IT Operations Automation
League needed to accelerate product development and internal operations in a regulated healthcare environment, but prompt tuning, multi-agent experimentation, and back-office work still required manual effort and strict security controls.
Reported impact
League cut product development cycle times in half, pushed Claude adoption from 80% to 98% across the company, automated prompt repair for underperforming agents, and expanded AI-driven automation into finance, design, compliance, and legal workflows without relaxing healthcare-grade controls.
League standardized on Claude Enterprise and Claude Code to make agentic work safe enough for a regulated healthcare software company. Engineers use Claude through an internal Swarm orchestration layer that can run up to 50 parallel agents against real build tasks, write code, run tests, and open pull requests, while engineers review every change before it ships. For a critical prompt-optimization workflow, League paired Fable exploration with Claude Code implementation across four codebases and two languages so a failing agent can read evaluation scores and rewrite its own prompt automatically. The rollout also extends beyond engineering: finance automates month-end, quarter-end, and payroll workflows, and teams use Claude Design for branded decks. Claude Security controls model tiers, connectors, and usage boundaries across the company.
League feeds evaluation scores, source code, and MCP-connected tool context into Claude Enterprise and Claude Code. Fable explores prompt-improvement strategies, Swarm agents implement changes across codebases, and engineers review the resulting pull requests before release; separate finance and design flows run on the same controlled platform.
All eight merged, roughly 48,000 lines across 271 files.
Our engineers do that through Swarm, our own orchestration layer built on Claude Code.
They now have more than 60 automated processes for month-end, quarter-end, and payroll
roughly 98% of League's code is AI-authored.
we can tier model access, manage which connectors are available and to whom, and set the boundaries we need.
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
Anthropic Sample Oct 13, 2026 DXC Technology DXC needed to build a production-scale agentic operations platform for managed services where specialized AI agents could coordinate complex enterprise IT workflows while meeting enterprise security, compliance,… 20% to 30% Claude Code · DXC OASIS · ITSM agent 
Salesforce Sample Sep 16, 2026 TeleSpecialists TeleSpecialists outgrew an under-configured CRM and no-automation marketing stack just as hospital, physician, and payer relationships made credentialing, onboarding, and growth increasingly exponential to manage. 10x increase in marketing campaign click-through and engagement rates Agentforce · Marketing Cloud · MuleSoft Anthropic Sample Archived · Sep 15, 2026 Carvana Carvana wanted AI to investigate production alerts inside Slack, but its homegrown Slackbots consumed engineering time and lacked the team-scoped access controls needed to expand safely across retail science,… 56% fewer alerts Claude Tag · Opus 5 