AI Tool Results & Artifacts
When the assistant does something rather than only replying, the result shows up as one of two things: a structured card in the conversation, or a document that opens as an editable artifact.
How a tool result appears
Every tool the assistant runs — searching, reading, creating, editing or deleting a record, or reaching outside eFirm to the web or a knowledge base — renders through the same card: an icon, a title, and a live status badge, with the tool-specific content inside. Most cards start collapsed; search-style cards start open so results are visible right away.
| State | What you see |
|---|---|
| Running | A progress indicator, with an elapsed-seconds counter on anything long-running so it never looks stalled. |
| Completed | A label naming the operation — created, retrieved, updated, deleted or completed — plus the result. A delete shows its badge in a distinct, destructive style. |
| Error | A separate error panel, shown both for a tool that fails outright and for one that runs but reports back that it didn't succeed — either way you see the same panel with the message it returned. |
| Empty | Nothing to show, so the whole card — including its title — is removed from the conversation rather than left as an empty shell. A running or errored card is always shown regardless. |
When the assistant reads or writes a record
Create, read, update and delete tools all render their result as one or more record cards for whatever they touched — a contact, invoice, expense, receipt, hearing, legal matter, milestone, file, folder, task or configuration. Reading a single item shows one detail card; reading several shows a list of compact ones, chosen automatically by how much data came back. Writing shows a confirmation banner naming the operation and the record it touched.
Where a record can be opened in eFirm, hovering it reveals a link that jumps straight to the full record there — see HAQQ eFirm. A result the assistant can't locate a record for simply doesn't show that link.
| Record | Status values |
|---|---|
| Invoice | Paid, pending, overdue or cancelled |
| Expense | Approved, rejected or reimbursed |
| Legal matter | Won, lost or settled |
| Milestone | In progress, completed or cancelled |
Searching eFirm, your knowledge base and the web
A search across your eFirm records opens already expanded, showing the keyword you searched in its header. Results are grouped by record type, with a filter pill for each type — plus an All pill — and a count on every one; picking a pill narrows what's shown without re-running the search. Each result opens in eFirm the same way a record card does.
A search of your knowledge base works the same way but organizes results into four independently collapsible sections — company, team, personal and public — each filterable by type, with results routing to a type-specific card (contact, task, legal matter, invoice, expense, calendar item, lead, milestone, reminder, file, a prior AI response, or a generic fallback).
A web search just shows the query and the source links it returned.
When the assistant drafts or edits a document
The moment the assistant creates or updates a document, an inline card appears in the conversation with its title, kind and a body preview. While it's still being written, the preview grows live and auto-scrolls to the newest text — you can't scroll or select it yet — and a status line says whether the assistant is creating a new document or updating an existing one. Maximize and Download stay disabled until drafting finishes; the moment it does, the card becomes clickable and the full panel opens on its own.
Once finished, the card shows a clamped preview, a stamp for how long ago it was last updated, a Maximize action, and a Download menu for PDF or DOCX — disabled while there's no content yet. If that same document is already open in the panel, the card in the conversation collapses to a single line instead of repeating the preview. For how a first draft gets generated in the first place, see AI Drafting.
The artifact panel
Clicking the card, or Maximize, opens the document as an overlay: on a wide screen it splits into your live conversation on the left and the document on the right; on a narrow one it fills the screen with the document alone. The header shows the title and either a “Saving changes” note or how long ago it was last updated, with a close control.
| Action | What it does |
|---|---|
| View changes | Toggles a diff highlighting insertions in green and deletions struck through in red, against the version before it. Disabled on the first version. |
| Previous / Next version | Steps through saved versions. Previous is disabled on the first version, Next on the latest. |
| Copy to clipboard | Copies the document's current content. Disabled while there's no content. |
| Download | Exports to PDF or DOCX with a progress toast. Disabled while there's no content. |
The body is an editable rich-text view of the document — but only while you're looking at the latest version. While the assistant is actively writing, the body switches to a read-only rendered view, and editing resumes once it stops.
Version history and comparing versions
Stepping to an older version makes the document read-only, dims the conversation column behind an overlay, and adds a footer reading “Viewing previous version” with a version indicator and a “Back to latest” button that returns you to the current version and to edit mode.
Asking the assistant to revise the document — from the conversation column beside the open panel — is the other way a version changes, distinct from editing the text yourself directly. While it works, a status banner tracks progress through analyzing the request and applying each change, then confirms once every change is applied, dismissing itself after a few seconds. If nothing needed to change, it says so and dismisses the same way.
What to know before you rely on it
- Edits to the latest version save on their own, a short pause after you stop typing — there's no save button and no confirmation.
- Only the latest version is editable; open an older one and you're read-only until you return to latest.
- A read-only viewer — someone on a shared, read-only conversation — can open a document, browse and compare versions, copy and download, but can't drive new edits.
- A document can't be opened into the full panel while it's still being written; the inline card carries the live progress instead, and the panel opens itself once drafting finishes.
- Text documents — essays, emails, general writing — are what actually opens in the editable panel today. A spreadsheet is referenced by its own icon in a tool result card, but its editor isn't enabled yet, so it doesn't open the same way.