Build a Blueprint from an Existing Document Using AI
This tutorial walks through using AI Design to create a new blueprint from an existing manual report — a Word template or PDF that your team has been filling in manually. Instead of building the blueprint field by field, you let AI analyse the document and suggest the structure.
Scenario: Your team fills in a monthly Word document called "Monthly Progress Report". It has sections for project status, a cost summary table, a risk register table, and a photo gallery. You want to digitise this into Report Forge.
Before you start
- Have the existing report template ready (Word .docx, PDF, or a filled-in example).
- Confirm the ✨ AI Design button is visible in the Blueprint Designer header. If not, contact your Workspace Admin to enable AI access.
- Decide on a Short Code for the blueprint (e.g.
MPRfor Monthly Progress Report).
Step 1 — Create a new blueprint
- Navigate to Report Forge → Blueprints.
- Click + New Blueprint.
- Enter the blueprint details:
- Name:
Monthly Progress Report - Description:
Monthly project status, cost, risks, and photos for site managers. - Short Code:
MPR
- Name:
- Click Save.
The Blueprint Designer opens. The tabs (Design, Form Designer, Preview) are now visible.
Step 2 — Open AI Design
- Click ✨ AI Design in the header (purple button, robot icon).
- The AI Blueprint Design panel opens on the right (titled "✨ AI Blueprint Design").
Step 3 — Upload the source document
- Drag your Word template (.docx) onto the upload area, or click to browse.
- If you also have a filled-in PDF example (which shows the actual data, not just field labels), upload that too — AI extracts better field types when it sees example values.
Tip: Upload both the template and a filled-in example if available. The template gives field names; the example gives context for field types.
Supported formats: JPEG, PNG, WebP, PDF (all pages), Excel (.xlsx), Word (.docx). Up to 20 files per session.
Step 4 — Add instructions
In the Instructions field, type guidance to help AI make better decisions:
This is a monthly progress report for construction projects.
Sections: Project Status (single entry), Cost Summary (table with monthly rows),
Risk Register (table with one risk per row), Photo Gallery (images with captions).
For Cost Summary, column "This Month" = PeriodCost (currency), "Budget" = BudgetCost (currency).
Risk severity values are: Critical, High, Medium, Low.
RAG status values are: Red, Amber, Green.
Specific instructions significantly improve AI accuracy, especially for section types and field type classification.
Step 5 — Select the AI model
- GPT-4o ($$) — good for standard Word templates and PDFs.
- Claude Sonnet 4.6 ($$$) — better for complex documents with dense tables, mixed content, or ambiguous structure.
For a standard monthly report template, GPT-4o works well. If you get poor results, re-run with Claude Sonnet 4.6.
Step 6 — Analyze the document
Click Analyze. The AI processes your files (5–30 seconds).
Step 7 — Review the suggested structure
The review panel shows the AI's proposed blueprint structure:
Example output:
Section: Project Status [single_row] [High confidence]
├── Project Name text required [High]
├── Report Date date required [High]
├── Overall Status select required [High] Options: On Track, At Risk, Delayed
├── RAG Status select required [High] Options: Red, Amber, Green
└── Executive Summary textarea — [Medium]
Section: Cost Summary [multi_row] [High confidence]
├── Period text required [High]
├── PeriodCost currency required [High]
├── BudgetCost currency required [High]
└── Variance number — [Medium]
Section: Risk Register [multi_row] [High confidence]
├── Risk ID text required [High]
├── Description textarea required [High]
├── Severity select required [High] Options: Critical, High, Medium, Low
├── RAG Status select required [High] Options: Red, Amber, Green
└── Mitigation textarea — [Medium]
Section: Photo Gallery [multi_row] [Medium confidence]
├── Photo image — [Medium]
└── Caption text — [Medium]
Review each section and field:
- Confirm section types:
single_rowvsmulti_row. - Check field types: is "Variance" better as
currencythannumber? (Change it after applying.) - Check required flags: are all required fields actually required?
- Review confidence tags: Green = high confidence; Gold = medium; Orange = low (review carefully).
Step 8 — Modify a suggestion before accepting
You notice "Variance" was suggested as number but should be currency. You also want to rename "Photo Gallery" to "Site Photos".
You can:
- Accept the suggestions as-is and change them in the Blueprint Designer after applying (easier for most changes).
- Or note the changes to make immediately after applying.
For this tutorial, accept the suggestions and fix them after.
Step 9 — Accept and apply
Click Accept & Apply.
The AI creates:
- All suggested sections (with correct single/multi-row settings).
- All suggested fields (with types, required flags, and options).
- A progress bar shows: "Creating sections... Adding fields..."
- Result summary: "Created 4 sections and 16 fields."
The blueprint's Design tab updates immediately to show all sections and fields.
Step 10 — Review and refine the created structure
Open the Design tab. Click each section to review its fields.
Fix the issues identified in Step 8:
-
Click the Cost Summary section → click the Variance field → open Field Settings → change type from
NumbertoCurrency. Click Apply. -
Click the Photo Gallery section → click the section name inline → rename to
Site Photos. Press Enter.
Add a field AI missed:
The Word template had a "Reporting Period" field (e.g. "March 2026") in the Project Status section that AI didn't extract:
- Select Project Status section.
- Click + Add Field.
- Choose Text type.
- Name it
Reporting Period, mark as Required, add placeholdere.g. March 2026.
Step 11 — Check section display modes
In the Section Settings panel (right column), verify each section's display mode:
- Project Status — display mode: Form (since it has a long narrative text field).
- Cost Summary — display mode: Grid (tabular data, many rows).
- Risk Register — display mode: Grid (tabular data, many rows).
- Site Photos — display mode: Form (file upload with caption is easier in form mode).
Change any that need adjusting via the Display mode setting in Section Settings.
Step 12 — Preview the blueprint
Click the Preview tab to see how the blueprint will look to contributors. Verify:
- Sections appear in the right order.
- Fields are in a logical sequence within each section.
- Section display modes look correct.
Reorder sections or fields in the Design tab if needed.
Step 13 — Publish the blueprint
When satisfied with the structure:
- Click Publish in the Blueprint Designer header.
- Confirm the action.
- The status badge changes to Published.
Contributors can now create editions from this blueprint.
Time comparison
| Task | Manual approach | With AI Design |
|---|---|---|
| Create 4 sections | 5 minutes | 30 seconds |
| Add 16 fields with types | 20 minutes | 30 seconds |
| Review and fix suggestions | — | 5 minutes |
| Set display modes and order | 5 minutes | 3 minutes |
| Total | ~30 minutes | ~9 minutes |
AI Design is most valuable for blueprints with many fields. For blueprints with fewer than 5 simple fields, manual creation is equally fast.
Related
- AI Design reference — complete feature reference including troubleshooting
- Blueprint Designer Overview — all Blueprint Designer panels
- Fields Panel — adding and configuring fields manually
- Publishing and Lifecycle — what publishing means for editions
- AI Fill tutorial — use AI to fill an edition from a source document