# Needs You inside the Assistant

Needs You draws into the Assistant's single scroller alongside Chat, Standing instructions, Memory and Files, so you can review a held reply or an operational problem and decide on it without leaving the panel. It is the same queue and the same approval path as the full screen — ranked rather than chronological, and draining by itself as causes clear.

- **Status:** Available
- **Audience:** both
- **In the app:** #/needs-you, #/approvals, #/inbox
- **Last verified:** 2026-09-10
- **Canonical:** https://connectbyjbrh.com/docs/assistant/needs-you-in-assistant/

## What is in the queue

Needs You is not an inbox of messages. It is everything currently waiting on a human judgement, which is a mixture of kinds that would otherwise be spread across four screens.

- **Held replies** — finished drafts the channel's autonomy mode will not send without a person, and sends waiting on an allowance that has been spent.
- **Approvals** — one held action apiece, each waiting on a yes or a no.
- **Questions** raised for a person, including ones the Assistant put there when it could not settle something itself.
- **Operational problems** — voice line health escalates into the same queue, rendered for both audiences from the same place, so a line that cannot take calls is a decision rather than a graph nobody is watching.

The queue is ranked, not chronological. A held reply to a first-time enquiry is a different kind of debt from a held reply on a thread that has been running a week, and ordering by arrival time buries the first under the second.

## Deciding without leaving the panel

1. Open Needs You in the sidebar — or from the More menu if the window is narrower than 600px.
   - Result: It draws into the same scroller as the conversation, so you are not navigating away from the work that raised the question.
2. Read the item with what it concerns. A held reply is shown with the message it answers.
   - Result: You can judge the answer against the question without opening a second screen.
3. Edit if you want to, then approve or reject.
   - Result: Your edit is what goes; the same send boundary a person's own send uses carries it, and the state stays uncertain until the provider acknowledges it.
4. Ask the Assistant about the item in the same tab if you need context before deciding.
   - Result: It reads the relationship, the history and any commitments — an ordinary read that changes nothing while you make up your mind.

That last step is the reason for putting the queue here rather than only on its own screen. Most held decisions are held because something is unclear, and the fastest way to clear it is to ask about the account without losing sight of the item.

## What it does not change

**The authority** — Deciding in the panel is deciding as yourself. It does not let the Assistant approve on your behalf, and it does not reach anything the Assistant may not do.
**The record** — The approval is attributed to the person who made it and lands in the decision log with the rule that applied — identical to deciding on the full screen.
**The ranking** — Working an item from here does not promote it. The order is the queue's and is the same wherever you look at it.
**The draining** — Entries clear themselves as their causes clear. An operational item that resolves — a line that comes back — leaves without anyone dismissing it, which is why the queue is worth trusting as a signal.

> **Careful** An item still in the queue means the customer is still waiting. A held reply is no reply rather than a slow one: nothing has been handed to any provider, there is no delivery attempt, and the clock the customer is counting has not stopped.

## When the panel is the wrong place for it

Two situations are better on the full screen. The first is a backlog: working twenty items is a task with its own rhythm, and the wider view shows more of them at once. The second is anything you need to compare — two held replies to the same customer, or a queue you are trying to understand rather than clear.

The panel is at its best for the single item that has just interrupted you, and for the decision that needs one question answered first. Both screens are the same queue, so nothing is lost by moving between them mid-way.

## Questions

### Is this a copy of the Needs You screen?

It is the same queue over the same data, drawn into the panel. There is one implementation behind both, which is also why a decision made in either place behaves identically and is recorded identically.

### Can the Assistant clear the queue for me?

No. Each item is a decision for a person, which is the definition of being in this queue. What the Assistant can do is read around an item — the history, the commitments, what was promised before — so the decision takes seconds rather than minutes.

### Why did an item disappear without me touching it?

Entries drain as their cause clears. An operational problem that resolves itself, or a send that becomes possible when the allowance resets, leaves the queue without anybody dismissing it.

## Related

- [What Connect may do](https://connectbyjbrh.com/docs/autonomy/)
- [Proposed actions](https://connectbyjbrh.com/docs/assistant/action-cards/)
- [Running an action](https://connectbyjbrh.com/docs/assistant/executing-actions/)
- [Assistant tabs](https://connectbyjbrh.com/docs/assistant/tabs/)
- [Human approval without blocking low-risk work](https://connectbyjbrh.com/research/approval-without-blocking/)
- [Email in Connect](https://connectbyjbrh.com/docs/email/)
- [Phone and voice in Connect](https://connectbyjbrh.com/docs/phone/)

## What this page is based on

- `docs-source/sources/GENERAL.md` §5 — Needs You, ranking, line health escalation and self-draining
- `docs-source/sources/GENERAL.md` §8 — the single scroller, the sidebar at 600px and the More menu
- Connect capability registry (`docs-source/facts.py`) — `needs_you`, `approvals`, `held_draft_approval`, `line_health`
