Dashboard assistant

Overview

The dashboard assistant is the AI on Home. Ask in plain language — it searches tenders, opens pursuits, works with requirements and quotations, reads team settings, and runs multi-step bid workflows. Sensitive writes pause for your approval before executing.

The assistant respects team membership, subscription limits, and credit balance. Higher intelligence settings use more credits per reply.

Tip: For reliable results:

  • Confirm the right active team before asking about pursuits or settings

  • Keep web search on for market questions; off for strict internal-only tasks

  • Read approval cards carefully before approving deletes or settings changes

  • Use New chat when switching unrelated tasks — not to cancel a running action

Important: Teammates' chats in Recents open read-only. Start New chat for your own writable session.

Composer

The composer sits at the bottom of Home in both start screen and conversation mode.

Text area

Placeholder pattern:

"Search tenders, manage pursuits, edit requirements, quotations, documents, or team settings"

Supports multiple lines, drag-and-drop attachments, and multi-file selection.

Disabled when:

Condition

Why

Approval/input card visible

Must resolve checkpoint first

Prior approval submitting

Prevents duplicate submits

Read-only shared chat

Not your session

Action menu (+ button)

Opens upward:

Item

Behaviour

Add photos & files

Attach images or documents

Web search

Toggle online search (default on; checkmark when enabled)

Intelligence

Submenu for response quality (uses credits)

Intelligence levels

Level

When to use

Auto (default)

Everyday tasks; balances cost and quality

Fast

Quick lookups; may make more mistakes

Balanced

Middle ground

Maximum

Complex multi-step bid workflows

Selection persists for the session. See Billing for credits.

Speech and send

Control

Behaviour

Speech

Voice-to-text; disabled when composer disabled

Send (idle)

Sends message

Stop (square, streaming, empty input)

Cancels current run

Send (streaming, text entered)

Queues message instead of stopping

Save (editing queued item)

Updates queue entry

Queued messages

Sending while the assistant is still replying queues your message instead of interrupting.

UI element

Meaning

Queued messages header + count

Collapsible panel, default expanded

Send next (↑)

Jump to front; may stop current run first

Edit

Load into composer

Remove

Delete from queue

When the run finishes, queued messages send in order (FIFO) unless you stopped the run.

Starter and follow-up suggestions

Starter suggestions (empty chat)

Four default chips, e.g.:

  1. Search recent tenders for your industry

  2. Open latest pursuit and summarize missing requirements

  3. Create quotations for unmapped requirements

  4. Review team legal representative and banking

Click sends immediately with current intelligence and web search settings.

Follow-up suggestions (after replies)

Up to four context-aware chips may appear above the composer (e.g. Open the most relevant tender).

Hidden while assistant is working, when an approval card is shown, or on read-only chats.

Message thread

Your messages

Right-aligned bubbles with text and optional attachments.

Resolution bubbles appear when you approved or rejected a sensitive action: "You approved this action" / "You rejected this action" with request details.

Assistant messages

Left-aligned; may stream token-by-token.

Element

Meaning

Thinking... shimmer

Waiting for first content

Reasoning (collapsible)

Extended thinking with duration

Sources (collapsible)

External links with count

What I am doing

Action step panel

Checkpoint

"Automatically compacted context" on long threads

Open tender buttons

Tender links as buttons, not raw URLs

Attachments

Download deliverables from message

Hover actions: timestamp, Copy, Retry (not while streaming), version navigation when multiple replies exist. Scroll to bottom appears when you scroll up.

Tool steps — "What I am doing"

Collapsible panel with user-facing labels and statuses: In progress, Queued, Completed, or Could not complete this step. Auto-expands while running.

Approval and input cards

Cards appear above the composer; composer and suggestions disable until resolved.

Input required

Badge: Input required

Shows agent message, read-only context fields, and editable fields (text areas, inputs, checkboxes).

Continue submits answers. Cancel run aborts.

Decision required

Badge: Decision required

Radio (single choice) or checkboxes (multiple) with optional descriptions.

Submit decision / Cancel run

Confirmation required

Amber card for sensitive actions — e.g. delete requirement, create pursuit, update team settings.

Approve (primary) / Reject (outline)

Examples requiring confirmation:

Action type

Delete requirement

Create pursuit from tender

Import extracted requirements

Start quotation generation

Update company profile, legal rep, banking

Create or AI-fill pursuit documents

What the assistant can do

Domain

Examples

Notices

Search, open, summarize tenders

Pursuits

Create, open, summarize projects

Requirements

List, create, update

Quotations / financial offer

Match products, generate quotes

Team settings

Read/update company profile, banking, legal rep (with approval)

Statistics

Company and authority intelligence

Alerts

Configure or inspect alert rules

Web

External search when enabled

Scheduled tasks

Create and inspect

All actions respect permissions, plan limits, and approval gates.

Read-only sessions

Teammates' chats in Recents:

Capability

Writable session

Read-only session

View history

Yes

Yes

Send messages

Yes

No

Queue / stop

Yes

No

Approve actions

Yes

No

Rename / delete

Yes

No

Start New chat for your own session.

Session menu (Recents row)

⋯ menu or right-click:

Action

Behaviour

Last updated

Date/time

Rename

Inline edit; Enter saves, Escape cancels

Delete

Removes session; if active, switches to another or New chat

Disabled on read-only sessions.

Errors and streaming

Event

Feedback

Send failure

Toast Failed to send assistant message; text restored to composer

Stop failure

Toast Failed to stop assistant response

Approval failure

Toast Failed to continue assistant run

Delete/rename failure

Matching error toasts

History load failure

Inline Try again (not toast)

Streaming: replies fill token-by-token. Retry on assistant messages creates new versions with previous/next navigation.

Long threads may show a short summary message when the conversation gets very long — your saved pursuits, settings, and documents are unchanged.

FAQ

Can the assistant change team settings without asking?

Sensitive updates show an amber confirmation card. Always read the summary before Approve.

Why is send disabled?

Read-only chat, active approval card, or submission in progress.

Does stopping cancel pursuit changes already saved?

Stop cancels the current AI run. Approved actions already committed persist.

How do I fix a wrong answer?

Continue the same thread with a correction — do not use New chat for fixes.

Why did my message queue instead of sending?

The assistant was still replying. It will send when the current run completes, or use Stop first.

Which intelligence level should I use daily?

Auto for most work; Maximum for complex multi-document bid sessions when credits allow.

  • Dashboard (Home)

  • Assistant session

  • New chat

  • Search chats

  • Billing

  • Team settings

  • Working with AI assistant