{"ok":true,"engine":"aicost.msp-ai-roi","inputs":{"tasksPerMonth":1200,"minutesPerTask":8,"eligiblePct":70,"successPct":80,"reviewMinPerTask":1.5,"reworkPct":12,"loadedHourlyUsd":68,"apiMonthlyUsd":220,"platformMonthlyUsd":300,"governanceMonthlyUsd":150,"integrationOneTimeUsd":6000,"modelSlug":"","inputTokensPerTask":0,"outputTokensPerTask":0},"result":{"apiMonthlyUsd":220,"apiBasis":"manual input (estimated*)","netHoursPerMonth":57.3,"netUsdPerMonth":3229.39,"technicianEquivalents":0.41,"paybackMonths":1.9,"scenarios":[{"name":"Conservative","successPct":60,"hrs":34.6,"usd":1683.34},{"name":"Expected","successPct":80,"hrs":57.3,"usd":3229.39},{"name":"Aggressive","successPct":90,"hrs":68.7,"usd":4002.42}],"decision":"DEPLOY WITH REVIEW GATES","decisionDetail":"Clears costs even at conservative settings. Deploy with human review gates and track actual vs modeled monthly.","laborRepriceNote":"Deflection is repricing seat labor across the industry (2026: $90/seat labor dropping to $60–$70 with AI-assisted triage*) — feed the released hours back into the Minimum Profitable Price calculator.","verdict":"POSITIVE ROI","verdictReason":"Net $3229/mo at expected settings and still $1683/mo at conservative — the automation clears its own costs.","memo":"Gross minutes saved are haircut by review (1.5 min/task), rework (12%), and failed-run cleanup. API basis: manual input (estimated*). Released hours are capacity, not cash, unless a hire is avoided or the time is billed.","questions":["Is the success rate measured on your tickets or quoted from a vendor deck?","Who reviews automated output, and is their time in the model?","What is the rollback plan when the model degrades? Governance cost covers monitoring — keep it non-zero."],"assumptions":["Failed-run cleanup assumed at 30% of task time for non-completed automations.","Technician-equivalent uses 140 productive hours/month.","Manual API figure is estimated*; supply a model + tokens for SSOT pricing."]}}