Anyone quoting a number without seeing your workflow is guessing. But you can get a long way with the cost structure, because it's the same on every project.
Someone has to map the workflow, wire the AI into your systems, test it against real work, and make it safe to run. A typical mid-complexity SMB build runs well under the cost of one month of the salary it's saving.
Our Builds: from $4,000 + GST, 6 months of support included
AI models charge per use, like electricity. A document-heavy automation that runs every day usually costs dollars to tens of dollars a month in model usage. The AI is rarely the expensive part; support and monitoring is the meaningful line.
Support retainer: from $200/month after the included period
Per-seat AI licences (Microsoft 365 Copilot at roughly A$45 per user per month, Claude Team around A$40 per seat) are great value for daily users and dead money for everyone else. Twenty unused Copilot seats is about $10,800 a year.
We audit tenants and find this regularly
We rebuilt a construction PM firm's monthly Project Control Group reporting on Claude and Azure, inside their own Microsoft 365 tenant. A roughly 3-hour report task now takes about 30 minutes. At full rollout across ~120 reports a month, that's projected to free up around 3,600 hours of senior PM time a year – roughly $360K of capacity. Read the case study.
The build cost was a fraction of the first year's benefit. That ratio isn't unusual when the workflow is chosen properly; it's the choosing that most businesses get wrong.
Take one repetitive task. Multiply: how many times it happens per month, how long it takes, and the hourly cost of the person doing it. That's your monthly leak.
A 2-hour task done 40 times a month by someone costing $75/hour leaks $6,000 a month. Against a build in the $10–20K range with a few hundred dollars a month to run, payback is measured in weeks. Our ROI Calculator does this in 30 seconds, no email required.
If the leak is under about $1,000 a month, an AI build probably isn't worth it yet. A cheaper fix (a template, a Power Automate flow, deleting the task) usually is. We tell clients this when it's true.
A few things reliably move a build from the bottom of the range to the top: messy or scattered source data, systems without decent APIs, workflows where the rules live in someone's head, and anything needing compliance review before output goes out the door. None of these are dealbreakers. They're just work, and a fixed-price quote should name them up front.
That's why we scope before we quote. Our AI Readiness Sprint (from $4,950 + GST, two weeks) maps the workflows, checks the data, and hands you a board-ready one-pager with real numbers – whether or not you build with us.
Be wary of open-ended day rates for AI "discovery", per-seat licences rolled out before use cases exist, and any quote where you can't see the deliverable. You should see the price and the deliverable before you commit. We publish ours.
If the maths doesn't stack up for your scale, we'll say so – "don't build yet" is a real outcome here too.