A useful AI budget starts with one business process. Here’s how to compare quotes, account for ongoing work and decide whether the improvement is worth paying for.
The short answer
Among five Australian providers that publish prices (checked 26 October 2026), a paid discovery or audit costs roughly A$2,000 to A$10,000, and a build for one defined workflow starts at about A$3,000 to A$5,000, with multi-system builds running higher. AI features in tools such as Google Workspace, Microsoft 365 Copilot or ChatGPT Business cost roughly $10 to $40 per user per month. But there is no single meaningful price for “AI implementation”. Configuring an existing feature, connecting two business tools and building a custom operating system are different jobs. Ask for a scoped setup price, an ongoing cost breakdown and a way to check whether the workflow actually improves.
First, work out which job you’re buying
If you run a plumbing company and quotes go unanswered, the requirement might be as simple as setting up reminders in your job software. If enquiries arrive in several places and someone manually works out which quote, customer and next action belong together, the job may involve an integration. Those should not receive the same proposal.
Separate the work into three possible starting points. These are scoping categories, not fixed AppliedAgents packages or market price bands.
| Starting point | What you’re buying | What to check |
|---|---|---|
| Configure what you have | Settings, templates and training inside an existing product. | Does the feature cover the actual process, including when to stop or hand work back? |
| Connect one workflow | A defined trigger, supported connections, checks and a clear destination. | Can the systems exchange the required information reliably? Who handles exceptions? |
| Build a custom system | New software or a broader process spanning several tools, roles and permissions. | What is the smallest useful first release? Who owns, maintains and supports it? |
AI is one possible component. A fixed rule can move an approved record between systems; AI may help interpret an unstructured enquiry or prepare a draft. Ask what decision needs AI and what happens when its output is uncertain.
Separate setup, running costs and your team’s time
A cheap build can become expensive if it needs constant checking. A larger initial quote may include testing, handover and support that another supplier has left out. Compare the same boundaries.
- Discovery and process design: understanding the current steps, choosing a baseline and agreeing what “done” means.
- Implementation: configuration, connections, data preparation, permissions and any custom interface.
- Testing and launch: realistic examples, duplicate or missing records, failed connections, approval rules and a controlled rollout.
- Ongoing services: software subscriptions, usage charges, hosting, monitoring, support and future changes. Ask which are included and which are billed separately.
- Internal time: giving access, explaining exceptions, reviewing outputs and teaching the team the new process.
For an Australian quote, ask whether amounts are in AUD, whether GST is included, and whether any underlying subscriptions are charged in another currency. Request a normal-usage estimate and a higher-usage example based on your own enquiry or job volume.
What Australian providers and AI tools publish
Most implementation firms, including AppliedAgents, quote after scoping. A few publish their prices, which gives you real reference points. The figures below come from each provider’s own website on 26 October 2026. They describe each firm’s own packages, so they are not like-for-like quotes, and including them is not an endorsement.
| Provider | Discovery or audit | Build | Ongoing |
|---|---|---|---|
| Hourglass AI (Sydney and Melbourne) | $2,500 flat, 7–14 days; credited to a build started within 30 days | $5,000–$25,000 per build, about two weeks each | About $1,000 a month per build; AI usage billed separately |
| Sandlabs (Melbourne) | A$1,999 discovery and roadmap, ex GST; credited if you build within 30 days | From A$5,000 ex GST, fixed price, 2–6 weeks | Not listed as a standard fee |
| Better Automations (Melbourne) | Not listed | $3,000–$8,000 fixed for one process; $8,000–$15,000 for a system spanning several tools | Hosting described as usually tens of dollars a month in your own accounts |
| Cipher Projects | A$2,650 audit, exclusive of tax; credited to a later build | Quoted as a fixed cost after the audit | Not listed |
| Works | A$5,000 for up to 50 people; A$10,000 for more than 50; about two weeks | Quoted separately | Not listed |
Three patterns are worth noticing. Paid discovery is common, and it is often credited against a build. A “build” usually means one defined process, so a business-wide system is several builds or a larger engagement. And running costs are listed separately from the build, including AI usage, which most of these providers do not include in their fees.
The cost of AI tools you may already be paying for
Before commissioning anything, compare the cost of switching on AI features in software your team already uses. These are list prices per user per month on the vendors’ own pages, checked on the same date.
| Tool | Published price | Notes |
|---|---|---|
| Microsoft 365 Copilot Business | AU$31.40 a month paid yearly, or AU$37.68 billed monthly | An add-on to an eligible Microsoft 365 business plan. A first-year promotional price of AU$26.91 runs until 31 December 2026. |
| Google Workspace | Business Starter A$9.90; Business Standard A$19.80 | Gemini features are included; Starter covers Gemini in Gmail and the Gemini app, Standard adds Docs, Meet and more. |
| ChatGPT Business | $20 a seat billed annually, or $25 monthly | Minimum two users. OpenAI’s page does not name the currency; check your billing currency. |
| Claude Team | $20 a seat billed annually, or $25 monthly | Standard seat; 2–150 seats. Check the billing currency at checkout. |
A tool subscription and an implementation solve different problems. Seats give your team a capable assistant. An implementation connects that capability to your enquiries, records and handovers, with checks around it. Many businesses need the first before they know whether they need the second.
Prices change, promotions end and currency or GST treatment varies. Confirm current pricing with each provider before relying on it. The figures here are published list prices, not quotes for your project.
What changes the scope for different workflows?
Two projects described as “one automation” can involve very different work. Compare the information sources, decisions and exceptions before comparing the price.
| Workflow | A bounded first scope | What adds work |
|---|---|---|
| Consulting client onboarding | One accepted engagement type passed into a project record and an internal checklist | Different service checklists, inconsistent scope documents, access approvals and several project systems |
| Ecommerce order enquiries | One support channel preparing a case from verified order information for a person to review | Multiple stores or warehouses, split shipments, stale carrier information and changes to live orders |
| Quote follow-up | One quote source, an agreed reminder sequence and clear stop conditions | Conflicting quote versions, replies in different inboxes and unclear ownership of the next action |
Ask what information the supplier needs to confirm the quote and which unresolved assumptions could change it. A cheaper proposal that excludes the difficult handovers may cover less of the job you actually need.
Use the consulting onboarding example or ecommerce order-support example to write a specific scope before asking providers to estimate it.
A worked budget example
Suppose a business is considering automating part of a six-hour weekly admin task. The figures below are deliberately hypothetical: they are not our prices, market averages or a client result. The setup figure sits near the low end of the published build prices above.
| Assumption | Illustrative value |
|---|---|
| One-off setup | A$3,000 |
| Ongoing system cost | A$150 per month |
| Current task time | 6 hours per week |
| Time actually released after checking | 50%, or 3 hours per week |
| Value assigned to that time | A$45 per hour |
| Monthly capacity value | 3 × 4.33 weeks × A$45 = about A$585 |
| Capacity value less running costs | About A$435 per month |
| Setup divided by that monthly value | About 6.9 months |
That last number is a capacity-value comparison, not a cash-payback promise. A salaried employee finishing the task sooner does not automatically reduce payroll. The value depends on what the business can do with the released time. The simple model also leaves out internal launch time and any later changes.
Stress-test it. If only one hour is released each week, the same calculation gives about A$195 of monthly capacity value. After A$150 of running costs, roughly A$45 remains: the setup comparison stretches to about 67 months. That is a very different decision.
Use a week of real observations to replace the assumptions. Count the time spent checking and correcting the new workflow, not just the steps it removes. Keep possible extra sales separate until you have evidence of an incremental contribution.
Ask every provider the same seven questions
- What exact process is included? Name the trigger, systems, users, output and work that remains manual.
- What can our existing software already do? Ask for the native option alongside any proposed custom work.
- What access or plan changes are required? Confirm permissions and integration feasibility before treating them as included.
- How will we accept the work? Agree representative examples and failure cases the system must handle before launch.
- Who sees mistakes? Establish an owner, a clear alert and a way to pause the workflow.
- What will we own and be able to export? Clarify accounts, documentation, source code where relevant, and the exit or handover process.
- What changes the bill? Identify usage allowances, exclusions, support boundaries and how new requests are quoted.
Choose a pilot you can judge
Start with a repetitive task that has a clear input and a clear end. “Get AI into the business” is too broad. “Prepare an enquiry record from the website form and send it to the office for review” is something you can scope and observe.
Before a pilot, record the current weekly volume, handling time, common mistakes and who owns the next step. During it, count completed handovers, exceptions and review time. Use those observations to decide whether to adjust, expand or stop.
For a trade business using ServiceM8, our guide to native and custom quote follow-up is a concrete place to start. For a wider process, see how we approach workflow automation and integrations.
How AppliedAgents approaches a quote
We start by understanding where work stalls, which tools you use and what you want to improve. We check the existing-tool option before recommending a connection or custom build. Scope, testing, handover and support should be clear in the proposal.
We do not publish a universal implementation price because a single follow-up workflow and a custom business system involve different work. Bring the task, approximate weekly volume and a list of your tools to a free strategy call. Those details make a useful discussion possible without having to arrive with a technical specification.
This is an AppliedAgents planning framework. All example costs and time values are invented for the calculation and should be replaced with your own evidence before committing to a project.
Practical thinking on AI, automation and the work behind a business.
