Pigment’s AI Model Builder: Capabilities, Use Cases & Limitations

Overview

The AI Model Builder is a conversational interface within Pigment that allows users to create and manage planning models using natural language prompts. It connects directly to Pigment’s underlying data model and supports both new builds and the expansion or auditing of existing applications. The AI layer is powered by a large language model and integrates with Pigment’s block and formula architecture.

Generative AI Capabilities

For new builds, the AI can create dimensions, metrics, tables, transaction lists, and folders, while also generating boards and views automatically. It leverages a pre-defined use case library to accelerate model scaffolding, offers build suggestions, and can add sample data, inputs, and list items on demand. External files can also be attached to help inform model structure.

For existing models, the AI supports expansion and audit capabilities, including writing, debugging, and optimizing Pigment formulas. It can restructure applications to align with best-practice architecture, update metadata and naming conventions, adjust levels of detail while retaining inputs, and create validation flags and missing-input alerts.

External AI Integration

Pigment’s AI Model Builder connects to external large language models, allowing users to describe intent in plain English while the system translates that into Pigment operations. It infers block types, formula logic, and structural changes, and provides proactive suggestions throughout the build process.

Pre-Defined Use Cases

The platform includes a growing library of pre-defined use cases, which can be used to quickly generate initial model structures. Our testing across fund forecasting, cash deployment, and operating expense planning showed that these templates provide a meaningful head start, though they benefit from further refinement through prompting.

Capability Testing

Across testing scenarios, the AI demonstrated a strong ability to interpret both conversational inputs and structured requirement lists. It supported tasks such as building models from natural language, adjusting levels of detail, generating sample data, identifying unused metrics, and creating validation flags. The ability to generate sample datasets and input structures significantly accelerated model readiness and reduced what would typically be multi-day build efforts.

Limitations

Several limitations remain. The AI cannot configure data imports or connections, manage security settings, or build iterative calculations. It has limited awareness of downstream dependencies, cannot interpret non-Pigment formulas such as Excel logic, and currently does not offer undo functionality for AI-driven changes.

Overall Assessment

Pigment’s AI Model Builder represents a meaningful step forward in reducing the manual effort required for financial model construction. It accelerates model development, supports governance through audit and restructuring capabilities, and introduces more intuitive ways to interact with planning models. While some constraints require careful oversight, this is a capability worth exploring for teams already using the platform.