Compute your workload's LDI

The same three-pillar math used for the federal benchmarks, applied to a workload you describe. AI cost is pulled live from the Compute CPI basket; everything else is yours to supply.

Inputs
BLS OEWS mean wage for the role. Loaded 1.3× for benefits (matches ECEC).
Wall-clock minutes for a trained human to handle one task.
Tickets, interviews, claims — whatever your unit is.
Pulls live cost/unit from the Compute CPI basket.
Share of volume plausibly already routed to AI. Default is the current composite LDI.
Your workload's LDI
Human cost / unit
—
—
AI cost / unit
—
—
Cost ratio
—
human vs AI per unit
Structural cost gap (100% sub)
—
(human − AI) × annual volume
Wage income displaced
—
at your observed sub rate
FTE equivalent of task hours
—
2,080 hr/yr · full-substitution ceiling

What the math is

Human cost
hourly = wage / 2080
loaded = hourly × 1.3
cost_per_unit = loaded × minutes / 60
AI cost
cost_per_unit = basket[task_type]
.cost_per_1k / 1000
Live from compute-cpi.json.
Wage income displaced
(sub_rate / 100) ×
annual_volume ×
human_cost_per_unit

This is not a savings estimate. The structural cost gap is the upper bound if substitution were total and AI cost scaled linearly — it reflects the pricing differential, not a forecast. The realized figure (wage income displaced) depends on your observed substitution rate. The federal composite LDI is — at last run.

Data access

All nine federal benchmark workloads are published one-JSON-per-workload at /data/ldi/workloads/ with an index at /data/ldi/workloads/index.json. Flat CSV: /data/ldi/workloads.csv.

Client-side math only — nothing you enter leaves your browser.