The decision
Our builds start at $3,500 AUD, typical projects are $3,500 to $12,000 AUD and complex AI agent systems are $12,000 to $25,000 AUD. Care is $750 to $2,500 AUD per month. These are our published ranges, not an Australian market average. Your fixed quote follows the audit.
Build a budget with four separate lines
Start with the one-off build fee. It covers the agreed process and the work required to test and hand it over. A quote for a working business process is different from a quote for a prototype. Ask whether the price includes the difficult records, failure handling, documentation and the checks you will use to accept the work.
Then add ongoing care. Someone must watch the system, investigate exceptions and maintain the connections after launch. Our care range is $750 to $2,500 AUD per month. Its level depends on how much you run and what your business loses when the process stops. Do not infer response times or service coverage from a price band.
Third, account for the software. The automation platform, hosting, AI model usage, message delivery and application plan changes can all add costs. You keep these accounts in your name. If you already pay for an application, include only the additional cost created by the automation when estimating incremental payback.
Finally, allow for your team's time. Someone needs to explain the current process, settle ambiguous rules, review tests and learn the new way of working. This effort is easy to leave out because it may not appear on an agency invoice. It still affects when the project starts producing useful capacity.
| Cost line | What to establish |
|---|---|
| Build | The agreed outcome, exclusions and acceptance tests |
| Care | Monitoring, repair scope, escalation and review |
| Software | Platform, infrastructure and usage billed to your accounts |
| Internal effort | Process decisions, test review, training and exception handling |
Scope is a better price signal than the number of workflows
One process can be divided into many workflows, or several operations can live on one canvas. That makes a workflow count a poor basis for comparing quotes. Instead, follow one business event. Identify the input, the records it touches, the decisions it needs and the output someone will rely on.
A request that contains all required fields and creates an internal task is relatively bounded. The same request becomes more complex if it arrives as a PDF, conflicts with an existing customer record and needs approval before a financial action. The increased cost comes from interpretation, safe matching and decision control, not the extra boxes on the diagram.
The data can matter more than the apparent size of the task. If two applications do not share a stable customer ID, you need a matching policy. Decide what happens when names are similar, records disagree or required information is absent. Cleaning and reconciling historical data is separate work from keeping new records in sync.
Be specific about what happens outside the normal path. A quote that excludes all unusual records may be suitable for a limited pilot, but it is not equivalent to one that covers everyday exceptions. Ask both suppliers to price the same boundary before deciding that one is expensive.
AI adds interpretation and evaluation work
A conventional rule can decide whether an amount is above an agreed threshold. An AI model may help read an unstructured brief or draft a summary. These are different jobs. Using a model for arithmetic or permission checks can add uncertainty where an explicit rule would be clearer.
The budget for an AI step should include representative examples, a definition of an acceptable output and a way to handle missing evidence. If the model reads a document, what source supports each field? If two attachments disagree, should it ask a person? How does the system prevent a confident but unsupported answer becoming an external action?
Some costs recur with use. Larger inputs, longer outputs and repeated attempts can increase model usage. A document-heavy process should be estimated using realistic document sizes and exception rates. A cheap run on one short sample does not establish the operating bill.
Human review also has a cost. If each generated answer needs nearly as much checking as doing the work manually, the build may return little capacity. Measure the review effort in the pilot and include it in the payback estimate. Automation can still improve consistency or traceability, but those benefits should be named separately.
Compare the unit you are actually buying
n8n's paid plans use workflow executions alongside plan features and limits. Zapier uses tasks with current rates depending on the action or product involved. Make uses credits, with many ordinary non-AI operations using a fixed credit amount and some AI or advanced features using dynamic amounts. These units are not interchangeable.
Estimate a complete business event using your intended design. Include repeated records, scheduled checks, error recovery and AI usage. Then compare a normal month and a busy month against the current vendor plan. This gives you a more useful number than comparing three entry prices in different currencies.
A self-hosted deployment can change the budget. You may pay for compute, a database, backups and the work of operating them. There may also be paid features or a licence requirement for your intended use. Treat a free software edition as one line in the budget, rather than a promise that the whole system costs nothing to run.
Check whose account pays each bill, what happens at a usage limit and who sees the warning. A system can be well built and still stop because the card paying for a dependency expired. Billing ownership belongs in the operating handover.
Calculate capacity payback before calling it savings
Measure the time the process consumes now. Include preparing inputs, moving information, checking results and repairing mistakes. Then estimate the work that remains with automation, including approvals and exceptions. The difference is the capacity potentially returned to your team.
To express that capacity in money, multiply hours returned per week by your loaded hourly cost and by 52 divided by 12. Subtract monthly care and additional software costs. Divide the build fee by the remaining monthly value. If the remaining value is zero or negative, this simple model does not show payback.
This calculation values time; it does not prove an expense will fall. If your salaried team works the same hours after the build, the immediate benefit is available capacity. To call it a cash saving, identify an actual cost that will reduce. If you intend to use the time to serve more customers, explain what demand and margin support that plan.
Keep uncertain benefits separate. Faster replies may help win work, but attributing every later sale to the automation would overstate the case. Avoided errors may matter, but their value needs an observed baseline. Begin with a conservative calculation you can inspect and describe the other benefits in their own terms.
A worked example, including the weaker case
Assume a hypothetical $6,000 build, 10 hours returned each week, a loaded hourly cost of $60, $750 monthly care and $150 of additional monthly software. Monthly capacity value is 10 × $60 × 52 ÷ 12, or $2,600. After $900 in operating costs, the net capacity value is $1,700. Simple payback is $6,000 ÷ $1,700, about 3.5 months.
Now change only the time returned to five hours a week. Monthly capacity value falls to $1,300. With the same $900 operating cost, the net becomes $400 and payback becomes 15 months. If the process returns three hours, the $780 capacity value does not cover the $900 recurring cost. There is no positive capacity payback on those assumptions.
These hypothetical inputs illustrate the arithmetic. They are not a quote, an observed customer outcome or an industry benchmark. They exclude transition effort, tax effects and seasonal variation. Your written estimate should use your workload, expected adoption and actual proposed operating costs.
The comparison can change the scope. A smaller build that tackles a repetitive bottleneck may make sense before a broader system. Alternatively, the process may matter primarily because of traceability or risk control. If so, agree how you will evaluate that value rather than forcing every benefit into an optimistic hours-saved calculation.
| Hours returned weekly | Monthly capacity value | After $900 operating cost | Simple payback |
|---|---|---|---|
| 10 hours | $2,600 | $1,700 | About 3.5 months |
| 5 hours | $1,300 | $400 | 15 months |
| 3 hours | $780 | Negative $120 | No positive payback |
What to ask for before accepting a quote
Ask the supplier to state the process boundary in ordinary language. The scope should name what starts the work, what counts as complete and what remains outside it. It should identify the software involved, the permissions required and the person who can settle a business-rule question.
Next, ask how you will accept the work. Which ordinary and difficult records will be tested? What should happen when a destination is unavailable? How do you know an attempted write was not duplicated? These questions reveal whether the quote covers operation or only initial assembly.
Confirm who owns the accounts, workflows, exports and documentation. Establish the recurring care fee and separate vendor costs. Ask what is included in repairs, what counts as an enhancement and what happens if you stop the care plan. Confirm the final tax treatment and payable amount in writing.
You do not need a perfect technical specification before a first call. Bring a process description, rough volume and the names of the systems involved. Describe where people intervene and why. That is enough to begin assessing whether an audit and a scoped build are worthwhile.
- A defined outcome and a clear list of exclusions
- A fixed quote and explicit assumptions
- Acceptance tests covering ordinary work and exceptions
- Build, care and software shown separately
- Ownership, handover and an operating responsibility after launch
When spending less is the better decision
A process that happens rarely may not justify a custom system. A changing internal policy may need to settle before it is encoded. A built-in feature of your existing software may solve the problem with less maintenance than another connection. We check those possibilities before recommending a build.
If the opportunity is real but the uncertainty is high, narrow the first scope. Choose a complete, useful outcome with a clear acceptance test. Keep a manual path available during the transition and observe the exceptions. Expand when you understand the workload, rather than buying complexity in anticipation of demand that has not arrived.
Ongoing support is part of that decision. The process must be worth operating after the launch work is finished. Read the maintenance scope alongside the build price and make sure the people relying on it know what happens when it stops.
Before you decide
Your questions, answered
Are these Australian market averages?
No. The price ranges are our published SAA build and care prices. We do not present them as a survey of Australian agencies or an industry hourly-rate benchmark.
Do I have to use AI to automate a process?
No. Many repeatable tasks can use explicit rules and supported app connections. AI can help with unstructured information where interpretation is useful, but it also introduces evaluation and review work.
Does a positive payback calculation guarantee savings?
No. The example values capacity using stated assumptions. Cash savings require a real expense to fall. Adoption, remaining review time, usage and operating costs can change the outcome.
What ongoing costs should I expect?
Budget for care, your platform plan and any additional hosting, AI usage, messaging or application access. Your own team will also spend time on approvals and exceptions. The proposed design determines which lines apply.
Are the published prices inclusive of GST?
Confirm tax treatment and the final payable amount in the written scope. Our build and care figures are in AUD.
Sources and checks
References support the specific facts discussed above. Check dates show when each source was consulted. Vendor features and rules can change.
- Our published build and care prices ↗ · Checked
- n8n pricing ↗ · Checked
- Zapier task rates ↗ · Checked
- Make credit billing ↗ · Checked