ROI Framework for AI Automation Projects

ROI for AI automation should be boring and measurable. If you cannot name the hours, revenue, or risk you will change, you are buying a demo. ## Formula Annual value ≈ (hours saved × fully loaded hourly cost) + (conversion lift × average deal value × volume) + (risk avoided) − (build + run cost) ## Inputs to gather before build - Current cycle time for the process - Error / rework rate - Volume per week - Who touches the process today - Cost of delay (especially speed-to-lead) ## Example (illustrative) If SDRs spend 8 hours/week copy-pasting leads, at $40/hour, that is ~$16,640/year before you count missed follow-ups. A two-week pilot that cuts that in half often pays back inside a quarter. ## What not to count as ROI - "We feel more innovative" - Token spend without successful task completion - Deflection that creates repeat contacts ## Governance Revisit metrics 30 and 90 days after launch. Kill or redesign workflows that do not move the numbers. ## Next step We use this framework in discovery so quotes map to outcomes, not vibes.