LegalOn Technologies AI Cost Optimization Case Study

LegalOn Technologies reduced its estimated daily AI costs by approximately 65% by transitioning from a single high-performance model to a tiered selection strategy based on task complexity. This optimization allowed the company to maintain development speed while aligning AI spending with the growth stages of different business units.

Tiered Model Selection Strategy

LegalOn Technologies shifted from using GPT-5.5 (Fast mode) for all tasks to a structured selection process managed by the AI-powered Development Center of Excellence (AID CoE). The company now matches specific models to the complexity of the task to optimize cost and performance:

  • GPT-6 Luna: Used for everyday execution, simple tasks, analysis, and code implementation (primarily as a subagent).
  • GPT-6.1 Sol: Used for standard software design, routine numerical analysis, document creation, and tasks requiring faster completion times than Luna.
  • GPT-6 Astra: The most capable model, reserved for complex analysis, architecture design, orchestration, and coordinating agents.

Resource Allocation and Governance

To maintain budget control without hindering productivity, LegalOn implemented a combination of system controls and strategic budget caps:

Budgetary Controls

  • Departmental Limits: Monthly usage limits were introduced for individuals and departments, monitored and adjusted by the AID CoE.
  • Differentiated Budgeting: Mature business units were required to improve cost efficiency by approximately 20%, while new businesses in the launch phase received generous budgets to prioritize speed and output over immediate cost efficiency.

Performance Optimization

To offset the restriction of "Fast mode" by default, development teams adopted parallel task execution to ensure that development speed remained constant despite the change in model access.

Measuring AI ROI via Feature Releases

LegalOn is moving beyond measuring AI success by development speed or usage volume. The company is developing a new metric to calculate the Return on Investment (ROI) based on customer value delivered per feature release.

By treating a feature release as a single unit of work, the company aims to link the actual AI costs invested in that specific release to the customer value it generates, removing the "black box" of total cost per piece of work.

Organizational Impact and Future Governance

AI adoption has led LegalOn to rethink its workforce planning and hiring, with a new emphasis on AI skills. To scale individual expertise, the company is building a knowledge base to share best practices regarding model combinations for design, implementation, and review.

Regarding the balance between control and agility, Yuta Tokitake, Senior Engineering Manager at LegalOn Technologies, stated:

"Excessive restrictions through rules and budgets can undermine an organization's momentum. What we need is a flexible operating model that effectively balances risk control with the speed teams need."

This governance framework is managed collaboratively by the AID CoE and the security team to balance performance, cost, and risk.

Sources