LegalOn cuts Codex costs by 65% through strategic model selection

LegalOn Technologies achieved a 65% reduction in estimated daily AI costs while maintaining development speed by strategically selecting among GPT-6 Astra, GPT-6.1 Sol, and GPT-6 Luna based on task complexity. Previously relying on unlimited access to GPT-5.5 in Fast mode, the company’s AI-powered Development CoE (AID CoE) introduced guidelines that shifted teams from using the highest-capability model for every task to choosing the right model for the job.

Model Tiering Strategy

The new governance framework categorizes models by capability to optimize resource allocation:

• GPT-6 Luna: Handles everyday execution, simple automations, and code implementation with clear requirements.
• GPT-6.1 Sol: Supports standard design, routine numerical analysis, and document creation.
• GPT-6 Astra: Reserved for complex analysis, architecture design, and agent orchestration.

Alongside model selection, LegalOn restricted Fast mode by default, allowing individual requests only when necessary. Teams maintained performance by running tasks in parallel, ensuring that cost controls did not hinder productivity.

Budget Governance and ROI

The company implemented budget caps at the department, group, and individual levels. Mature business areas were tasked with improving cost efficiency by approximately 20%, while new businesses in the launch phase received generous budgets to encourage active AI adoption. This dual approach allowed LegalOn to direct resources toward growth areas while controlling overall spending.

Looking ahead, LegalOn is developing a new metric to measure the true ROI of AI investment by linking customer value delivered by feature releases to the actual AI costs invested in them. The company also plans to build a knowledge base to share best practices for model combinations across the organization, turning individual engineer expertise into a company-wide capability.