Why We Invested in Fieldguide: Building the Agentic AI OS for Audit and Advisory

The audit and advisory industry is experiencing a structural collision between escalating demand and declining talent base. CPA graduates have declined 27% over the last decade, while the Department of Labor projects 4% annual labor growth needed through 2032 just to offset retirement exodus. Meanwhile, regulatory burden intensifies—from heightened SEC enforcement to mandatory ESG assurance requirements—creating an unsustainable capacity constraint. This isn’t a cyclical challenge; it’s a secular transformation demanding fundamental re-platforming of how professional services operate. 

That’s why Geodesic Capital is investing in Fieldguide, the company building the agentic AI operating system for audit and advisory.

The Fragmentation Tax on Professional Services

Audit and advisory firms operate on a billable hours model that has become structurally constrained. The global industry generates $230 billion annually, yet commits only 3-7% of revenues to software spend—the remainder attributable to human labor, which is collapsing under its own weight.

The technology infrastructure hasn’t meaningfully evolved in decades. 52% of accounting firms rely on three or more disconnected systems, often cobbling together spreadsheets, email, static portals, and legacy practice management tools—creating version control chaos, duplicative work, and zero real-time visibility. The cost of siloed systems is staggering: 70% of employees spend over 20 hours weekly navigating disconnected systems, translating to $2.5 million in annual productivity losses for mid-to-large organizations, making it ripe for disruption.

System of Record Meets System of Action

While most breakout vertical AI companies possess a couple of control points at best:—proprietary compounding data, system-of-record/action positioning, domain-tuned models solving superhuman tasks, robust integrations, closed-loop systems, and end-to-end workflow automation—Fieldguide has assembled all six, which becomes a scaling advantage.

The company functions as both the repository for audit engagement data and the execution layer where work actually happens. This dual positioning creates network effects that are exceptionally difficult to replicate. By serving as the end-to-end platform across planning, document intake, testing, reporting, and close, Fieldguide becomes embedded in practitioners’ daily workflows. Users average 7 hours of active engagement per day across 5 days per week—a level of stickiness that signals true product-market fit.

The Data Moat: 20,000 Audits Across 80 Types

Fieldguide has assembled the deepest corpus in the industry: structured engagement data, workpapers, test procedures, and audit outcomes spanning 20,000 audits across 80 engagement types. This proprietary dataset becomes the foundation for domain-tuned models that deliver measurably better-than-human results.

The technical architecture is instructive. Fieldguide’s Testing Agent achieves 98% accuracy versus 54% for human auditors and 90% for GPT-4, while operating 94% faster. This isn’t incremental improvement—it’s a step-function change. The platform employs multi-model optimization, routing tasks to the best-performing LLM based on benchmarked performance, combined with advanced techniques including chain-of-thought reasoning, consensus mechanisms, and self-reflection.

As the company processes more engagements, this data advantage compounds, creating a widening moat that strengthens with every audit executed on the platform.

How It Works: From Co-Pilot to Autonomous Execution

We are moving from the state of “CoPilot” to “CoWork”. 

In that context, Fieldguide’s agentic capabilities move beyond passive chatbots (“CoPilot”) to active agents executing complex, multi-step workflows with minimal human intervention (“CoWork”). The platform’s orchestration layer manages multiple AI sub-agents that collaborate to plan and execute traditionally manual workflows, compressing weeks of work into days with dramatically higher quality.

Today, Fieldguide operates in “Guided Automation” (Level 3)—where AI agents execute full engagement workflows with structured checkpoints and human oversight. Practitioners no longer perform the bulk of tasks; instead, they guide and supervise the process, intervening when judgment is required. On average, Fieldguide customers unlock approximately 2x capacity—staff accomplish twice the work in the same timeframe.

Our belief is that the company will get to “Strategic Automation” (Level 5), where AI agents plan and execute full engagements while practitioners operate at the governance layer.

Go-to-Market Velocity with Capital Efficiency

Fieldguide has penetrated 40 of the Top 100 CPA firms—including 7 of the Top 10. The company grew close to 200% year-over-year. Critically, it is important for efficient scaling in Vertical AI, and Fieldguide has demonstrated it.

Why Fieldguide Wins

Our investment conviction rests on four pillars.

First, category-defining position: Fieldguide possesses all six key characteristics outlined in the Vertical AI thesis —proprietary compounding data, system-of-record/action positioning that compounds with the 80+ robust integrations, combining that data moat with domain-tuned models to deliver superhuman performance, closed-loop systems, and end-to-end workflow automation.

Second, founders with vertical domain expertise: Jin Chang (CEO) spent five years in assurance and advisory at EY, before joining Atrium, one of the earliest legal tech companies. His co-founder, Chris Szymansky (CTO) was VP of Engineering at Atrium. Critically, 25% of Fieldguide’s FTEs are former auditors, ensuring the product is built by practitioners for practitioners.

Third, competitive moat defensibility: Legacy vendors are bolting on AI but lack production-grade capabilities and remain burdened by technical debt. Point solutions are not end-to-end platforms and cannot achieve system-of-record positioning. Newer entrants lack the data corpus, distribution, and live deployment scale that Fieldguide has built over the last four years.

Fourth, a massive labor augmentation/substitution opportunity: The market opportunity is substantial; we assess that Fieldguide addresses a $13B TAM in 2025 representing junior staff labor spend among Top 5-500 CPA firms in the US alone, which can be augmented by agentic AI, in the context of an increasing structural talent gap. 

Audit and advisory services will never sunset—they are fundamental functions for organizational transparency, internal controls, and regulatory compliance. As regulatory burden increases and labor supply contracts structurally, we at Geodesic firmly believe Fieldguide’s agentic AI platform is positioned to capture the future of how professional services operate. 

Read more about Fieldguide’s funding news in Fortune: Goldman Sachs leads $75 million funding round for Fieldguide, an AI-native accounting and audit platform