Software Second

AEM Published economic model

Automation economics,
without invented savings.

A transparent method for following workflow time through capacity, explicit economic capture, technology cost, quality, cash flow, and uncertainty.

Model version: AEM-1.0.0Published August 10, 2026Use the calculator

01 Governing distinction

Time saved → capacity created → value captured → economic result.

Hours removed from a task do not become cash merely because they have a loaded labor rate. If payroll, overtime, contractor spend, planned hiring, throughput contribution, or another explicit economic outcome does not change, the model reports resource-equivalent capacity—not direct cash savings.

02 Variables

The workflow and the proposed system.

Rates are decimals from 0 to 1. Money is entered in the user’s working currency; the public calculator presents USD.

VAnnual workflow units
m₀Current human minutes per unit
e₀Current human error / rework rate
rw₀Human rework minutes per error
F₀Fixed annual human hours
wUser-supplied loaded hourly cost
aAutomation coverage: share of units attempted
sStraight-through success rate
rReview rate on successful outputs
mrMinutes per human review
mfExtra failure / exception overhead minutes
e₁Residual automated-output error rate
rw₁Residual rework minutes per error
F₁Fixed annual oversight hours

03 Human effort

Count what disappears.
Keep what remains.

Baseline time per unitT₀ = m₀ + e₀ × rw₀

Normal handling time plus expected human rework time.

Baseline human hoursH₀ = V × T₀ ÷ 60 + F₀

The current unit-driven and fixed annual labor requirement.

Proposed human hoursH₁ = V ÷ 60 × [(1−a)T₀ + a(1−s)(T₀+mf) + as(rmr+e₁rw₁)] + F₁

The three terms count units never automated, failed attempts returning to the full human workflow with extra overhead, and successful outputs that still require review or later residual rework.

Capacity changeHcap = H₀ − H₁

Negative capacity is preserved. It means the proposed workflow creates more human work.

Labor-equivalent capacity valueHcap × w

A resource-equivalent metric. It is never added automatically to cash benefits.

04 Capture ledger

Claim each recovered hour once.

Allocated capacity cannot exceed positive capacity created. The server rejects an overallocated ledger.

Reduced overtime

hours avoided × marginal overtime cost

Reduced contractor spend

hours avoided × contractor cost

Headcount cost removed

hours actually removed × loaded cost

Future hire avoided

planned capacity avoided × planned loaded cost

Throughput reallocation

hours reallocated × incremental contribution per hour

Gross revenue is converted to contribution using a user-supplied margin.

Unused capacity

max(Hcap, 0) − allocated hours

Visible operational capacity with no claimed direct financial effect.

05 Quality and technology

Separate error consequences from rework labor.

Current non-labor quality costQ₀ = V × e₀ × c₀
Proposed non-labor quality costQ₁ = V × [(1−a)e₀c₀ + a(1−s)(e₀c₀+cf) + ase₁c₁]

Includes existing-process errors on untouched and failed-fallback units, automation-induced failure cost, and residual automated-output error cost.

Quality valueQ₀ − Q₁

May be negative.

Annual variable technologyV × a × variable cost per attempted unit
New recurring technologyfixed + variable + other recurring

06 Two economic views

Resource efficiency and cash impact answer different questions.

Resource view

How much operational resource does each workflow consume?

Baseline resource cost − proposed resource cost

Includes labor-equivalent resource use, expected non-labor quality cost, and relevant recurring technology cost. The result is labeled resource-equivalent economic change.

Cash / capture view

What financial effect has an explicit mechanism?

captured capacity + quality value + retired recurring − new recurring

The labor-equivalent capacity value is excluded. This is the steady-state annual net cash impact used by cash flow, NPV, and payback.

07 Timing and investment

Model the month the economics actually begin.

  1. 01

    Linear ramp

    0 before go-live; then min(1, (m−delay) ÷ max(ramp,1))

    When ramp is zero, full performance begins immediately after go-live.

  2. 02

    Monthly cash flow

    benefits × ramp − variable cost × ramp − fixed operating cost

    Fixed operating cost begins at go-live. Implementation is a month-zero outflow.

  3. 03

    NPV

    −I + Σ CFₘ ÷ (1 + ((1+d)^(1/12)−1))ᵐ

    The annual effective required return is converted to an effective monthly rate.

  4. 04

    Simple ROI

    (total benefits − total incremental costs) ÷ total incremental costs

    Undiscounted and explicitly labeled. It is not available when incremental cost is zero.

  5. 05

    Payback

    first month cumulative cash flow ≥ 0

    Simple and discounted payback are calculated separately. No crossing means no payback within the horizon.

  6. 06

    IRR

    solve Σ CFₘ ÷ (1+r)ᵐ = 0

    A bracketed numerical solver returns an annualized IRR only when one economically meaningful root exists.

08 Break-even and sensitivity

Ask what must be true—and what drives the answer.

Break-even

A sampled range and bisection solve selected-horizon NPV = 0 for automation coverage, straight-through success, human-review rate, recurring technology cost, and workflow volume. The model reports when economics stay positive or negative throughout the range instead of manufacturing a threshold.

One-way sensitivity

Selected-horizon NPV is recalculated at a mechanical low and high assumption, normally −10% and +10% within legal bounds. Results are ranked by absolute NPV swing. These are sensitivity tests, not probability forecasts.

Scenario simulation

Optional low, most-likely, and high inputs define triangular distributions. A canonical hash seeds 5,000 server-side scenarios so identical assumptions reproduce identical P10, P50, P90, positive-NPV, and payback results.

09 Interpretation example

“The workflow creates 2,000 hours of capacity. You identified mechanisms for capturing 600 hours. The remaining 1,400 hours are operational capacity—not booked cash savings.”

The calculator then shows review and failure burden, recurring technology cost, quality change, NPV, payback, break-even thresholds, and the assumptions with the greatest NPV effect. Every major output includes the formula and substitution used.

10 Limitations

Transparent arithmetic does not make assumptions true.

  1. 01

    Results are conditional on user-supplied assumptions. The model does not verify workflow volume, labor time, performance, costs, contribution, or implementation feasibility.

  2. 02

    The linear ramp is intentionally simple. It does not model adoption S-curves, seasonality, correlated implementation delays, or capacity constraints outside the supplied assumptions.

  3. 03

    Sensitivity changes one variable at a time. It is not a confidence interval and does not reveal interactions among assumptions.

  4. 04

    Scenario simulation uses triangular distributions selected by the user. Its probabilities are conditional scenario outputs—not forecasts or empirical likelihood estimates.

  5. 05

    Non-labor error economics exclude human rework time because rework labor is already included in workflow hours.

  6. 06

    Avoided hires are counterfactual. They create value only from the month the hire would otherwise have occurred and require evidence that the hire was genuinely planned or necessary.

Use AEM-1.0.0 to expose the assumptions. Validate them before treating the result as a decision.

Open the calculator