AI & Carbon Intelligence¶

Carbon Connect uses artificial intelligence and carbon science at every stage of the platform. This page explains how these technologies work in plain language, focusing on what they deliver rather than how they are built.

The Matching Algorithm: Five Dimensions of Intelligence¶

When Carbon Connect recommends a grant to a company, it evaluates five different dimensions and combines them into a single match score. Think of it like a hiring manager evaluating a candidate -- they do not just check one thing; they consider qualifications, experience, cultural fit, timing, and references.

Dimension 1: Rule-Based Eligibility (30% of score)¶

What it does: Checks the hard requirements -- the things that would immediately disqualify you.

How it works: The system compares your company's factual data against each grant's stated criteria:

• Country -- Does the grant cover your country? If not, you are immediately disqualified (score = 0).
• Company size -- Does the grant accept companies of your size (micro, small, medium, large)? If not, disqualified.
• Industry codes -- Do your NACE codes overlap with the grant's target sectors? Full match scores highest; same-sector partial match scores 50%.

Why it matters: This eliminates the "false hope" problem. You never waste time on grants you cannot qualify for.

Dimension 2: Semantic Understanding (25% of score)¶

What it does: Reads and understands what your business does and what the grant funds, then measures how well they align.

How it works: The platform converts both your business description and each grant's description into mathematical representations called "embeddings" -- essentially, a numerical fingerprint of meaning. The system then measures how similar these fingerprints are.

For example, a company description mentioning "solar panel installation for commercial buildings" would score highly against a grant funding "renewable energy deployment in the built environment" -- even though they use completely different words. The AI understands they are talking about the same thing.

Why it matters: Traditional keyword search misses grants that use different terminology. Semantic understanding finds relevant opportunities regardless of how they are worded.

Dimension 3: Carbon Alignment (25% of score)¶

What it does: Measures how well your carbon profile matches what the grant is designed to fund.

How it works: The system evaluates multiple carbon-specific factors:

Additionally, grants that are classified as carbon-focused receive a 1.2x bonus -- meaning the final score is boosted by 20% for grants specifically designed to fund decarbonization. This ensures carbon-relevant opportunities rise to the top.

Why it matters: This is what makes Carbon Connect fundamentally different from generic grant platforms. A company with a strong carbon profile will discover funding opportunities that a keyword-based search would never surface.

Dimension 4: Peer Intelligence (10% of score)¶

What it does: Learns from what similar companies have done and uses those patterns to improve recommendations.

How it works: When companies in similar sectors, of similar sizes, in similar countries interact with grants (view them, save them, apply to them), the system identifies patterns. If manufacturing SMEs in Germany consistently save grants from a particular program, similar German manufacturers will see those grants ranked higher.

This is the same principle behind "customers who bought this also bought..." recommendations, applied to grant discovery.

Why it matters: Human behavior signals contain valuable information that algorithms alone cannot capture. Peer intelligence surfaces opportunities that your peers have found valuable.

Dimension 5: Timing and Urgency (10% of score)¶

What it does: Prioritizes grants with approaching deadlines so you do not miss opportunities.

How it works:

Why it matters: A perfect grant match is worthless if the deadline has already passed. Urgency scoring ensures time-sensitive opportunities get the attention they deserve.

Claude AI: The Application Writing Engine¶

What Claude Does¶

When a user selects a grant to apply for, the platform sends a carefully constructed request to Anthropic Claude (specifically, Claude Sonnet 4) that includes:

• The company's profile and carbon data
• The grant's requirements, criteria, and objectives
• The specific application section being generated (executive summary, project description, methodology, etc.)
• Domain-specific instructions about grant writing best practices

Claude produces a professional-quality first draft that the user then reviews and refines.

Why Claude¶

Model Selection¶

The platform supports three Claude models for different use cases:

Climatiq: The Carbon Calculator¶

What Climatiq Does¶

The platform integrates with the Climatiq API to convert raw activity data (kWh of electricity, liters of fuel, kilometers driven) into verified greenhouse gas emissions measured in tonnes of CO2 equivalent (tCO2e).

How It Works¶

1. User enters activity data -- for example, "50,000 kWh of electricity consumed in Germany"
2. Platform calls Climatiq with the activity data and region
3. Climatiq applies the correct emission factor -- for German grid electricity, this is approximately 0.338 kgCO2e per kWh
4. Result is returned -- 16.9 tonnes of CO2 equivalent for Scope 2 emissions

GHG Protocol Compliance¶

All calculations follow the Greenhouse Gas Protocol, the international standard for carbon accounting:

Regional Precision¶

Emission factors vary significantly by country. Electricity in France (largely nuclear) produces far fewer emissions per kWh than electricity in Poland (largely coal). Climatiq provides country-specific and regional emission factors to ensure calculations are accurate for each company's actual location.

How It All Connects¶

The three AI and carbon components work together as an integrated system:

1. Climatiq calculates your carbon profile from raw activity data
2. The matching algorithm uses that profile (along with four other dimensions) to find your best grant matches
3. Claude uses both your profile and the grant details to write your application

This integration is what enables the "10 minutes to set up, seconds to match, minutes to apply" experience that defines Carbon Connect.