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

CFOs We are Implementing AI Backwards

SaaS Metrics School

CFOs We are Implementing AI Backwards