AI Automation ROI Calculator

Before you spend anything on AI, find out whether the task you have in mind is worth automating at all. Put in how many people do the job, how long it takes them and what an hour of their time costs, and this works out the annual saving, the payback period and where you stand after three years — including the running costs most estimates quietly leave out.


Will automating this actually pay for itself?

Put your own numbers in. Nothing is sent anywhere and there's no email gate — the figures update as you move the sliders, including when the answer is no.

Salary plus on-costs — roughly annual package ÷ 1,750.

The rest is exception handling and review. Presets start conservative.

Net saving per year

Payback period
Net after 3 years
Time returned
Equivalent to

Indicative only. Build costs vary widely with how messy the source data is, and running costs are estimated at 18% of build per year for licences, usage and support.

What counts as a good payback?

Under 12 months

The shape of a project worth doing. A first build that pays for itself inside a year is the one that earns you the internal credibility to do the next one.

12 to 36 months

Pays back, but slowly. Fine for a second or third automation once you have proven the approach — a poor choice for your first, where you need a visible win.

Over 36 months

Hard to justify. Systems change, suppliers change formats and priorities move. A payback beyond three years is usually a signal to pick a different task.


Use — Guide

How to use this tool

  1. Pick the task you are thinking about automating. The preset sets a realistic starting point for how much of it automation can actually remove, and an indicative build cost — both deliberately conservative.
  2. Set how many people do this task and how many hours each of them spends on it in a normal week. Use real numbers, not best-case ones; if you are not sure, ask the people doing it.
  3. Set the loaded hourly cost — salary plus superannuation, leave, payroll tax and overheads. A rough shortcut is annual package divided by 1,750.
  4. Adjust the "realistically automated" slider. Nothing goes to 100%: there is always exception handling, review and the cases the system has to escalate to a person.
  5. Read the payback period and the three-year net. A first automation that pays back inside a year is the kind that funds the next one; anything past about three years is usually the wrong first project.
  6. If the numbers do not work, that is a useful result — the task may simply be too small to justify a build. Try a task with more hours in it before spending anything.

FAQ — Questions

Frequently asked questions

01How accurate is this calculator?

It is an indicative model, not a quote. The two inputs that move the result most — how much of the task can genuinely be automated, and what the build actually costs — both depend on specifics this tool cannot see, above all how clean and consistent your source data is. What the calculator is good for is sorting candidates: it will reliably tell you whether a task is in the right order of magnitude to be worth automating, and it will tell you clearly when it is not.

02Why does it assume 46 working weeks instead of 52?

Because nobody works 52 weeks a year. Four weeks of annual leave plus public holidays and typical sick leave puts most Australian employees close to 46 productive weeks. Using 52 would inflate every annual saving by around 13% — a small dishonesty that compounds into a badly wrong business case.

03What is the 18% annual running cost?

Automation is not a one-off purchase. Once it is live you pay for licences and API or inference usage, monitoring, and the ongoing work of handling changes when a supplier alters a document format or a system gets updated. Eighteen per cent of build cost per year is a reasonable planning figure for a maintained workflow. A model that ignores running costs will show a payback that never arrives in reality.

04Why is the time saving discounted?

The calculator applies a realisation factor of 80% to the theoretical saving, because recovered time is rarely converted into productive work one-for-one. This matters most when a saving is spread thinly — twenty people each getting fifteen minutes back is far harder to convert into real capacity than one person getting a full day back. If your saving is concentrated on a few people, the true figure will sit at the optimistic end.

05The result says it does not pay back. What now?

Take it seriously — it is usually correct, and finding out here costs nothing. The most common reasons are that too few hours go into the task to justify a build, or that the build is being scoped too ambitiously for a first project. Look for a task with more total hours across more people, or a simpler approach: for many businesses a properly configured and trained Microsoft 365 Copilot rollout covers a lot of ground at a fraction of a custom build.

06Does Peritus Digital build these automations?

Yes. We work with businesses across Newcastle, Lake Macquarie, the Central Coast and Port Stephens on document processing, workflow automation, Microsoft 365 Copilot rollouts and AI-assisted analytics — and we secure what we build, which matters because the account an automation runs under is exactly the kind of access attackers look for. If the numbers here look promising, we will test the approach against your real documents before you commit to anything.