Results on customer projects

1 day

Onboarding a new developer

• ProblemA new hire used to spend 2–3 weeks learning the project and its environment, then another 2–3 weeks at half speed.
• What we didThe developer and their agent get the team's decisions and rules while working on each file.
• ResultNew developers pick up tasks on day one.
• How we measuredCompared with the last few new hires before AI Council.

30–45%

One task, clean test

• ProblemThe same task on the same project.
• What we didThe same agent completed the task twice: once with AI Council and once without.
• Result30–45% saved in time and in tokens.
• How we measuredCompletion time and token usage across the two runs.

40–60%

A group of tasks: a team and several agents

• ProblemA series of tasks worked on in parallel by several developers and agents.
• What we didThe team worked with a shared AI Council knowledge base and rules.
• Result40–60% saved in time and in tokens. Knowledge gathered in some tasks helps with the next ones.
• How we measuredTime and token usage on a series of tasks with and without AI Council.

4x faster

A company building ERP and data warehouse systems, two teams

• ProblemTasks spanning two connected systems: integrations, bug investigations, writing specs.
• What we didAI Council takes both projects and the links between them into account.
• ResultCross-system tasks moved 4 times faster.

Fewer interruptions

An industrial software company (machine tools)

• ProblemEngineers kept getting interrupted with the same questions about configuration.
• What we didThe project chat answers these questions from the project's knowledge base.
• ResultDevelopers get pulled out of their tasks less often, and answers to repeated questions are saved.
• How we measuredThe team's own assessment.