Measured manual and AI review workload

Worked example

A team evaluating AI-assisted drafting can count drafting, human review and rework over the same workload. Compare errors using the audited denominator for each process.

For 100 fictional records, manual review takes 2 staff-hours and AI-assisted review takes 1.5. The audited error rates are 1% and 3%, respectively. A half-hour saved comes with a higher error rate in this example.

Try it yourself

The calculator opens with the fictional example already filled in. Change one input, choose Calculate locally, and compare the result to the original values below. Use Reset fictional fixture to start over.

An unknown result means the available inputs do not support a value. Check the calculator’s Warnings tab for what is missing or uncertain.

Measured manual labor 2 staff-hours
Measured AI draft, review and correction labor 1.5 staff-hours
Manual minus AI-assisted labor 0.5 staff-hours
Manual audited error rate 1 % of audited items
AI-assisted audited error rate 3 % of audited items
Manual labor cost at supplied wage Undefined for these inputs

Go to the calculator Browse all calculators

Calculator

Approach (label only)Changing the label does not improve capacity, speed, error rates or confidence. Enter measured differences explicitly.

Comparison conditions and human review

Equal-resource comparison conditionsRecord budget, staffing, resources, information release dates, scenario conditions and evaluation criteria. Material differences can make options not comparable.

Evaluation period

Geography / exercise boundary

Human reviewers and roles

This evaluation depends on a real transaction

Supplied human authority gateThis declaration is not legal certification. Blocked, pending or assumed gates stop transaction-dependent evaluation.

Keep units consistent within each table. Enter zero when the amount is known to be zero. Leave optional amounts blank when they are unknown. If a required value is missing, the calculator will ask for it.

Advanced: complete input envelope and optional fields

Edit optional fields using the method dictionary. Apply validates and recalculates the entire input. Unknown values remain null.

Versioned input JSON

Import local JSON or CSV

Files stay in this browser. Maximum 1 MiB. JSON imports are recalculated. CSV supports flat observation tables, with column types taken from the displayed fixture. Empty cells become null.

Workbench JSON / input envelope

CSV target table

CSV file, matching the target table headers

Compare the arrangements

Keep physical resources and evaluation conditions equal. With identical inputs, all three approaches produce identical arithmetic. The comparison refuses material differences in declared conditions; it does not verify those declarations.

Show only differing metric rows

Privacy: comparison inputs may contain sensitive notes or operational details. Storage is off by default. Do not enter personal records or sensitive infrastructure details. Reproduction exports include supplied inputs and notes, including raw small-cell inputs suppressed from CORE-009 result tables. Suppression of displayed results does not make a reproduction export safe to publish.

Save comparison inputs on this device until I clear them

Executive brief and reproducible export

Public exercise notes

Review for personal, confidential or sensitive material before saving or sharing. No automatic submission occurs.

Read the generated brief

Method, inputs and limits

Compare measured manual work with AI drafting, human review and correction for the same explicitly identified task. Method version 0.2.0. Infer quality or AI uplift from mode labels, speed, assumed productivity or unmeasured downstream outcomes.

CORE-011 input dictionary
Input	Unit / type	Required interpretation
tasks — Paired measured tasks	work-items / json	Each task: id, scope, items, evidenceId, manualMinutes, aiDraftMinutes, reviewMinutes, correctionMinutes, manualErrors {errors,checked}, aiErrors {errors,checked}. Optional audit sampleId labels identify the same items audited in each process. Null error counts remain unknown.
wagePerHour — Optional common measured wage	currency/staff-hour / number	Parameters.wagePerHour and explicit parameters.currencyUnit, including base year, are required together for labor costing. Costs are rounded to two currency decimals. Minimum 0.
currencyUnit — Wage currency and base year	currency unit / text	Use the supplied schema and declared units.
reviewQueue — Optional measured review queue	work-items / json	Observations.reviewQueue: itemsWaiting, minutesPerItem, reviewers, availableMinutesPerReviewer, periodLabel. Whole tasks are allocated per reviewer; no new arrivals or inferred productivity.

Visible mathematics / decision rules

AI-assisted minutes = draft minutes + human review minutes + correction minutes

staff-hour difference = (manual minutes − AI-assisted minutes) / 60

verified error rate = audited errors / audited items, only when audited items > 0

optional labor cost = measured minutes / 60 × supplied common wage per staff-hour; two currency decimal places

optional queue reviewable items = min(items waiting, reviewers × floor(available minutes per reviewer / measured minutes per item)); pooled clearance periods = total work minutes / total reviewer minutes, a lower bound only

What this result cannot establish

Timing and errors must be measured on the same workload and eligible labor scope.
A positive time saving does not establish better output quality; an AI process may be slower or have more audited errors.

What would falsify the assumptions?

Repeat paired measurement using independently audited work and identical task definitions.
A process with longer total time or more errors demonstrates added cost or risk.
A non-AI process can produce identical results with the same measured resources.
Assumptions fail if the supplied definitions, units, period, scope or source records are wrong or incomplete.
An independent source audit or replication may contradict an input or the recorded dependency.
This method creates no resources. Physical conservation is checked in the inventory and transfer ledgers; descriptive graphs and checklists are not resource-flow models.
The same equations and review checklist can be reproduced with a spreadsheet or structured human-only process.
This is deterministic arithmetic, a descriptive comparison or a structured governance check, not a fitted causal effect or forecast. Fictional inputs remain an exercise.
Equal or worse measured errors, net review effort, disclosure burden or fallback under MAR would show no added benefit or added risk; no improvement is assumed.

Reproducible fictional fixture and independently calculated expectations

This fixture is synthetic. It does not describe a country, government, company or event.

{
  "observations": {
    "unit": "work-items",
    "origin": "synthetic",
    "sources": [
      {
        "id": "SYN-001",
        "kind": "synthetic",
        "label": "Fictional paired workload timing",
        "url": null
      }
    ],
    "tasks": [
      {
        "id": "T1",
        "scope": "Same 100 fictional records",
        "items": 100,
        "evidenceId": "SYN-001",
        "manualMinutes": 120,
        "aiDraftMinutes": 30,
        "reviewMinutes": 40,
        "correctionMinutes": 20,
        "manualErrors": {
          "errors": 1,
          "checked": 100
        },
        "aiErrors": {
          "errors": 3,
          "checked": 100
        }
      }
    ]
  },
  "parameters": {}
}

Expected checks

{
  "manualStaffHours": 2,
  "aiStaffHours": 1.5,
  "savedStaffHours": 0.5,
  "manualErrorRatePercent": 1,
  "aiErrorRatePercent": 3
}

Download input fixture

Authoritative background references

NIST: AI Risk Management Framework — Voluntary governance background; not validation of MAR. Reviewed 2026-11-09. Reference, not endorsement or validation.

These sources inform definitions or review practices. They do not validate this implementation or endorse MAR.

Open the original Scenario Studio Examine original hypotheses Find related policy modules