AI-paired delivery has compressed build cost 3-5× on well-defined work — the threshold has shifted from "build only above $500K/yr SaaS" to "consider build above $200K/yr." Fourteen probing questions on tool category, scale, switching cost, strategic importance, lock-in tolerance. Output: 3-yr TCO for both paths, payback period, risk-adjusted Build / Buy / Hybrid verdict for the board.
SaaS TCO uses your stated annual cost compounded over 3 years with the chosen growth rate, plus seat-creep adjustment based on active-seat ratio. Build TCO uses an AI-paired delivery model (3-5× faster than 2023 baselines for well-defined work), with complexity / regulated-overlay multipliers, plus 3-year run cost (hosting + 0.5 FTE platform). Risk weighting accounts for switching cost, strategic importance, and lock-in tolerance.
The threshold has dropped from $500K/yr SaaS to ~$200K/yr because AI-paired delivery compressed build cost 3-5× on well-defined work. Below $200K it's still buy. Above $500K it's almost always build. The middle band depends on strategic factors.
Under 12 months: build is usually right. 12-24 months: depends on strategic factors. Over 24 months: buy.
Switching cost (data migration, retraining), opportunity cost of engineering attention, and tool-specific integrations the build needs to replicate. The memo addresses each.