiet morning different from one on a day with a full Needs You queue.

That is the honest limit of what this page can say. Connect's documentation does not record the exact ranking that decides which suggestions appear, so none is described here. What matters operationally is the behaviour, and the behaviour is unambiguous: a suggestion is text, and text is not an action.

## What they are not

- **Not a capability list.** There are 66 tools behind the Assistant and only a few prompts on screen. Never conclude something is impossible because it was not suggested — ask for it.
- **Not a queue.** Suggestions do not accumulate, expire or need clearing. Ignoring every one of them costs nothing.
- **Not a decision.** Selecting one does not approve anything. If the request would write, you still get a card with editable arguments and a confirmation.
- **Not advice about your business.** A suggestion is a phrasing shortcut. The recommendation that carries reasoning behind it is the next-best-action answer, which cites what it was built from.

> **Note** The distinction between a prompt and a proposal is the one to keep. A prompt is what you might ask. A proposal is what Connect intends to do and is waiting for you to allow — see [Proposed actions](/docs/assistant/action-cards/).

## Using them well

1. Read the suggestion as a starting point and edit it before sending.
   - Result: You get the specificity of your own question with none of the typing — 'summarise this thread' becomes 'summarise what they have asked us for and what we have promised'.
2. Pin the subject first if the suggestion is about one account.
   - Result: The suggestion resolves against the pin rather than the page behind the panel, which is the usual source of a right answer about the wrong account.
3. Ignore them entirely when you know what you want.
   - Result: Typing the request directly is always available and is never worse.

A suggestion is most useful in exactly the situation people describe as being stuck: you can see that something needs doing, and you are not sure what to ask for. It is least useful once you have a habit, because your own phrasing is better than any generic one.

## When a suggestion looks wrong

**It is about an account you have finished with** — The tab is still pinned to it. Unpin, or open a fresh tab.
**It offers something you do not have** — Suggestions follow the screen and the workspace state, not your permissions in every detail. If the underlying action is beyond the Assistant's authority, the refusal comes with the reason — see [What the Assistant may not do](/docs/assistant/assistant-authority/).
**There are no suggestions at all** — That is not a fault. The box works exactly the same when empty; type the question.
**The same one keeps appearing** — Usually because the thing behind it is still true — an item still in Needs You, a follow-up still late. Deal with the cause rather than the prompt.

## Questions

### Does clicking a suggestion run it immediately?

It puts the request in front of the Assistant. A question is answered; anything that would change a workspace record comes back as a card you confirm, exactly as if you had typed it yourself. There is no path from a suggestion straight to a write.

### Can I add my own suggestions?

Connect's documentation does not record a way to author them, so this page does not claim one. The durable way to shape what the Assistant does by default is a standing instruction, which is a direction rather than a prompt.

### Are suggestions different for a customer workspace?

The mechanism is the same one for both audiences. What differs is workspace state — a queue with items in it produces different prompts from an empty one — and, for a customer, the plan's allowance bounding how much work follows.

## Related

- [Connect Assistant](https://connectbyjbrh.com/docs/assistant/)
- [Screen awareness](https://connectbyjbrh.com/docs/assistant/screen-context/)
- [Proposed actions](https://connectbyjbrh.com/docs/assistant/action-cards/)
- [Answers that change nothing](https://connectbyjbrh.com/docs/assistant/read-only-answers/)
- [What Connect may do](https://connectbyjbrh.com/docs/autonomy/)
- [Making a next-best-action defensible](https://connectbyjbrh.com/research/evidence-for-next-action/)

## What this page is based on

- `docs-source/sources/GENERAL.md` §8 — the Assistant panel and its 66 tools
- `docs-source/sources/GENERAL.md` §5 — Needs You as the ranked queue
- Connect capability registry (`docs-source/facts.py`) — `assistant`, `needs_you`