How to write effective prompts for Build with AI

Build with AI transforms your process descriptions into complete carbon accounting systems. The quality of your prompt directly impacts the accuracy and completeness of your generated system. This guide shows you how to write prompts that produce professional-grade results.

Understanding How Build with AI Works

Build with AI analyzes your description and uploaded documents to create a comprehensive process flow model following Life Cycle Assessment (LCA) methodology. The AI:

• Maps all material and energy flows through your process
• Identifies emission sources and connects them logically
• Structures processes in sequential manufacturing order
• Applies appropriate units and calculations
• Validates that all inputs appear in the final system

Your role: Provide clear, complete information so the AI can build an accurate model.

The Four Essential Elements

Every effective prompt includes these four components:

1. System Title (Required)

What it is: The clear, specific name of your system

Best practices:

• Use descriptive, industry-standard terminology
• Include the product or process name
• Keep it concise (3-8 words)

Examples:

Good titles:

• "Aluminum Window Frame Production"
• "Corrugated Cardboard Box Manufacturing"
• "Office Building HVAC System Operation"
• "Plastic Injection Molding - Automotive Parts"

Avoid:

• "Manufacturing Process" (too vague)
• "My Company's Main Product" (not descriptive)
• "Process 1" (meaningless)

2. Detailed Description (Required)

What it is: A comprehensive explanation of your process, inputs, outputs, and operational context

Structure your description with these elements:

Industry Sector and Location

Tell the AI what industry you're in and where you operate:

Industry: Automotive component manufacturing
Location: Stuttgart, Germany
Facility type: ISO 14001 certified manufacturing plant

Why it matters: Regional differences in energy grids, transportation infrastructure, and material availability significantly impact carbon footprints. A steel mill in Germany uses a different electricity mix than one in China.

Main Inputs

List all materials, energy sources, and resources that enter your process:

Materials:

• Raw materials (metals, plastics, chemicals, natural resources)
• Intermediate products (semi-finished components)
• Packaging materials
• Process aids (lubricants, coolants, cleaning agents)

Energy:

• Electricity (specify if renewable)
• Natural gas
• Diesel/gasoline
• Steam
• Compressed air

Example:

Main inputs:
- Aluminum ingots (6061 alloy): 50 kg per window frame
- Electricity (German grid mix): 45 kWh per batch
- Natural gas for heating: 15 m³ per batch
- Protective coating materials: 2.5 kg per frame
- Packaging: corrugated cardboard, 3 kg per unit

Key Outputs and Byproducts

Describe what your process produces:

Primary outputs:

• Main product with specifications
• Quantity produced per cycle/batch

Byproducts and waste:

• Co-products that have value
• Scrap materials (identify if recyclable)
• Waste streams requiring disposal
• Emissions to air or water

Example:

Outputs:
- Window frames: 100 units per production run
- Aluminum scrap: 8 kg per batch (recycled on-site)
- Wastewater from coating process: 150 liters per batch
- Packaging waste: mixed materials, sent to recycling facility

Process Steps (Recommended)

Describe major stages in your production workflow:

Manufacturing sequence:
1. Aluminum extrusion and cutting
2. CNC machining for joints and hardware holes
3. Surface treatment and anodizing
4. Protective coating application
5. Quality inspection and testing
6. Packaging and palletization

Why it matters: Helps the AI structure processes in the correct sequential order and identify where emissions occur.

Standards and Certifications

Mention relevant compliance requirements:

Standards:
- ISO 14067 (Product Carbon Footprint)
- EN 15804 (EPD for construction products)
- Cradle-to-gate system boundary

3. Supporting Documents (Highly Recommended)

What to upload: Technical documents that provide detailed specifications

Supported formats: PDF, JPG, PNG, WEBP (up to 5 files, 20 MB each)

Document Types and What They Provide:

Bill of Materials (BOM)

• Complete list of components and quantities
• Part numbers and specifications
• Assembly structure

Why critical: The AI validates that 100% of BOM items appear as inputs in the generated system. Missing a BOM item means incomplete carbon accounting.

Environmental Product Declarations (EPDs)

• Verified environmental data for materials
• Carbon intensity values
• Life cycle impact information

Technical Specifications

• Process parameters (temperatures, pressures, cycle times)
• Equipment specifications
• Quality standards

Process Flow Diagrams

• Visual representations of your process
• Material flow sequences
• Equipment connections

Example document package for aluminum window frame production:

1. BOM_WindowFrame_6061.pdf (material list with quantities)
2. EPD_Aluminum_Extrusion.pdf (verified aluminum data)
3. TechSpec_CoatingProcess.pdf (coating material specifications)
4. ProcessFlow_Manufacturing.jpg (visual process diagram)

Document preparation tips:

• Ensure text is clear and readable (not handwritten or blurry scans)
• Include units of measurement
• Highlight or annotate critical information
• Provide current, up-to-date documents

4. Carbon Estimation Display (Optional)

What it does: Shows estimated carbon footprint directly in your system title

When to enable:

• Comparing multiple design scenarios quickly
• Presenting rough estimates to stakeholders
• Initial assessments before detailed data entry
• Workshop sessions needing immediate feedback

When to disable:

• Final reporting or compliance documentation
• When precise calculations are required
• Regulatory submissions

Remember: These are estimates for planning purposes only. Complete data entry and verification are required for accurate carbon footprinting.

Complete Prompt Examples

Example 1: Food & Beverage Manufacturing

System Title: "Organic Pasta Production - 500g Package"

Description:

Industry: Food manufacturing (organic pasta production)
Location: Bologna, Italy
Facility: HACCP certified, 5,000 m² production area
 Main inputs:
- Organic durum wheat semolina: 0.52 kg per package
- Water: 0.15 liters per kg of dough
- Electricity (Italian grid, 30% renewable): 0.8 kWh per batch (200 packages)
- Natural gas for drying: 1.2 m³ per batch
- Packaging: cardboard box (40g), plastic window (5g)
 Key outputs:
- 500g pasta packages: 200 units per batch
- Production waste (broken pasta): 2% of output, sent to animal feed
- Packaging waste: collected for recycling
 Process steps:
1. Ingredient mixing and dough preparation
2. Extrusion and shaping
3. Drying (60°C, 12 hours)
4. Quality control and metal detection
5. Packaging and labeling
6. Palletization for distribution
 Standards: ISO 22000 (Food Safety), organic certification
System boundary: Cradle-to-gate (ingredients to packaged product)

Documents uploaded:

• BOM_Organic_Pasta_Recipe.pdf
• EPD_Durum_Wheat_Italy.pdf
• Packaging_Specifications.xlsx

Example 2: Electronics Manufacturing

System Title: "Smartphone Circuit Board Assembly"

Description:

Industry: Electronics manufacturing (PCB assembly)
Location: Shenzhen, China
Facility: ISO 14001, RoHS compliant
 Main inputs:
- Bare PCB (FR-4, 6-layer): 1 unit per phone
- Integrated circuits: 15 components (see BOM for details)
- Resistors and capacitors: 85 surface-mount components
- Solder paste (lead-free, SAC305): 2.5 grams per board
- Electricity (Guangdong grid): 0.25 kWh per board
- Nitrogen gas for reflow: 0.5 m³ per batch (50 boards)
 Key outputs:
- Assembled circuit boards: 50 units per batch
- Solder waste: 50 grams per batch (hazardous waste)
- Component packaging waste: mixed materials
 Process steps:
1. Solder paste application (stencil printing)
2. Component placement (pick-and-place automation)
3. Reflow soldering (nitrogen atmosphere, peak 245°C)
4. Automated optical inspection (AOI)
5. Functional testing
6. Conformal coating application
7. Final inspection and ESD packaging
 Standards: IPC-A-610 (assembly standards), ISO 14067
System boundary: Component arrival to assembled PCB

Documents uploaded:

• BOM_Smartphone_MainBoard_v3.xlsx (complete component list)
• PCB_Specifications.pdf
• Assembly_Process_Flow.pdf
• Component_EPDs_Compilation.pdf

Example 3: Construction Materials

System Title: "Ready-Mix Concrete Production - C30/37 Grade"

Description:

Industry: Construction materials (concrete batching)
Location: Munich, Germany
Facility: 50,000 m³ annual production capacity
 Main inputs per cubic meter:
- Portland cement CEM I 42.5 R: 320 kg
- Natural aggregates (0-16mm): 800 kg
- Sand (0-4mm): 900 kg
- Water: 185 liters
- Plasticizer admixture: 1.5 kg
- Electricity for mixing: 5 kWh per m³
- Diesel for truck delivery: 2.5 liters per m³ (average 15km delivery)
 Key outputs:
- Ready-mix concrete: 1 m³ (2,400 kg)
- Wash water from mixer cleaning: 50 liters per batch (recycled)
- Returned concrete waste: 2% (recycled as aggregate)
 Process steps:
1. Aggregate and sand delivery and storage
2. Material weighing and batching
3. Mixing in rotating drum (6 minutes)
4. Loading into delivery truck
5. Transportation to construction site
6. Placement and finishing (not included in boundary)
 Standards: EN 206 (concrete specification), ISO 14025 (EPD)
System boundary: Raw materials to delivery at construction site
Functional unit: 1 cubic meter of C30/37 ready-mix concrete

Documents uploaded:

• Material_Certificates_Cement_Aggregates.pdf
• EPD_Portland_Cement_Germany.pdf
• Concrete_Mix_Design_C30.pdf

Common Mistakes and How to Avoid Them

Mistake 1: Vague or Generic Descriptions

Problem:

"We make windows. We use aluminum and glass. 
We also use some electricity."

Why it fails:

• No quantities specified
• Missing location and industry context
• No process steps
• Incomplete input list

Solution:

Industry: Residential window manufacturing
Location: Portland, Oregon, USA
 Main inputs per window unit:
- Aluminum profiles (6063-T5): 12 kg
- Double-pane glass: 2.5 m² (4mm each pane)
- Argon gas fill: 0.015 m³
- EPDM gaskets: 0.8 kg
- Stainless steel hardware: 0.5 kg
- Electricity (Pacific Northwest grid, 70% hydro): 8 kWh per unit
 Process steps:
1. Aluminum profile cutting and machining
2. Glass unit assembly and argon filling
3. Frame assembly and gasket installation
4. Hardware installation
5. Quality testing (water resistance, air leakage)
6. Packaging with protective corner guards
 Standards: NFRC certification, ENERGY STAR qualified

Mistake 2: Inconsistent Units or Missing Quantities

Problem:

Main inputs:
- Steel: some amount
- Electricity: a lot
- Paint: enough for coating

Why it fails:

• No specific quantities
• Cannot calculate carbon footprint
• Impossible to validate completeness

Solution:

Main inputs per 1,000 units:
- Cold-rolled steel sheet (1.5mm thickness): 500 kg
- Electricity (Texas grid): 150 kWh
- Powder coating paint: 25 kg
- Natural gas (curing oven): 40 m³

Mistake 3: Uploading Incomplete or Illegible Documents

Problem:

• Handwritten BOM that's difficult to read
• Scanned document with poor resolution
• Partial information (missing quantities or specifications)
• Documents in unsupported formats
• Documents more than 200 pages

Solution:

• Use typed or clearly printed documents
• Ensure high-resolution scans (minimum 300 DPI)
• Include all relevant pages (don't crop critical information)
• Convert unsupported formats to PDF before uploading
• Annotate documents to highlight key information

Mistake 4: Ignoring Regional Context

Problem:

"Manufacturing process using electricity and natural gas"

Why it fails:

• Energy grids vary dramatically by region
• German electricity has different carbon intensity than Chinese electricity
• Transportation distances affect distribution emissions

Solution:

Location: Shanghai, China
Energy sources:
- Electricity: East China Power Grid (coal-heavy mix, ~0.7 kg CO₂/kWh)
- Natural gas: pipeline supply from West-East Pipeline
 Transportation:
- Raw materials: average 500 km by truck from Jiangsu suppliers
- Finished goods distribution: 80% domestic (average 1,200 km), 
  20% export via Port of Shanghai

Mistake 5: Mixing Different Functional Units

Problem:

"Production includes 100 bottles per hour, 
electricity is 50 kWh per day, 
plastic pellets are ordered in 25 kg bags"

Why it fails:

• Inconsistent time periods (per hour vs. per day)
• Mixed units (bags vs. kg)
• Cannot normalize to functional unit

Solution:

Functional unit: 1,000 PET bottles (500ml capacity)
 Inputs per functional unit:
- PET resin pellets: 35 kg
- Electricity: 12 kWh
- Compressed air: 5 m³
- Cooling water: 150 liters

Validation Checklist

Before clicking "Build," verify your prompt includes:

Essential elements: 
☐ Clear, descriptive system title
☐ Industry sector specified
☐ Geographic location provided
☐ All material inputs listed with quantities
☐ All energy inputs listed with quantities
☐ Main outputs and byproducts described
☐ Process steps outlined (at minimum, major stages)

Quality indicators: 
☐ Quantities include specific units (kg, kWh, m³, etc.)
☐ Consistent functional unit throughout
☐ Supporting documents uploaded (BOM strongly recommended)
☐ Regional context for energy sources provided
☐ Standards or certifications mentioned (if applicable)

Completeness check: 
☐ Every item in uploaded BOM appears in description OR description clearly references "see uploaded BOM for complete list"
☐ System boundary clearly defined
☐ Time frame specified (per unit, per batch, per hour, etc.)

What Happens After You Click "Build"

AI Processing Steps:

1. Analyzes your description for industry, location, inputs, outputs, and process structure
2. Extracts data from uploaded documents (BOM, EPDs, specifications)
3. Maps material and energy flows through sequential process steps
4. Validates completeness by checking that all BOM items appear as inputs
5. Structures the system as a directed acyclic graph (DAG) following LCA methodology
6. Assigns appropriate units from standardized LCA nomenclature
7. Generates the system model with all processes, inputs, outputs, and connections

Processing time: Typically 30-60 seconds, depending on system complexity

What You'll See:

Your generated system opens in Design Mode with:

• All processes arranged in logical sequence
• Input nodes connected to processes
• Output nodes showing products and byproducts
• Direct emission sources identified (in preparation)
• Visual process flow diagram

Next steps:

1. Review the structure in Design Mode
2. Switch to Entry Mode to add specific emission data and quantities
3. Use "Suggest LCI with AI" to quickly populate emission factors
4. Verify completeness by checking that all expected inputs appear
5. Analyze results once data entry is complete

Troubleshooting

Problem: Generated system is missing inputs I described

Possible causes:

• Input was mentioned ambiguously in description
• Quantity not specified
• Item mentioned in document but not cross-referenced in text

Solution:

• Edit the system in Design Mode to add missing inputs manually
• Regenerate with clearer description: explicitly list each input with quantities
• Ensure uploaded documents are readable (not blurry scans)

Problem: Process sequence doesn't match my actual workflow

Solution:

• In Design Mode, rearrange processes by dragging nodes
• Add intermediate processes if needed
• The initial AI generation is a starting point—customize to match your specific operations

Problem: AI suggested inappropriate units for my materials

Possible cause:

• Material type was unclear in description
• Ambiguous wording

Solution:

• In Entry Mode, change units using the dropdown
• Regenerate with explicit unit specification: "Cold-rolled steel: 500 kilograms (not tonnes)"

Problem: Carbon estimate seems unrealistic

Remember:

• Initial estimates use database averages
• Accuracy improves when you add supplier-specific data in Entry Mode
• Regional variations can cause significant differences
• Check that location and energy sources were correctly specified

Next steps:

• Review system in Entry Mode
• Replace generic emission factors with supplier-specific data
• Use "Suggest LCI with AI" for better regional matches
• Verify process-specific parameters

Best Practices Summary

For Optimal Results:

1. Be specific with quantities - Include exact amounts with units
2. Provide regional context - Location affects energy mix and transportation
3. Upload complete BOMs - The AI validates every item appears as input
4. Describe process sequence - Helps AI structure flows logically
5. Include supporting documents - EPDs and specifications improve accuracy
6. Mention relevant standards - Guides system boundary and methodology choices
7. Use consistent units - Normalize everything to the same functional unit
8. Specify byproducts and waste - Complete accounting includes all outputs

Remember:

✅ More detail = better results
✅ Documents enhance accuracy
✅ Clear structure helps AI organize processes
✅ You can refine the generated system in Design Mode
✅ Entry Mode is where you'll add precise supplier data

Related Resources

Getting started with systems:

• Quick Start: Build Your First System
• Design Your System (Design Mode)
• Add Data to Your System (Entry Mode)

AI features:

• Build with AI: AI-Powered System Builder
• Suggest LCI with AI: AI-Powered Emission Data Selection

Understanding carbon accounting:

• Reference Data Library
• Scope 1, 2, and 3 Emissions

Need help? Contact support@carbonsig.com or use the in-app Feedback tool.

Last updated: November 2025