KPMG US Tax Pros Build Client Software via 'Vibe Coding' Pilot

A six-week internal pilot at KPMG US has turned tax professionals into software builders, training them to produce production-grade client tools using AI-assisted "vibe coding" techniques rather than conventional programming. The initiative, which bypassed KPMG's traditional software engineering talent pool in favor of domain experts who know tax compliance from the inside, has since been woven directly into the firm's client service offerings.

Vibe coding, a method that lets practitioners generate and refine working code through natural language prompts rather than hand-written syntax, has been creeping into accounting and professional services for months. KPMG itself noted in January 2026 that developers are now using approaches such as "vibe coding" and "agentic coding" that allow them to generate and refine code using natural language, "fundamentally reshaping the software development process and required skillsets." The internal pilot took that observation a step further: instead of retraining engineers to understand tax, KPMG trained tax professionals to build software.

The distinction matters on the partner track. Tax associates and managers have spent years learning the substance underneath the code, including transfer pricing logic, CAMT calculations, and global minimum tax mechanics. Turning that expertise into deployable software, without a six-month handoff to an engineering team, compresses the cycle between insight and client deliverable in ways that could reshape how engagements are staffed and scoped.

KPMG's broader push in this space has been visible. The firm's Tax AI Accelerator Program, announced in February 2026, combines technical training with practical tax application and is led by instructors who are both technologists and experienced tax professionals. Fortune 150 company Duke Energy is among more than a dozen organizations participating in the program. The vibe coding pilot represents the internal counterpart: KPMG professionals themselves acquiring the same build-versus-describe shift the firm is now selling to clients.

Each participant organization in KPMG's AI tax infrastructure receives a custom deployment of KPMG's Digital Gateway platform, built on Microsoft Azure OpenAI, providing a secure private sandbox where teams can experiment with AI tools and test use cases against real compliance challenges without exposing sensitive data. That sandboxed architecture also underpins what KPMG's own tax teams used to develop and validate the tools now embedded in client engagements.

For KPMG professionals watching how AI restructures headcount and project economics across the Big Four, the pilot signals something concrete: the line between tax technician and software developer is collapsing faster than most busy-season schedules would suggest, and the firms moving quickest are not waiting for engineers to learn tax.