LegalOn Technologies Cuts Codex Costs Without Compromising Development Speed

LegalOn Technologies has taken a strategic step to reduce costs in its Codex integration used for global legal and business processes. By implementing model selection based on task complexity and tiered budget limits, the company reduced costs by approximately 65 percent without compromising development speed.

Cost Control and Model Optimization

LegalOn Technologies faced significant challenges in cost control while integrating AI tools to support its global business processes.

To optimize costs without slowing down development speed, the company adopted a task complexity-based selection strategy among the GPT-6 Astra, GPT-6 Luna, and GPT-6.1 Sol models.

AI Center of Excellence Guidance

The company's in-house AI Development CoE developed comprehensive guidelines to select the right model for each task rather than using the highest-capacity model.

Under the newly implemented controls, GPT-6 Luna handles open-requirement code implementations, while GPT-6.1 Sol supports standard design and data analysis.

Budget Limits and Departmental Allocation

Alongside model selection, the Fast mode was restricted by default, and strict budget limits were introduced at the department, group, and individual levels.

Mature business lines were asked to increase cost efficiency, while broader budgets were allocated to newly established business lines to encourage active usage.

Future Goals and Investment Metrics

Moving forward, LegalOn is developing investment metrics focused on feature releases and customer value rather than relying solely on development speed and usage volume.

The company also plans to create a knowledge base to share practical experiences and place greater emphasis on AI skills in the hiring processes.