6.2 Generating a Document from an Existing Document
Key Takeaways
Reference an existing file when drafting so Copilot rewrites verified content instead of generating unverified prose from the model's own knowledge.
In Word, referencing a source file turns a blank-page task into a transformation task, which materially lowers fabrication risk.
Rewrite, tone adjustment, expand, shorten, and Visualize as a Table act on text you have selected, not on the whole document.
Reusing an approved source document preserves the numbers and commitments that a from-scratch draft would invent.
Always review generated transformations against the source: Copilot preserves meaning well but can drop qualifiers that carry legal weight.
6.2 Generating a Document from an Existing Document
Quick Answer: The blueprint separates "create a new document from a prompt" from "generate a document from an existing document" because the risk profiles differ. Drafting from a prompt asks the model to supply both structure and substance. Drafting from a source document asks it only to re-express substance that already exists — which is why source-grounded generation is the default choice for anything containing figures, dates, or commitments.
Source-Grounded Generation (File-Referenced Drafting)
Source-grounded drafting anchors Copilot's generative process in verified organizational assets. By referencing existing files, Copilot extracts specific domain context, project milestones, budget figures, and stakeholder responsibilities, synthesizing them into a new document structure.
- Ideal Use Cases: Converting a raw project charter into a formal client proposal, synthesizing multiple meeting transcripts into an executive briefing memo, or turning a technical specification into a customer-facing product overview.
- The
/(Forward Slash) Referencing Mechanism: Typing/in the "Draft with Copilot" dialog opens a dynamic file picker displaying recently accessed documents from OneDrive for Business and SharePoint Online. Users can select up to three reference files to ground the generation.
| Operational Dimension | Blank-Canvas Generation | Source-Grounded Generation (/) |
|---|---|---|
| Information Source | Base LLM training knowledge + User prompt instructions | Explicit enterprise files retrieved via Microsoft Graph |
| Prompt Complexity Required | High; must supply context, domain background, and parameters | Medium; prompt defines desired structure and transformation goal |
| Hallucination Risk | Moderate to High if asked for specific corporate facts | Low; model strictly synthesizes grounded source text |
| Citation Provenance | None; content is newly synthesized | Generates references grounded in source documents |
| Reference Limit | N/A (No files attached) | Up to 3 cloud-stored files (.docx, .pptx, .pdf, transcripts) |
Exam Warning: Copilot in Word can only reference files stored in OneDrive for Business or SharePoint Online. Files stored on a local computer drive (e.g.,
C:\Users\Desktop\file.docx) that have not synchronized with the Microsoft 365 cloud tenant cannot be resolved by Microsoft Graph and will not appear in the/file picker.
1. In-Line Transformation & Iterative Refinement
Beyond drafting complete sections, Copilot in Word serves as an intelligent in-line editor. By highlighting any section of existing text, users unlock contextual transformation tools.
┌─────────────────────────────────────────────────────────────────────────────┐
│ IN-LINE TRANSFORMATION TOOLS │
├──────────────────────────┬──────────────────────────┬───────────────────────┤
│ REWRITE WITH COPILOT │ TONE ADJUSTMENT │ VISUALIZE AS A TABLE │
│ │ │ │
│ [Selected Paragraph] │ [Selected Text] │ [Selected Prose/List] │
│ │ │ │ │ │ │
│ ▼ │ ▼ │ ▼ │
│ Alternative phrasing │ • Professional │ Formatted Word Table │
│ with side-by-side │ • Casual │ with auto-generated │
│ comparison view │ • Concise │ column headers │
└──────────────────────────┴──────────────────────────┴───────────────────────┘
1. Rewrite with Copilot
Highlighting a sentence or paragraph and selecting the Copilot icon > Rewrite triggers an automated linguistic refactoring. Copilot generates multiple candidate revisions that enhance clarity, eliminate passive voice, and improve syntactic flow while preserving the core message.
2. Tone Shifting & Register Modification
Copilot provides pre-configured tone settings that instantly adapt selected text to different business contexts:
- Professional: Enhances formal vocabulary, enforces objective phrasing, and removes conversational idioms. Ideal for client proposals, executive summaries, and legal disclosures.
- Casual: Softens rigid structures, introduces conversational transitions, and enhances warmth. Suitable for internal team updates, newsletters, and informal project retrospectives.
- Concise: Aggressively removes filler words, eliminates redundancy, and condenses compound sentences into direct, punchy assertions. Designed for busy executives and high-impact briefing memos.
3. Modifying Length (Expand vs. Shorten)
- Expand: Instructs Copilot to elaborate on core concepts, add supporting justifications, or provide illustrative enterprise examples for high-level bullet points.
- Shorten: Compresses multi-paragraph explanations into executive summaries or single-sentence takeaways while preserving essential data points.
4. Visualize as a Table
Dense narrative text containing comparative metrics, product features, or step-by-step responsibilities can be overwhelming to read. Selecting unstructured text and choosing Visualize as a Table prompts Copilot to:
- Analyze the underlying categorical dimensions in the text.
- Create a formatted Word table with structured column headers.
- Populate the cells with extracted facts, preserving numerical figures and owner attributions.
Reviewing a Derived Document
Transformation is not lossless. Copilot reliably preserves the main assertions of a source but can quietly drop the qualifiers that lawyers and auditors care about — "up to", "subject to board approval", "in the pilot region only". Before a derived document leaves your hands:
- Diff the numbers. Every figure in the output should appear in the source. Any figure that does not is either inferred or invented.
- Check the hedges. Search the source for conditional language and confirm it survived.
- Confirm the scope. A summary that silently promotes a regional pilot into a company-wide programme is the most common and most damaging transformation error.
| Task | Better approach | Why |
|---|---|---|
| Board paper from an approved strategy memo | Reference the memo | Figures and commitments are already agreed |
| Client-facing summary of a signed SOW | Reference the SOW | Scope language must match the contract exactly |
| Brainstorm of possible campaign themes | Prompt-only drafting | No source of truth exists yet, and invention is the point |
On the Exam: When a scenario involves an existing approved artifact — a contract, a policy, a signed proposal — the expected answer references that file rather than describing its contents in the prompt.
An operations lead has drafted a 4-paragraph narrative in Word describing team deliverables, milestone target dates, and assigned owners. The lead wants to convert this unstructured text into an editable 3-column table. Which native Microsoft 365 Copilot feature should be used?
Select the text and choose 'Visualize as a Table' from the Copilot menu.
Copy the text into Excel, prompt Copilot to convert it to a CSV string, and paste it back into Word.
Use 'Rewrite with Copilot' and set the tone parameter to 'Formal Tabular'.
Open the Copilot Side Pane and submit a script command to execute VBA table formatting.
A manager must produce a client-facing summary of a signed statement of work. Which approach best reduces the risk of misstating scope?
Describe the engagement from memory in a detailed prompt
Reference the signed statement of work as the source for the summary
Ask Copilot to search the web for similar statements of work
Draft from a prompt, then ask Copilot to check its own accuracy
Which review step most directly catches the highest-impact error when Copilot derives a summary from a source document?
Checking that the summary is shorter than the source
Confirming the tone matches the intended audience
Verifying that qualifiers and scope limits from the source survived into the summary
Making sure the summary uses headings and bullet points
Sections you finish are checked off in the contents.