1. Scope
Compares monthly support cost between human-only and AI-first (with human escalation) models. Includes token spend per AI ticket and escalation overhead.
2. Inputs and outputs
Inputs
- ticketsPerMonth number
- avgHumanMinutesPerTicket number
- humanHourlyCost number ($)
Loaded rate (wage + benefits + tools + QA).
- aiResolutionRate number (0–1)
- tokensPerAiResolved number
- aiInputPricePerMtok number ($)
- aiOutputPricePerMtok number ($)
- escalationOverheadMinutes number
Outputs
- monthlySavings
Human-only minus AI-first.
- costPerAiTicket
AI-first total / tickets.
- savingsPercent
Savings / human-only baseline.
Engine source: src/lib/ai-vs-human-support-cost/engine.ts
3. Formula / scoring logic
human_only = N × (handle / 60) × hourly
ai_token_$ = tokens × ((in + out) / 2) / 1_000_000
ai_first = N × ai_token_$ + N × (1 − rate) × ((handle + escalation) / 60 × hourly)4. Assumptions
- 50/50 input/output token split.
- Escalated tickets pay full human handle time plus escalation overhead.
- Loaded human cost is 1.4–1.8× wage; user enters loaded rate directly.
5. Data sources
- BLS — customer service representatives wages as of 2024-05
- Zendesk CX Trends Report as of 2024
6. Known limitations
- Doesn't include QA cost on AI responses, tooling fees (Zendesk / Intercom), or training time.
- Token figures ignore retries, tool calls, and chained agent runs that can 2–10× real spend.
7. Reproducibility
Input
4000 tix, 12 min, $35/hr, 65% rate, 4000 tok, $0.50/$1.50, 3 min escalation.
Expected output
humanOnly $28k, savings >$15k.
8. Change log
- 2026-05-08 methodology first published.
Worked example
Run live against the same engine this site ships
(/engines/ai-vs-human-support-cost.js).
The inputs and outputs below are recomputed on every build and
independently re-verified in CI — they are never hand-authored.
Input
- tool
- ai_vs_human_support_cost
- tickets_per_month
- 4000
- avg_human_minutes_per_ticket
- 12
- human_hourly_cost
- 35
- ai_resolution_rate
- 65
- tokens_per_ai_resolved
- 4000
- ai_input_price_per_mtok
- 0.5
- ai_output_price_per_mtok
- 1.5
- escalation_overhead_minutes
- 3
Output
- humanOnlyMonthlyCost
- 28000
- costPerHumanTicket
- 7
- aiFirstMonthlyCost
- 12266
- costPerAiTicket
- 3.07
- monthlySavings
- 15734
- savingsPercent
- 56.19
- aiCheaperPerTicket
- true
- aiResolvedCount
- 2600
- escalatedCount
- 1400
Frequently asked questions
- What does the AI vs Human Support Cost calculate?
- Compares monthly support cost between human-only and AI-first (with human escalation) models. Includes token spend per AI ticket and escalation overhead.
- What inputs does the AI vs Human Support Cost need?
- It takes 8 inputs: ticketsPerMonth, avgHumanMinutesPerTicket, humanHourlyCost, aiResolutionRate, tokensPerAiResolved, aiInputPricePerMtok, aiOutputPricePerMtok, escalationOverheadMinutes. Outputs returned: monthlySavings, costPerAiTicket, savingsPercent.
- What formula does the AI vs Human Support Cost use?
- The exact computation is: human_only = N × (handle / 60) × hourly; ai_token_$ = tokens × ((in + out) / 2) / 1_000_000; ai_first = N × ai_token_$ + N × (1 − rate) × ((handle + escalation) / 60 × hourly)
- Can I verify the AI vs Human Support Cost with a worked example?
- Yes. With 4000 tix, 12 min, $35/hr, 65% rate, 4000 tok, $0.50/$1.50, 3 min escalation. the tool returns humanOnly $28k, savings >$15k.
- Where does the AI vs Human Support Cost get its benchmark data?
- Reference data is sourced from: BLS — customer service representatives wages (as of 2024-05); Zendesk CX Trends Report (as of 2024).
- What can the AI vs Human Support Cost not tell me?
- Known limitations: Doesn't include QA cost on AI responses, tooling fees (Zendesk / Intercom), or training time. Token figures ignore retries, tool calls, and chained agent runs that can 2–10× real spend.