LegalOn Technologies, a legal technology startup, says it reduced estimated daily AI costs by 65% while maintaining development speed. The company achieved the reduction by matching GPT-6 Astra, GPT-6 Luna, and GPT-6.1 Sol to tasks, restricting Fast mode by default, and setting budgets according to business needs. The results compare costs with GPT-5.5; mature business areas were asked to improve cost efficiency by up to approximately 20%.

LegalOn initially gave developers unlimited access to GPT-5.5 in Fast mode for design, implementation, and everyday work. As adoption expanded, its AI-powered Development CoE (AID CoE) tested models and shared selection guidance, allowing engineers to choose models based on task complexity and development stage. The company also added administrator controls with monthly usage limits for departments and individuals, which AID CoE monitors and adjusts as needs change.

Under the guidelines, GPT-6 Luna handles code implementation with clear requirements and everyday automations; GPT-6.1 Sol supports standard design, data analysis, and document preparation; and GPT-6 Astra is used for complex analysis and architecture design. Teams can move from lighter to more capable models as tasks become more complex. LegalOn also limited Fast mode by default, while teams used parallel task execution to maintain performance. Established businesses were asked to pursue up to approximately 20% greater cost efficiency, while new businesses in launch phases received generous budgets to encourage AI use.

Senior Engineering Manager Yuta Tokitake said new businesses prioritized business speed over cost efficiency, aiming to use AI extensively to increase output and drive growth. LegalOn reports that model selection, Fast mode restrictions, and tailored budgets together cut estimated daily costs by approximately 65%, while maintaining development speed. Teams also report that AI is shortening the development-to-release cycle, though the company says it has not yet established whether faster development delivers more customer value.

To measure that value, LegalOn is developing a metric that treats each feature release as a unit, linking the customer value it delivers with the AI costs invested in producing it. The company is building a pipeline to calculate the metric, which Tokitake said is intended to make the return on investment visible and accurate. LegalOn also plans a knowledge base for sharing effective model combinations across design, implementation, and review, and has begun emphasizing AI skills in hiring. Its AID CoE and security team are working together on a flexible governance framework balancing risk control with teams' ability to experiment.

LegalOn Technologies, a legal technology startup, says it reduced estimated daily AI costs by 65% while maintaining development speed. The company achieved the reduction by matching GPT-6 Astra, GPT-6 Luna, and GPT-6.1 Sol to tasks, restricting Fast mode by default, and setting budgets according to business needs. The results compare costs with GPT-5.5; mature business areas were asked to improve cost efficiency by up to approximately 20%.

LegalOn initially gave developers unlimited access to GPT-5.5 in Fast mode for design, implementation, and everyday work. As adoption expanded, its AI-powered Development CoE (AID CoE) tested models and shared selection guidance, allowing engineers to choose models based on task complexity and development stage. The company also added administrator controls with monthly usage limits for departments and individuals, which AID CoE monitors and adjusts as needs change.

Under the guidelines, GPT-6 Luna handles code implementation with clear requirements and everyday automations; GPT-6.1 Sol supports standard design, data analysis, and document preparation; and GPT-6 Astra is used for complex analysis and architecture design. Teams can move from lighter to more capable models as tasks become more complex. LegalOn also limited Fast mode by default, while teams used parallel task execution to maintain performance. Established businesses were asked to pursue up to approximately 20% greater cost efficiency, while new businesses in launch phases received generous budgets to encourage AI use.

Senior Engineering Manager Yuta Tokitake said new businesses prioritized business speed over cost efficiency, aiming to use AI extensively to increase output and drive growth. LegalOn reports that model selection, Fast mode restrictions, and tailored budgets together cut estimated daily costs by approximately 65%, while maintaining development speed. Teams also report that AI is shortening the development-to-release cycle, though the company says it has not yet established whether faster development delivers more customer value.

To measure that value, LegalOn is developing a metric that treats each feature release as a unit, linking the customer value it delivers with the AI costs invested in producing it. The company is building a pipeline to calculate the metric, which Tokitake said is intended to make the return on investment visible and accurate. LegalOn also plans a knowledge base for sharing effective model combinations across design, implementation, and review, and has begun emphasizing AI skills in hiring. Its AID CoE and security team are working together on a flexible governance framework balancing risk control with teams' ability to experiment.