How I Decide Whether an Automation Is Worth Building

Measure the current process

Begin with executions per month, active minutes per execution and loaded hourly cost. Monthly manual cost is executions × minutes ÷ 60 × loaded cost. Keep waiting time separate: a two-day cycle may contain only twenty minutes of labour, but delay may still carry commercial value.

Measure a representative sample instead of relying on the smoothest SOP. Include research, copying, checking, rework, exception chasing and supervisory review.

Estimate coverage, not “automation percentage”

A workflow rarely removes every human minute. Estimate the share of executions eligible for the automated path and the review minutes each eligible run still needs. Then account for exceptions that remain fully manual.

For 600 monthly executions at 12 minutes each, the baseline is 120 hours. If 75% are eligible and require two minutes of review, those runs fall from 90 hours to 15 hours. The other 30 hours remain. Capacity released is 75 hours, not 108.

Price ongoing ownership

Add provider usage, hosting, monitoring, credential rotation, API changes, model regression tests and expected modification work. A fragile $20/month Zap that consumes ten hours of emergency support is not cheap.

Separate predictable monthly maintenance from major process change. A workflow owner should review exceptions and operational metrics even when Bahman provides managed maintenance.

Include error and delay economics carefully

Some workflows create value through fewer errors, faster response or recovered throughput. Count these only when they can be observed. Duplicate refunds have a direct cost; “better decisions” needs a measurable proxy and a baseline.

Also price failure. If a wrong automated action is expensive or irreversible, coverage may need to stay low and review high. The safest economically positive design may automate preparation rather than execution.

Calculate simple payback

Monthly value equals released capacity at loaded cost, plus evidenced error/delay savings, minus monthly operating cost. Simple payback is build cost divided by monthly value. It is a screening tool, not a financial forecast.

Example: 110 hours released at $55/hour yields $6,050 of monthly capacity. If operation costs $550, net monthly value is $5,500. An $18,000 build has a simple payback of roughly 3.3 months. This does not guarantee the organisation converts every released hour into cash.

Run sensitivity, then check readiness

Use low, base and high assumptions for volume, coverage, review time and cost. If the project only works under the optimistic case, it is probably not ready. The interactive calculator on this site exposes these assumptions rather than hiding them.

Finally check ownership, API access, process stability, data quality and willingness to handle exceptions. A high-value unstable process may need redesign before automation.

• Build now: frequent, stable, measurable process with accessible systems and a clear owner.

• Audit first: attractive value but uncertain volume, policy or integration surface.

• Do not automate yet: rare process, rapidly changing rules, poor data or failure cost beyond available controls.

Capacity is not automatically cash

Released hours may become faster service, more throughput, less overtime or reduced hiring pressure. Say which one the buyer expects. Do not label all time savings as profit.

The honest business case is usually strong enough: make assumptions visible, measure after launch, and expand only when observed economics support it.