User Guide

Overview

ForgeAI Workshop is an AI-powered manufacturing intelligence platform. Upload engineering documents and get structured, actionable data back — scope takeoffs, parts lists, cost estimates, BOM diffs, specs, and readiness scores.

Quick Start

1. 1Sign in or create an account at workshop.forgesuite.ai
2. 2Click a tool card on the dashboard (e.g., Takeoff)
3. 3Upload your PDF document using the Browse Files button
4. 4Wait for processing (typically 15 seconds – 3 minutes depending on document size)
5. 5View results, then export as CSV, PDF, or Markdown

Suggested Workflow

While every tool works independently, here is the recommended flow from RFQ to quote:

BOM Compare and Spec Extraction are standalone tools you can use at any point.

Readiness Check

Score your RFQ package for completeness before you start quoting.

What it checks

The tool runs a two-pass analysis — fast rule-based checks followed by an AI review acting as a senior manufacturing quoting manager. It evaluates:

• Material & certification — alloy, temper, form, governing spec, mill certs, DFARS, ITAR
• Dimensions & tolerances — critical dims, general tolerances, GD&T consistency, unreasonable tolerances
• Finish & coating — surface finish (Ra), plating/coating spec, masking requirements
• Drawing quality — current revision, all views present, scale stated, unclear annotations
• Quality & compliance — FAI, source inspection, Nadcap/AS9100, lot traceability
• Quotability — quantities, delivery timeline, TBD fields, packaging requirements
• Service/construction RFQs — process specs, safety plans, hazmat, crane/rigging, insurance, inspection hold points

Understanding the score

Every issue gets a severity level. The score starts at 100 and deducts points per issue:

85–100: Quote-ready (green). 60–84: Usable with risk (amber). Below 60: Not ready — clarification needed (red).

Best input documents

Takeoff

Quantity takeoff — every line item, spec, risk, and requirement from your RFQ.

What it extracts

Upload an RFQ PDF (up to 10 files) and the AI reads every page, producing a structured breakdown. It automatically detects whether the RFQ is manufacturing, construction, or services and adjusts the output accordingly:

• Project context — project name, RFQ number, buyer info, due date, contract type, currency
• Line items (manufacturing) — part number, material, grade, certifications, dimensions with tolerances, finish, quantity, quantity breaks, lead time, inspection requirements
• Scope items (construction) — item number, scope category (Civil/Structural, MEP, Roofing, Finishes, etc.), specifications with standard numbers, conditions, exclusions
• Technical requirements — requirement text (verbatim), source section, which items it impacts, criticality level
• Risk flags — conflicts, ambiguities, missing info, tight schedules, unusual requirements, supply chain risks, regulatory issues
• Commercial terms — payment terms, liquidated damages, insurance, bonds, warranty, price variation, evaluation criteria
• Submission requirements — format, deadline, required documents, pre-bid meeting details
• Quote readiness — overall assessment (ready / needs info / not ready) plus a specific list of what is missing to produce a quote

Confidence scores

Every extracted item gets a confidence score (0.0–1.0):

• 0.95–1.0 Clear quantities, specs, and standard references
• 0.80–0.94 Some ambiguity — finish or dimensions inferred
• Below 0.80 Poor OCR, unclear specs, or missing information — manual review recommended

Best input documents

After extraction

Takeoff results can flow directly into Quote Builder. Click “Generate Quote” to have the AI produce a work breakdown with operations, materials, and labor estimates for each item — then apply your shop rates for a dollar quote.

Drawing Extraction

Upload engineering drawings and the AI visually reads each page to extract parts, materials, and build data.

What it extracts

Unlike Takeoff (which reads text), Drawing Extraction renders each page as an image and uses AI vision to identify components from balloon callouts, BOM tables, leader lines, and visual geometry:

• Parts list — item number, part number, description, quantity, material, size/dimensions
• Drawing metadata — drawing number, title, revision, scale, project name
• Document notes — special instructions (e.g., “Apply Loctite”, “Break all sharp edges”), certifications, organization
• Item classification — each item tagged as part (discrete component), feature (holes, pockets, chamfers), or annotation (GD&T, datum targets). Only parts feed into quoting.
• Cross-page deduplication — items appearing on multiple pages (common in multi-view drawings) are flagged as probable duplicates
• Confidence scores — 0.9+ from BOM tables, 0.7–0.9 from balloon callouts, 0.5–0.7 from inferred geometry

Best input documents

After extraction

Several follow-up actions are available directly from the results view:

• Generate Quote — creates a work breakdown and feeds into Quote Builder
• Review Requirements — AI checks each item for missing material, quantities, special requirements, and standard hardware
• Verify BOM — upload an existing BOM to cross-reference against the extracted drawing data

Quote Builder

Turn AI-generated work breakdowns into dollar quotes using your shop’s configured rates.

How it works

Quote Builder is a pricing engine that runs after Takeoff or Drawing Extraction. The flow is:

1. Takeoff or Drawing Extraction extracts scope from your documents
2. Work Breakdown (generated automatically) — the AI produces detailed operations, materials, and labor estimates for each item
3. Quote Builder matches operations against your shop rates and calculates costs
4. AI Validation reviews the finished quote for anomalies (cost spikes, zero labor, missing scope)

What it generates

• Per line item: material cost, labor hours, labor cost, setup cost, overhead, profit, line total
• Operations detail: each operation (CNC turning, milling, grinding, etc.) matched to your process rate
• Materials detail: each material matched to your configured unit costs with markup
• Quote totals: subtotals for material, labor, setup, overhead, and profit — plus grand total
• Confidence level: firm, budgetary, or rough order of magnitude — based on how complete the input data is
• Validation results: AI review with grade (A–F), warnings, and cost anomaly flags
• Unconfigured alerts: lists any operations or materials that didn’t match a rate, so you know what to add

Shop Config (required first)

Before generating quotes, configure your shop in the Shop Config section:

• Process rates — hourly rates for each operation (e.g., CNC Turning $85/hr, Welding $65/hr)
• Material costs — unit costs for raw materials (e.g., 6061 Aluminum $3.50/lb)
• Overhead multiplier — applied to labor cost (default 35%)
• Profit margin — applied to all costs (default 15%)
• Hours padding — buffer added to labor estimates (default 10%)

You can import rates from CSV, or use Suggest from Job after your first work breakdown — it discovers what operations and materials your job needs and creates placeholder entries for you to price.

Editing quotes

Quotes are fully editable after generation. Adjust any line item cost, add manual lines, delete items, or change overhead/profit margins. The system automatically recalculates totals. Re-quoting the same job increments the version number (Q-0001 v1 → v2).

Export formats

Quotes can be exported in five formats:

• CSV — line items table with subtotals
• PDF — professional printable quote with branding and confidence banner
• HTML — same as PDF, viewable in browser
• Markdown — clean text format with metadata and tables
• Work Order — shop-floor document with routing/operations per part, material requirements, risk flags, and inspection checkpoints

BOM Compare

Upload two BOMs and the AI matches parts across them — even when part numbers, naming, or column layouts differ.

How matching works

The tool uses AI to intelligently match parts across the two BOMs using multiple strategies:

• Exact part number — straightforward PN equality
• Normalized part number — handles revision suffixes (e.g., NRF52833-QIAA vs NRF52833-QIAA-R7)
• Manufacturer part number — matches even when internal PNs differ
• Description matching — when part numbering schemes are completely different
• Engineering knowledge — non-obvious matches like capacitor codes (100nF = 0.1µF = 104 in cap code)

Output categories

Conflicts include severity (minor/major), a list of specific changes (e.g., “Quantity: 17 → 18”), and an engineering explanation of why the change matters.

Supported file formats

• CSV (best) — auto-detects delimiter (comma, semicolon, tab) and skips metadata headers
• Excel (.xlsx, .xls) — reads calculated values from the active sheet
• PDF — uses AI to extract tabular BOM data from the document

Additional modes

• Drawing-to-Drawing Compare — compare extracted parts lists from two drawing jobs
• Drawing BOM Verification — upload a BOM to cross-reference against an AI-extracted drawing. Produces color-coded results: green matches, red conflicts, amber drawing-only, blue BOM-only

Spec Extraction

Extract structured specifications from component and material datasheets.

What it extracts

• Chemical compositions (element percentages with min/max ranges)
• Mechanical properties (tensile strength, yield strength, elongation, hardness)
• Physical properties (density, melting point, thermal conductivity, resistivity)
• Electrical specifications (voltage, current, data rates, timing parameters)
• Dimensional data (bore, diameter, width, package sizes)
• Performance ratings (load capacity, speed limits, temperature ranges)
• Governing standards (AMS, UNS, JEDEC, CE, RoHS, etc.)
• Applications and features

Best input documents

Understanding results

Results include:

• Component name and manufacturer — automatically identified from the document
• Governing specs — industry standards shown as badges (e.g., AMS 5604, UNS S17400)
• Applications — intended uses extracted from the document
• Specifications table — structured parameter/value/unit/conditions data
• Features — product features and capabilities
• Cross-references — equivalent parts and alternate designations

Supported File Formats

Limits & Billing

Usage is measured in pages processed per billing period. Each page of an uploaded PDF counts as one page. Quote generation counts as 1 page per quote.

Processing time varies by document complexity: 15–30 seconds for simple documents, 1–3 minutes for medium documents, up to 8 minutes for large drawing packages (50+ pages). BOM Compare counts 50 line items as 1 page. Your dashboard shows real-time page usage.

FAQ