KPMG ran a six-week pilot where tax professionals with no technical background built working software prototypes using vibe coding, and some of those tools are already live with clients.
When Brad Brown watched a room full of tax associates and managers start building functional software prototypes, he wasn't looking at a cohort of trained developers. These were certified public accountants and tax specialists, people whose daily reality involved parsing regulatory code, not writing programming code. Yet within six weeks, they had produced tools that KPMG US now deploys in client engagements.
The pilot, which ran from January through February, brought together roughly 30 tax professionals with minimal technical backgrounds and paired them with engineers in small teams. Their task was straightforward in concept but radical in practice: use vibe coding to build software that solved real problems they encountered in their daily work. Vibe coding, a term that has gained traction across the technology sector over the past year, refers to the practice of generating functional code through natural language prompts rather than manually writing syntax. Tools like Cursor, Windsurf, and Bolt have made it possible for non-engineers to describe what they want in plain English and receive working prototypes in return.
What makes KPMG's experiment significant is not the technology itself but the organizational shift it represents. Professional services firms have long operated on a model where domain experts identify problems and technical teams build solutions. That handoff creates friction. Tax professionals would describe what they needed, engineers would interpret those requirements, and the resulting product often required multiple iterations before it matched the original vision. By putting the prototyping capability directly in the hands of the people who understand the problem most intimately, KPMG has compressed that cycle dramatically.
The tools that emerged from the pilot are not hobby projects or internal experiments. According to KPMG, some of what these teams built has already been incorporated into software that clients use to manage tax workflows and data. The groups of four to six workers automated compliance processes, built integrations across multiple data systems, and created analytics tools that previously required significant manual effort. A manager from KPMG's tax technology team served as a coach for each group, stepping in when participants encountered technical obstacles, but the core ideation and initial build came from the tax professionals themselves.
Brown, who serves as chief digital officer for KPMG's tax practice, described the dynamic as reaching toward something akin to the concept of a tenfold engineer that permeates Silicon Valley discussions, only applied to consulting. The tax professionals are not replacing software engineers. Instead, they are creating what Brown called lightweight, workable tools that engineering teams then refine with proper safeguards, security infrastructure, and scalability. The engineers spend less time interpreting requirements and more time strengthening functional prototypes.
This division of labor matters for startups and established enterprises alike. The bottleneck in most organizations building internal tools is not engineering capacity alone. It is the translation cost between domain expertise and technical execution. Every round of clarification, every misunderstood requirement, every prototype that misses the mark represents wasted time and money. Vibe coding collapses that gap by letting the person with the domain knowledge take the first pass at the solution.
A Broader Movement Takes Shape
KPMG is not alone in exploring this approach. Checkr, the identity verification startup valued at over $4.6 billion, has similarly encouraged employees without technical backgrounds to experiment with building tools using AI-assisted coding. The trend reflects a broader shift in how enterprises think about software development as AI tools lower the barrier to entry for non-technical workers. GitHub Copilot, Replit Agent, and Anthropic's Claude have all contributed to an ecosystem where the ability to describe a problem clearly can be as valuable as the ability to write a function.
For KPMG, the next step is scaling what worked in the pilot. The firm is evaluating the creation of roughly a dozen permanent teams that would blend tax professionals with software engineers, formally structuring the hybrid model that proved effective during the initial experiment. If those teams perform as expected, the implications extend well beyond one firm's internal operations. Professional services is an industry built on billable hours and specialized expertise. When your domain experts can also prototype their own solutions, the economics of service delivery start to look fundamentally different.
The real question for competitors and startups watching this space is not whether vibe coding works. It is how quickly organizations can retrain their workforce to think like builders while maintaining the discipline to know when to hand off to professional engineers. KPMG's pilot suggests the learning curve is shorter than most would have predicted six months ago. The companies that figure out the organizational design fastest will have a meaningful edge in an industry where speed to solution increasingly determines who wins the engagement.
