What the assistant can do in your document

What the assistant can do in your Word document

The assistant works directly on the document you have open. It can read it (your selection, the outline, a full map of paragraphs, tables, images and charts), search it, insert rich content including tables and charts, edit or delete text, run find and replace, format text, and manage comments.

It also combines document work with your project data — for example, pulling open issues from your project and inserting them as a table in the document. Just describe what you want in plain language.

Reading your document — selection, outline, full map, search, statistics

The assistant can look at your document in several ways:

• Your current selection: select some text and ask about it, e.g. "Rewrite this selection more formally".
• The outline: it can read the heading structure to understand how the document is organised.
• A full document map: every paragraph, table, image, and chart, each with its position, so it can target edits precisely.
• Text search: "Find every mention of 'generator'".
• Statistics: word counts and similar figures about the document.
• Pending changes: "What is still waiting for review?" lists the tracked changes and proposals in the document.

Reading never changes your document and never requires approval.

Inserting rich content — headings, lists, tables, code, and charts

Ask the assistant to write content and it inserts properly formatted Word content, not plain text:

• Headings, bold and italic text, and nested lists.
• Tables — e.g. "Insert a table of the open issues for this asset".
• Code blocks.
• Charts: column, bar, line and pie, e.g. "Add a bar chart of issue priorities". Each chart is inserted as a picture with a description for screen readers and, if you ask, a numbered caption.

Content can be inserted at the end of the document or at a specific position (see the next section).

Inserting at a specific position

You control where new content goes. The assistant can insert at the end of the document or after a specific paragraph. Point it at the spot in plain language:

• "Insert the summary after the 'Scope' heading".
• "Add this table at the end of section 2".
• "Put a new heading right before the conclusion".

Because the assistant can map every paragraph with its position, it can place content precisely where you ask.

Editing text — replace, delete, find and replace, format

The assistant can change existing text:

• Replace or delete a range of paragraphs: "Replace section 3 with the updated procedure".
• Replace or delete your current selection: select text and say "Delete this" or "Rewrite this paragraph".
• Document-wide find and replace: "Replace every occurrence of 'Site A' with 'Site B'".
• Change wording inside a paragraph: "Fix the voltage in the second paragraph — it should be 400 V". Where possible, only the words that change are marked as changed, not the whole paragraph.
• Format the selection: change the font of the text you have selected.

In Review mode (the default), edits to existing text arrive as tracked changes that you accept or reject. See "Review modes" in the assistant help. Depending on your approval mode, an edit may also be shown to you for approval before it runs.

Deleting tables, charts, and images

The assistant can remove objects from the document, not just text. Tables, charts, and images are identified by their position in the document map, so you can say things like:

• "Delete the second table".
• "Remove the chart in section 4".
• "Delete the image after the introduction".

If you are unsure which object is which, ask for the document map first — the assistant lists every table, image, and chart with its index.

Tables, footnotes, paragraph styles, and ready-made blocks

Beyond plain text, the assistant can work with these parts of a document:

• Existing tables: change cell text, add or delete rows and columns, set a table style, repeat the header row, or set column widths — "Add a row for pump P-102 to the second table".
• Footnotes and endnotes: "Add a footnote after 'IEC 61439' with the full standard title". This needs a newer Word build.
• Paragraph styles and layout: apply a heading or other built-in style, change alignment and spacing, or make text bold, italic, or highlighted — "Make these three lines Heading 2".
• Ready-made blocks: a callout, a status table, a row of key figures, a figure with a numbered caption, a signature block, a cover page, or a checklist — "Insert a sign-off block at the end". Blocks use your document's own fonts and styles.

In Review mode, a new block arrives as a proposal you keep or discard, and table or footnote edits arrive as tracked changes.

Working with comments

The assistant can manage Word comments for you:

• List all comments with their surrounding context.
• Add a comment anchored to your current selection, to a specific paragraph, or to text it finds by searching — e.g. "Find every mention of 'generator' and add a comment asking for a spec reference".
• Reply to a comment.
• Resolve or re-open a comment.
• Delete a comment.

Comments are added straight away. They are not tracked changes, so they get no review card, whatever review mode you use.

Note: comment features need a newer Word build (Word API 1.4+). On older versions of Word, the assistant cannot work with comments while everything else keeps working.

Approvals — you confirm before the document changes

Every action that modifies your document is a write, and writes follow your chat's approval mode:

• Auto: edits run without asking.
• Ask for critical: only critical actions ask first.
• Ask for edits: every document edit shows an approval card first — nothing changes until you approve.

The approval card lists each pending action with a checkbox, so you can approve some and decline others. For document edits it also previews the text. Reading your document never asks for approval.

Approvals happen BEFORE an edit. Reviewing happens AFTER it: in Review mode the edit lands as tracked changes or as a proposal, and you accept, reject, keep, or discard it. See "Review modes" and "How to review the assistant's edits" in the assistant help.

Good example prompts for document work

Some prompts that show what the assistant does well in Word:

• "Summarise the open issues for asset UPS-001 and insert a table here".
• "Add a commissioning summary section with a bar chart of issue priorities".
• "Find every mention of 'generator' and add a comment asking for a spec reference".
• "Replace section 3 with the updated procedure".
• "Give me the document outline and statistics".

Combine project data with document edits in a single ask — that is where the add-in shines.

Limits and good to know

A few things to keep in mind:

• The assistant works on the document currently open next to the pane, and the pane must be open for it to act.
• Edits happen in YOUR open copy of the document — you keep normal control over saving.
• Charts are pictures, not Word chart objects: they do not open in Word's chart editor. To change one, ask the assistant to replace it.
• Comments need a newer Word build (Word API 1.4+).
• If a write times out, the assistant checks the document with a read before retrying, so an edit is not applied twice.
• An action that cannot finish fails with a clear error after about 4 minutes instead of hanging. An open dialog box in Word is a common cause.
• The assistant never accepts or rejects tracked changes for you. You decide, in the chat or on Word's Review tab.
• Tracked changes appear under your own name in Word, because Word records the signed-in user as the author.
• Footnotes, comments, and tracked changes need a newer Word build. On a Word version that cannot track changes for the add-in, edits are applied directly.
• Project actions (creating issues, reading documents, and so on) always respect your project permissions. In personal mode (My Workspace) there are no project actions.