For an Australian business, an AI chatbot can cost anything from a modest software subscription to a custom project with substantial setup and ongoing support. As broad, typical planning ranges, allow AUD $500 to $3,000 upfront for a simple website assistant, or AUD $15,000 to $50,000 for an assistant connected to business systems, plus ongoing costs.
The biggest price difference is not how convincingly the chatbot writes. It is what information it needs, what actions it can take and what happens when it gets something wrong. A bot that explains your opening hours is a different purchase from one that checks customer orders or searches confidential staff documents.
Start with the job you need done, then compare the full cost of delivering it safely. The ranges below are budget guides, not quotes or surveyed market averages. All amounts are in AUD and exclude GST.
AI chatbot costs in Australia: practical budget ranges
A useful budget separates the initial build from the cost of keeping the assistant available, accurate and secure. The monthly column below allows for a mix of software, usage and routine support, but providers bundle these differently.
These are broad, typical ranges for relatively contained business projects. Complex permissions, large document collections, sensitive information or difficult integrations can push costs beyond them. Ask every supplier what their estimate includes before comparing totals.
Typical planning ranges, AUD excluding GST
| Assistant scope | Upfront setup | Ongoing monthly | Main cost driver |
|---|---|---|---|
| Basic website FAQs | $500 to $3,000 | $50 to $500 | Content and configuration |
| Document-based answers | $3,000 to $15,000 | $200 to $2,000 | Content quality and testing |
| Connected customer assistant | $15,000 to $50,000 | $1,000 to $5,000 | Integrations and safeguards |
| Internal knowledge assistant | $5,000 to $25,000 | $300 to $3,000 | Permissions and document scope |
What your setup fee should actually cover
A setup fee should buy more than a chat window on your website. It should cover agreeing on the assistant's purpose, preparing its information, configuring its behaviour and testing it against realistic questions.
For a trades company, that might mean explaining service areas, collecting enquiry details and handing urgent jobs to a person. It should not mean inventing a fixed price for work that needs an inspection.
For a distributor, answering general delivery questions is relatively contained. Showing a customer's order status adds identity checks, access to the order system and handling for missing or delayed data. Those requirements explain much of the extra cost.
Ask for clear acceptance criteria. You should know what the assistant must answer, what it must refuse, when it must hand over and who signs off before launch.
The scope decisions that move your quote
Information quality Clear, current service pages are easier to use than conflicting PDFs and old policy documents. Someone still needs to decide which source is correct.
System connections Reading a public page is simpler than looking up an invoice or changing a booking. Each connection needs access controls, testing and a plan for outages.
Customer verification General answers may need no login. Customer-specific answers require a reliable way to establish who is asking and what they can see.
Risk and escalation A wrong opening time is inconvenient. A wrong clinical instruction or unauthorised account change is much more serious. Higher-risk tasks need tighter boundaries and review.
Channels and usage A website-only assistant has a different scope from one operating across messaging platforms and voice. Usage limits and channel charges also affect ongoing spend.
Monthly fees are more than the AI model bill
The AI model is only one running cost. Depending on the product, you may also pay for the chatbot platform, hosting, document search, conversation storage, monitoring and support.
Some subscriptions charge by conversation, message, staff seat or usage credit. These units are not interchangeable. Ask what counts as a billable conversation, whether a long exchange costs more and what happens when you reach the included limit.
Content maintenance matters too. If your prices, policies or service areas change, someone must update the source material and check the answers. Confirm whether that work is included, charged separately or left to your team.
Request a monthly estimate for normal use and a busier period, with the assumptions shown. Set spend alerts and agree on a cap or fallback behaviour where the platform supports it.
Choose a subscription or custom build by the job
An off-the-shelf subscription is often the sensible starting point when you want answers from approved website pages and a basic handover to staff. You still need to check answer quality, data handling and whether you can export your information when you leave.
A custom build becomes more relevant when the assistant needs unusual workflows, detailed permissions or connections that standard products do not support. Custom does not automatically mean better. It means you are paying for a closer fit and accepting more responsibility for maintenance.
An accounting practice looking for internal procedure answers may be well served by an assistant within its existing software environment. But it must respect staff access rights. Uploading every client folder into a shared chatbot is not a sensible shortcut.
Before commissioning anything new, check what your current software already offers and what its AI features cost. An existing option may be sufficient once configured properly.
Check whether the cost is worth it
Measure the problem before buying the solution. Look at the repetitive questions your team receives, how long they take to resolve and how many genuinely need a person's judgement.
Estimate the staff time the assistant could save, then subtract the time needed to review conversations, maintain content and handle escalations. Time saved is useful capacity, but it is not automatically a reduction in wages.
For a customer-facing assistant, track completed useful tasks rather than chat volume alone. Did the customer find the correct policy, submit a usable enquiry or reach the right staff member? A conversation that ends without a handover is not necessarily a successful answer.
Agree on a review point before launch. If the assistant produces frequent corrections or creates more follow-up work, narrow its scope before adding features.
What to ask suppliers before you approve a quote
Ask for an itemised scope Separate setup, integrations, licences, usage and support. Confirm GST treatment and whether any services are billed in a foreign currency.
Make testing explicit Provide real questions, including ambiguous requests and questions the assistant should not answer. Agree on acceptable responses and escalation behaviour.
Check data handling Ask where information is stored and processed, who can access it, how long conversations are retained and whether inputs may be used to train models.
Confirm the handover Find out who receives an escalation, what context they receive and what the customer sees outside business hours.
Understand your exit options Check ownership and export rights for your content, configuration, conversation records and custom code. Ask what stopping the service involves.
Start with a bounded pilot this week
Choose one audience and one repeatable job. For a clinic, that could be answering approved administrative questions about parking, opening hours and booking procedures, while directing symptoms and clinical questions to the appropriate human channel.
Collect the approved information, nominate a content owner and write down the questions you expect. Include awkward cases such as outdated information, incomplete requests and attempts to obtain private details.
Give suppliers the same brief and ask them to price the pilot separately from future integrations. That makes quotes easier to compare and gives you a way to test usefulness before committing to a broader rollout.
Related reading
Frequently asked questions
- Can I put a free AI chatbot on my business website?
Some products offer free plans or trials, but limits may apply to usage, branding, data sources and support. Check data handling and test the answers before exposing it to customers. Free software still takes staff time to configure and maintain.
- Do I need to train my own AI model?
Usually not for an assistant answering questions from business documents. Many systems retrieve relevant approved content and give it to an existing model when answering. Preparing that content and testing responses still requires work.
- Is Australian data hosting included in the price?
Do not assume it is. Ask separately about storage, model processing, backups and support access, because these may occur in different locations. Confirm your privacy and contractual requirements before choosing a platform.
- Can an AI chatbot replace my receptionist?
It may handle a bounded set of routine enquiries, but that does not make it a full replacement for a receptionist. Exceptions, distressed customers and sensitive requests still need people. Budget for a reliable handover rather than assuming every conversation can be automated.
The right chatbot budget depends on its responsibilities, not just its subscription price. Compare setup, usage, maintenance, data safeguards and human handover as one complete cost.
Your best next step is a short, specific brief: the audience, the questions, the approved information and the tasks the assistant must not perform. Price that scope first, prove it is useful, then decide whether more complex features are worth funding.
Talk to CU Digital Intelligence about a customer-facing or internal AI assistant, with a clear scope, realistic running costs and a practical plan for testing answer quality. See how we approach AI Chatbots
Working on something similar?
Tell us what you’re trying to build, improve or automate.