What would AI actually save you? In dollars.
Not a maturity score — a figure from your own processes, deliberately understated so it holds up. Plus an honest check on whether AI is already in use in your business without anyone governing it.
What would AI actually be worth to you?
Not a maturity score. A dollar figure, from your own processes — plus an honest check on whether AI is already being used in your business without anyone governing it.
The opportunity
List your manual jobs. Get hours and dollars per year, and a ranked list of what’s worth automating first.
The risk
Whether staff are already pasting client data into personal AI accounts — and a policy template to fix it.
We deliberately understate the savings. Every other calculator assumes automation removes 100% of a task. It doesn’t. We use the low end of realistic ranges and show the working, so the number holds up when someone challenges it.
Not everything is an AI problem
Each task gets matched to the technology that actually fits, with an honest range for how much time it removes. Two of these five need no AI at all — and they're usually the best value.
Rules-based workflow
Power Automate or similar. Deterministic, testable, and no AI involved — which makes it the most reliable and usually the cheapest thing to build.
System integration
Connect the systems so data stops being re-typed. Highest reduction of anything here because the manual step disappears entirely.
Document AI / extraction
AI reads the document and populates your system. A human still checks exceptions, so this never gets to zero.
Generative AI with human review
AI drafts, a person edits and approves. Genuinely useful, but editing takes real time — anyone promising you 90% here is selling.
Classification and triage
AI reads, categorises and routes — enquiries, emails, tickets. Removes the sorting, not the responding.
Percentages are the proportion of task time typically removed. We use the lower bound for your headline figure. There’s a sixth category — not worth automating— because sometimes that’s the honest answer.
How this works
The maths, the assumptions, and why our numbers are smaller than everyone else's.
How is the dollar figure calculated?
For each task: minutes per occurrence × how often it happens per year × how many people do it, giving annual hours. That's multiplied by an indicative hourly cost to the business — the midpoint of the band you select, which includes on-costs like superannuation, leave and overhead, not just wages. The saving then applies a realistic reduction percentage based on what kind of automation fits that task.
Why are your savings lower than other AI calculators?
Because most of them assume automation eliminates 100% of a task, which isn't true. Extracted invoice data still needs exceptions checked. AI-drafted quotes still need editing. We use the low end of realistic reduction ranges for the headline figure and show the upper end separately. A number that survives scrutiny in a meeting is worth more than a big one that doesn't.
What is shadow AI?
Staff using AI tools that the business hasn't approved or configured — typically personal ChatGPT, Claude or Gemini accounts. It matters because those accounts sit outside your tenancy: you can't see what's been shared, can't revoke access when someone leaves, and consumer terms may permit the provider to use inputs for model improvement. In most businesses it's already happening.
Is putting client information into AI a privacy problem?
It can be. Entering personal information into a third-party AI service is a disclosure of that information under the Privacy Act 1988, and it needs to be consistent with what you told people when you collected it. Enterprise services like Microsoft 365 Copilot and Azure OpenAI operate under commercial data protection terms and don't use your prompts to train foundation models — consumer tiers may differ.
Do I need AI, or just automation?
Often just automation. A large share of the savings this tool identifies comes from rules-based workflows and system integration — no AI involved, cheaper to build, and far more reliable because the behaviour is deterministic. The results screen tells you which of your tasks genuinely need AI and which don't. We'd rather sell you the boring thing that works.
What's in the policy template?
An editable Word document covering approved tools, information that must never be entered, Privacy Act considerations, human review requirements for customer-facing output, incident reporting and review ownership. Sections adapt to your answers. It's a starting template, not legal advice — have it reviewed before issuing it to staff.
The estimate is the easy part
An automation audit maps the processes properly, confirms the numbers against what’s actually in your systems, and gives you a costed build plan for the highest-return few. You keep the ranked list whether or not you engage us.