Economics / 2 min read

A better business case than “hours saved.”

Model useful capacity, review effort, operating costs, and payback without confusing an estimate with a guaranteed return.

Tern & Loom editorial · 20 October 2026

Define the eligible work

Start with monthly volume and the share that the proposed workflow can safely handle. Do not model the whole department if the pilot only handles routine invoices or one support category. Multiply eligible items by the current handling time to estimate the baseline effort.

Subtract the work that remains

Automation usually changes human work rather than removing all of it. Staff review output, resolve exceptions, maintain knowledge, and investigate failures. Estimate the minutes that remain per item and include ongoing administration separately.

Hours released = eligible monthly items × (current minutes − remaining minutes) ÷ 60. If the remaining work exceeds today’s effort, the model should show a negative result. That is useful information.

Price the whole operating system

Include model/API usage, OCR, workflow hosting, integration subscriptions, and support. Use a realistic loaded hourly cost to value capacity. Do not omit exception handling simply because the happy-path demonstration looks fast.

Net modeled capacity value = hours released × loaded hourly cost − monthly operating costs. This is an economic comparison, not a promise that payroll will fall by that amount.

Use payback carefully

Implementation cost divided by positive monthly net value gives a simple modeled payback period. When net value is zero or negative, there is no positive payback under those assumptions. Show that honestly instead of forcing a flattering number.

This model excludes financing, taxes, opportunity cost, and changes in demand. It is appropriate for an early operational discussion, not a formal investment valuation. Run conservative, expected, and optimistic scenarios before committing.

Measure quality alongside efficiency

A faster process that creates more incorrect orders or reopened tickets can be a worse process. Track a quality measure and an exception measure alongside capacity: duplicate rate, critical field errors, rework, reopen rate, or unresolved queue age.

Agree the baseline and success criteria with the business owner before the build. Then replace assumptions with observed pilot data and make the next investment decision from that evidence.

Ready to make this concrete? Explore the capacity calculator or prepare a workflow brief.

Further reading

This article offers Tern & Loom’s practical framework. The primary references below provide supporting technical and market context.

Microsoft: organizational readiness for AI ↗

McKinsey: The state of AI ↗

Keep exploring

More useful starting points.

A better way to work

Make room for
your next chapter.

Tell us where the work gets stuck. We’ll help define a worthwhile first step.

Plan your first automation