Introduction: The Architectural Mismatch
In the rapidly evolving landscape of enterprise software, the promise of Generative AI (GenAI) has often outpaced its practical application. While Large Language Models (LLMs) have proven adept at content creation, summarization, and basic coding tasks, a critical debate is emerging regarding their utility in high-stakes financial and enterprise planning.
Brian McCormac, Chief Revenue Officer at Fintastic, recently sat down with CB Insights to articulate a nuanced perspective that challenges the prevailing "AI-can-do-it-all" narrative. His core assertion—that asking a language model to compute over a full enterprise plan is an attempt to force a square peg into a round hole—highlights a fundamental architectural mismatch between probabilistic models and deterministic financial requirements.
This article explores the technical limitations of current AI models in enterprise environments, the strategic positioning of Fintastic, and the future of specialized versus generalized AI in the corporate sector.
The Core Problem: Probabilistic vs. Deterministic Computing
To understand McCormac’s skepticism, one must first distinguish between how LLMs function and how enterprise planning operates.
The Nature of LLMs
LLMs are, at their heart, probabilistic engines. They predict the next most likely token in a sequence based on vast datasets. They are designed for linguistic fluency, pattern recognition, and creative generation. They are not built for mathematical precision, relational integrity, or the rigid logic required for multi-dimensional financial modeling.
The Enterprise Planning Requirement
Enterprise planning, by contrast, requires deterministic outcomes. When a CFO reviews a revenue projection, a headcount budget, or a tax liability forecast, there is zero room for "hallucinations" or probabilistic variance. The logic must be traceable, auditable, and mathematically sound.
As McCormac notes, "Asking a language model to compute over a full enterprise plan is asking it to do something it was never really architected or intended to do." This architectural limitation is the primary hurdle for companies looking to integrate GenAI directly into their core financial engines.
Chronology: The Evolution of AI in Finance
The integration of AI into finance has been a multi-decade journey, currently punctuated by the GenAI boom.
- 2000–2010: The Spreadsheet Era. The dominance of Excel and rigid ERP systems. Data was manual, siloed, and error-prone.
- 2010–2018: The Rise of Cloud Planning. Companies like Anaplan and Adaptive Insights introduced cloud-based, collaborative planning. Logic became more centralized, but still required heavy manual configuration.
- 2018–2022: Predictive Analytics. Machine Learning (ML) models began to appear, offering regression-based forecasting. These were deterministic but often black-boxed, leaving finance teams unable to explain why a model reached a specific conclusion.
- 2023–Present: The Generative AI Gold Rush. The emergence of GPT-4 and other LLMs led to a frenzy of implementation. Many companies attempted to plug LLMs into their financial data lakes, leading to the "architectural mismatch" that McCormac highlights.
The current phase of the industry is one of recalibration. Organizations are realizing that GenAI is a powerful interface for data, but a poor processor for complex financial logic.
Supporting Data: The Efficiency Gap
Industry benchmarks from the past 18 months suggest a growing "Implementation Gap." According to recent research:
- Hallucination Rates in Financial Logic: Studies indicate that LLMs attempting complex multi-step arithmetic without external "chain-of-thought" frameworks or calculator tools exhibit error rates as high as 15–20% in financial reasoning tasks.
- Productivity Paradigms: While GenAI reduces the time spent on drafting emails and summaries by approximately 40%, its impact on the accuracy of financial planning is negligible or, in some cases, detrimental due to the need for human validation.
- Enterprise Trust: Surveys of CFOs indicate that fewer than 30% are currently willing to automate budget allocation decisions via AI without "human-in-the-loop" verification, citing auditability as the primary concern.
These figures underscore why Fintastic and similar innovators are focusing on "Hybrid Models"—systems where AI handles the interpretation and natural language queries, while deterministic, purpose-built engines handle the actual computations.
Fintastic’s Strategic Positioning
Fintastic, under the leadership of CRO Brian McCormac, has positioned itself as a specialized provider that respects the boundaries of AI. Rather than trying to be an "AI-everything" platform, Fintastic focuses on the utility of the platform for the end-user.
Market Definition
Fintastic operates in the intersection of Enterprise Performance Management (EPM) and Business Intelligence (BI). They define their market not by the technology they use (AI/ML), but by the problems they solve: agility in forecasting, real-time variance analysis, and audit-ready reporting.
Product Philosophy
McCormac’s approach is one of "Guided Intelligence." Instead of asking an LLM to "calculate our Q4 budget," Fintastic uses LLMs to translate a user’s natural language request into a query that hits a deterministic, rules-based engine. This preserves the accuracy of the calculation while providing the ease of use that modern enterprise software users expect.
Official Responses and Industry Sentiment
The perspective shared by McCormac is echoed by many in the CTO and CFO community. As enterprises move past the initial hype cycle, the sentiment is shifting from "how can we use AI for everything" to "where does AI provide the highest leverage?"
- The Case for Specialized AI: Experts argue that we are moving toward a modular architecture. In this future, LLMs act as the "Front Office" (customer service, data discovery, reporting), while specialized, domain-specific algorithms act as the "Back Office" (core calculations, risk management, compliance).
- The Auditability Mandate: Regulators are beginning to take note of AI in finance. The ability to explain a decision—the "explainability" factor—is becoming a regulatory requirement. Because LLMs are inherently opaque, relying on them for core financial decisions is increasingly viewed as a liability.
Implications: The Future of Enterprise Software
What does this mean for the future of business operations?
1. The Decline of "Black Box" Automation
The era of blind reliance on AI models is coming to an end. Businesses will prioritize "Glass Box" systems—platforms that show their work, allow for manual overrides, and provide clear audit trails for every calculation.
2. The Rise of Domain-Specific LLMs
We will see a transition from general-purpose models (like GPT-4) to specialized models trained on specific financial datasets and accounting standards. These models will be "constrained" by logic layers that prevent them from venturing into incorrect mathematical territory.
3. The Human-in-the-Loop 2.0
The human’s role is shifting from "data entry" to "logic auditor." Instead of spending time inputting numbers, employees will spend time reviewing the logic used by the AI to arrive at those numbers. This is a higher-value task that requires deep domain expertise.
4. Integration Challenges
Companies will continue to struggle with data silos. AI is only as good as the data it has access to. If an enterprise has fragmented, messy, or inconsistent financial data, even the most advanced AI will fail. The primary investment for most companies in the next 3–5 years will be data infrastructure, not just AI software.
Conclusion: Balancing Innovation with Integrity
Brian McCormac’s warning is a timely reminder for the enterprise software sector. While the allure of GenAI is undeniable, the fundamental requirements of business—accuracy, accountability, and logic—remain unchanged.
The companies that will win in the next decade are not those that chase the latest AI trend at the expense of performance. They are those that understand the strengths and, more importantly, the limitations of the tools they deploy. By treating LLMs as powerful linguistic assistants rather than omniscient financial brains, Fintastic and its peers are helping enterprises build a future that is both smarter and more reliable.
As McCormac implies, true innovation isn’t just about what you can do with technology; it’s about what you should do to ensure the longevity and integrity of the business. For now, the most effective enterprise strategy is to keep the math in the engine and the language in the interface.
