You've asked what a custom AI tool costs and been offered a discovery workshop. Here are the numbers instead: what the Australian market charges for custom AI builds, what moves the price, and our own figures, on the page, before anyone books a whiteboard.
Why AI pricing feels opaque
Partly for the usual reason: unpublished prices can flex to the client. But AI has an extra one: the field is new enough that plenty of sellers don't know what things should cost either, so the quote becomes a function of your budget and their confidence. The result is the widest quote spread we see in any category: the same chatbot brief can come back at $3,000 from a freelancer and $150,000 from a consultancy, and neither number tells you anything until you know what's inside it.
What the Australian market charges
Off-the-shelf tools: tens to a few hundred dollars a month. ChatGPT Team, Copilot, an AI feature inside software you already pay for. Not a build at all, and genuinely the right answer for a lot of jobs. If a subscription solves it, buy the subscription. The build conversation starts where these stop: your data, your workflow, your rules.
Freelancers and no-code builders: roughly $2,000 to $15,000. Automations wired together in tools like Zapier or Make, a chatbot on a platform, a GPT with your documents loaded. The good ones deliver real value fast. The constraint is what happens after delivery: these builds live or die on one person's availability, and the maintenance question usually goes unasked until something breaks.
Specialist agencies and studios: roughly $5,000 to $100,000+. Custom tools built on your data with proper engineering: testing, guardrails, a plan for who maintains it. This is where we sit. The width of the range is the scope, not the margin, and the tiers inside it are covered below.
Consultancies and enterprise builds: $150,000 into the millions. Sometimes the scale justifies it. Often what you're buying above the specialist tier is the machine around the work: the strategy phase, the steering committee, the workshop series. If you're not an enterprise, be suspicious of enterprise process at enterprise prices.
What moves the price
What the tool has to connect to. A tool that answers from a folder of documents is days of work. A tool that reads your CRM, writes to your job system, and emails your customers is a different project, because every integration adds a system to understand, an edge case to handle, and a way to fail that has to be caught.
The state of your data. AI tools answer from what you give them. If your knowledge lives in tidy documents and structured systems, the build starts immediately. If it lives in seven inboxes and a retiring employee's head, the first phase of the project is getting it out, and that phase is real work that belongs in the quote rather than discovered in month two.
The cost of being wrong. A tool that drafts internal summaries can be wrong occasionally and cost you a shrug. A tool that quotes prices to customers or touches compliance needs review steps, guardrails, and testing in proportion to the damage a bad output does. We build human approval into anything that acts on your behalf, and that engineering is part of what separates the specialist tier from the cheap one.
Who maintains it. Models update, APIs change, your business changes. A build with nobody on the hook for it degrades quietly until someone notices it's been wrong for a month. Ongoing ownership, whether your team or ours, is a line in the conversation from day one or a surprise later.
What we charge
Our AI builds run $5,000 to $100,000+. At the lower end: a focused tool doing one job well, a lookup tool answering from your own documents, a drafting assistant, an email router, built and shipped in weeks. The middle of the range covers tools with real integrations and approval workflows. The top end is a platform running a meaningful chunk of your operations. Where you land is driven by the four variables above, and we'll tell you which end you're at, and why, before you sign anything. Complex scopes can go beyond the range, and we'll say so upfront.
And the constraint that saves you the most money: if a spreadsheet solves it, we'll say so. Some of the best advice we've given on a first call was to not build anything, and the businesses that heard it came back when there was something worth building. How to pick a first AI project that survives contact with reality is covered separately, and it's the better question than the price.
When cheap AI costs more than it saves
The failure mode isn't the build breaking. It's the build being quietly wrong: a chatbot inventing a policy you don't have, an automation filing things into the wrong place for a month, a tool the team stopped trusting and silently worked around. Most abandoned AI projects die exactly this death, and the money wasn't the largest loss; the team's willingness to try the next tool was. The internal tools that stuck share a pattern: small enough to trust, useful enough to open daily, and owned by someone.
Tell us what's going on. We'll tell you what we'd build, or that you shouldn't build anything, and what it costs. If it makes sense, we start.
FAQ
How much does a custom AI tool cost in Australia?
Simple custom tools built by a specialist start around $5,000. Tools with integrations and approval workflows commonly run $15,000 to $50,000, and full platforms reach $100,000 or more. Below that sit freelancer and no-code builds from roughly $2,000, with the trade-offs in maintenance and ownership.
How much does an AI chatbot cost?
A platform chatbot with your documents loaded can be a few thousand dollars. A custom chatbot that answers from your systems, follows your rules, and hands off to a human properly typically runs $5,000 to $30,000 depending on what it connects to and the cost of it being wrong.
Should we build custom AI or use ChatGPT?
Start with the subscription. A custom build earns its cost when the job needs your data, your workflow, or your rules: answering from your documents, writing into your systems, or enforcing steps ChatGPT can't. If a subscription or a spreadsheet solves it, we'll tell you that on the first call.
What does it cost to run after launch?
Two lines: model usage, which for most business tools is tens to a few hundred dollars a month, and maintenance, which is the one to ask about, because models and APIs change and someone has to own that. We scope ongoing ownership into the build conversation rather than leaving it as a surprise.
FAQ
How much does a custom AI tool cost in Australia?
Simple custom tools built by a specialist start around $5,000. Tools with integrations and approval workflows commonly run $15,000 to $50,000, and full platforms reach $100,000 or more. Below that sit freelancer and no-code builds from roughly $2,000, with trade-offs in maintenance and ownership.
How much does an AI chatbot cost?
A platform chatbot with your documents loaded can be a few thousand dollars. A custom chatbot that answers from your systems, follows your rules, and hands off to a human properly typically runs $5,000 to $30,000 depending on what it connects to and the cost of it being wrong.
Should we build custom AI or use ChatGPT?
Start with the subscription. A custom build earns its cost when the job needs your data, your workflow, or your rules: answering from your documents, writing into your systems, or enforcing steps ChatGPT can't. If a subscription or a spreadsheet solves it, we'll say so on the first call.
What does it cost to run an AI tool after launch?
Two lines: model usage, which for most business tools is tens to a few hundred dollars a month, and maintenance, which is the one to ask about, because models and APIs change and someone has to own that. We scope ongoing ownership into the build conversation rather than leaving it as a surprise.
