SaaS AI Unit Economics & COGS Skill

Overview

Generic SaaS unit economics treats AI as one cost line. AI-aware investors in 2026 require the AI cost line decomposed into its constituent waterfall, separately modelled, separately sensitised, and separately reported. This skill installs that discipline.

The output is a four-layer view of the AI cost-of-revenue:

1. Direct AI COGS — tokens, embeddings, fine-tuning amortisation, vector store, GPU reservation.
2. Indirect AI COGS — eval pipeline cost, observability / safety infra, model-router infra, caching infra.
3. AI reserves — hallucination-liability reserve, retraining-cycle reserve, model-migration reserve.
4. AI overhead allocation — share of AI-team payroll, share of AI tooling, share of governance.

Together these produce the AI Gross Margin, AI Contribution Margin per tier, blended-GM impact, AI-cost-as-%-of-ARR, and the diagnostics investors and DFIs look for.

Use When