Are finance teams implementing AI the wrong way?

In episode #359, Ben Murray argues that many CFOs and finance leaders are approaching AI backward—focusing too much on prompts and quick wins rather than building the foundational data infrastructure required for meaningful, repeatable insights.

Drawing from recent AI webinars and his experience building softwaremetrics.ai, Ben explains why SaaS metrics, retention, and cohort analysis should not rely on AI. Instead, these should be computed through structured, deterministic systems first—then enhanced with AI for deeper analysis and pattern recognition.

Resources Mentioned

• My new metrics engine: https://softwaremetrics.ai/
• My SaaSpocalypse post: https://www.thesaascfo.com/the-saaspocalypse-ai-agents-vibe-coding-and-the-changing-economics-of-saas/

What You’ll Learn

• Why prompt-driven AI workflows are not scalable in finance
• The difference between deterministic systems and AI-driven analysis
• Why you don’t need AI to calculate core SaaS metrics like retention or CAC payback
• The importance of structured data and clean data pipelines
• How AI should be layered on top of computed financial data—not raw inputs
• Why context windows and token usage matter when working with large datasets
• How AI can uncover insights (like expansion opportunities) that FP&A teams may miss

Why It Matters

• Prompt-based workflows create inconsistency and lack of auditability
• Without structured data, AI outputs are unreliable and not repeatable
• Finance teams risk “prompt fatigue” without building scalable systems
• Deterministic calculations ensure accuracy for critical SaaS metrics and reporting
• AI delivers the most value when used for analysis—not basic computation
• Efficient data handling reduces token costs and improves performance