Fixing vs preventing with AI: fridge repair math shows $120K vs $200K a year

louisdorard · x · 2026-10-25

A worked example of quantifying AI value before building: for a fridge repair company where 1 in 5 jobs needs a second visit —

Fix faster: an AI agent tells the second engineer the likely fault and parts to bring, skipping the depot trip; 10% saved per visit ≈ $120K/year.
Prevent: an ML model trained on past jobs flags likely second visits so a senior engineer goes first; preventing half of them ≈ $200K/year.

Takeaway: run the numbers first — fixing and preventing have very different economics.

More from coding & agent