EY launched a context-integrated agent workflow that linked factory agents to its internal engineering standards, code repositories and compliance frameworks, featuring semi-autonomous agents designed to produce deployable code. The initiative was led by EY’s product development and engineering teams and followed an 18–24 month buildup of cultural adoption and technical integrations.
The rollout evaluated multiple agent platforms, including Factory, Replit and Lovable, and elevated Factory after pilot metrics showed strong developer adoption. EY implemented security guardrails, repo access controls and a workload classification framework that routed high-autonomy tasks—like code review and defect fixes—to agents while preserving human oversight for architecture and large refactors.
For developers, the shift reframed roles from sole coders to orchestrators directing agents, driving measured productivity gains across teams and enabling faster, standards-compliant delivery of audit, tax and financial platforms. The approach shows how embedding AI in a company’s "context universe" helps scale useful, deployable code generation.
Image Credit: Krot_Studio / Shutterstock
What's Driving This Trend
- Context-aware Coding Agents
- Embedding agents with organizational standards and repo context enables generation of deployable, compliance-aligned code at scale that can reshape internal engineering workflows.
- Developer-orchestrator Roles
- A shift from hands-on coding to agent orchestration creates new role definitions centered on agent supervision, policy configuration and high-level architectural judgment.
- Guardrails-driven Autonomous Coding
- Combining workload classification with security guardrails and access controls allows semi-autonomous agents to handle routine fixes and reviews while preserving human oversight for risky tasks.
Who This Affects Most
- Professional Services and Accounting
- Integrating context-aware agents into audit, tax and financial platforms promises faster delivery of standards-compliant solutions and could change how advisory firms scale technical service offerings.
- Enterprise Software Development
- Enterprise product teams stand to benefit from agent-assisted code generation that accelerates feature delivery while enforcing internal architecture and compliance constraints.
- Devops and CI/CD Tooling
- Tooling that routes tasks by autonomy level and enforces repo-level policies can enable continuous, secure code production pipelines augmented by semi-autonomous agents.
