Koi Engine:
AI-Accelerated Modeling

Last Updated: 2025-06-09

---

Overview

Koi offers unrivaled speed and accuracy in model delivery. As the first and only platform to provide on-demand, customized emissions forecasts, it is designed to equip innovators and investors with analytics that accommodate different use cases, timelines, and technologies. Koi models are built on a validated Data Lake that integrates extensively curated, foundational data from reliable sources such as the IEA, SBTi, and openLCA.

Using trustworthy emissions baseline and market data as a foundation, Koi's computational engine instantly constructs a climate solution technology model that includes anticipated life cycle intervention points and potential market uptake.

Use of AI

The Koi Engine rapidly scaffolds new forecasting models using AI on top of a foundation of deeply curated data. Our Koi Data Lake is continuously maintained and structured to capture the most relevant global, regional, industry, and company-specific data. The result: rapid models based on reliable data, consistent logic, and real-world context.

Koi uses AI to accelerate the modeling process. We will never compromise on transparency, though, so have gone to great lengths to ensure full tracability of data sources and quality.

Every model generated by the Koi Engine is tagged with multiple Quality Assurance (QA) levels ranging from AI-Only (AI0) to Human Verified (FULLVAL). These QA tags make it clear how much each component of the model has been reviewed or refined by experts, ensuring users understand what they’re seeing and how best to apply it. Whether you’re screening early-stage startups or performing detailed diligence on an investment opportunity, you know exactly how much weight to place on each output.

Model QA Label	Description
AI0	AI generated data for 1+ baseline parameters. No model quality assurance.
BASELINEVAL	Baseline GHG intensity and market size are human verified. Solution data are AI generated (in-part or full). Classified as essential quality assurance that bounds the model within a plausible range.
FULLVAL	All data are human verified. Classified as enhanced quality assurance.

Integrating Automation into Commissioned Forecasting Services

Koi's modeling approach is designed to accommodate a wide range of analytical needs and climate strategies. All Koi license holders have access to the Data Lake and previously constructed models via the Search Engine. Users can customize models in Koi Studio to incorporate their specific data and modify parameters as desired.

Koi lowers the barriers to obtaining customized, forward-looking emissions models. The status quo entails analysts maintaining bespoke spreadsheets to track reliable data sources (spanning everything from emissions intensities to market sizes). These spreadsheets are most often manually maintained, and act as data libraries for constructing analyses on a case-by-case basis - often lacking technology specificity and requiring weeks of manual research by specialists. This legacy approach is not only time- and cost-intensive but also prone to errors, difficult to maintain, and ineffective for collaboration. Nevertheless, expert manual review offers substantial benefits and is why we offer a hybrid approach with our expert forecasting services.

Forecasting Services

Forecasting Services are offered in multiple tiers for maximum flexibility across use cases. New models typically begin with auto-generating a draft based upon relevant data from the Koi Data Lake. Model drafts are then quality assured by a trained scientist via manual review, research and refinement.

Our fastest quality assurance option, Rapid Models, use the engine to autogenerate a model draft and then have a scientist perform essential quality assurance (see above criteria for BASELINEVAL). Models that have been human-reviewed meet this essential quality assurance standard. We also offer services for higher-tier models with additional manual refinements that add significant nuance (i.e., purchasing a Refined Model includes FULLVAL).

Use Case Considerations

Each use case requires specific considerations of technology nuance and detail.

Some Koi users model the potential impact of early-stage ventures, where significant uncertainty surrounds both the future product and target market. These users often find that rapid models are well-suited to their needs given the inherent uncertainty, the wide range of technologies under consideration, and budget constraints.

Conversely, other users require a deep understanding of potential investment impacts and prefer our higher-tier models, which integrate our scientists' technical expertise through stakeholder engagement and discovery, while still leveraging the efficiencies of automated modeling.

Starting with the auto-generated model, our scientists maximize time efficiency by focusing on the details that matter—whether it's regional specificity, incorporating insights from white papers or technical reports, or modeling unique lifecycle phases for the solution.

This approach allows for the generation of best-in-class models in a fraction of the traditional time. Further, anchoring models to the foundational Data Lake not only saves precious time but also ensures consistency and data intercompatibility across analyses both now and in the future as the underlying data are updated.

Next steps

• For more information about each model tier, check out the forecasting services pricing table and the informational popups for each feature.
• Allowing users to perform their own quality assurance in-platform via a guided process is on our Koi Roadmap. If this is an important feature to you, please consider providing feedback within the roadmap.