AI automation pricing. Know both numbers.

Start with the process costing you the most

You may be retyping customer details between a CRM and Xero, checking a shared inbox for missing documents or asking someone to compile the same report every Friday. A useful first build has a business owner, a recurring workload and an outcome you can recognise. It does not need to involve AI.

For a trade business, the starting point might be the handover from an accepted job to an invoice draft. For a professional services firm, it could be collecting the information needed before work starts. We check the exact fields, access and supported actions in your software before promising the connection. A product name alone does not establish that every job can be automated.

Choose one bounded process first if several teams disagree about how it should work. You can expand a tested foundation later. Paying to automate an unresolved process usually creates a faster version of the same confusion.

What the build should change in your working day

Enquiries can reach the right person with the context already attached. Quotes and invoice drafts can use approved source records instead of another round of typing. Systems can share a stable customer or job reference so the team can trace one piece of work across them.

Larger builds can check and reconcile batches of records, route sensitive decisions to a person and coordinate work across several teams. The scope should say which of those outcomes you are buying, how exceptions reach you and which actions remain manual. A workflow count does not answer any of those questions.

We build in your own n8n, Zapier or Make account, on your credentials. Monitoring and documentation are part of the build. Before sign-off, you watch the process run against your real data. Handover includes workflow documentation, exported files and a recorded walkthrough.

What moves a project through the range

The main drivers are the number of systems, the access each system gives us, the condition of the data and the exceptions the process must survive. Two systems with a suitable API and consistent records can be more straightforward than one system with no supported way to read the required information.

Data cleanup can be substantial work. Similar company names, duplicate contacts and missing customer IDs need an agreed matching rule. If the system cannot distinguish two records safely, the correct output is a review task. Guessing makes a tidy demo and a difficult accounting problem.

An AI step adds another kind of work. We need representative examples, a definition of an acceptable answer and a response when evidence is missing. Approving an invoice or sending a commitment to a customer needs stronger controls than drafting an internal summary. The number of prompts is a poor proxy for that responsibility.

Volume matters both during a normal week and at peaks. A monthly import, a seasonal rush and retries after an outage can change the design. The quote should identify expected workload and constraints rather than promise unlimited capacity.

• Connections and the exact read and write actions required
• Cleanup, record matching and migration scope
• Approval rules, exceptions and the cost of a wrong action
• Expected volume, peak workload and recovery after interruptions

Separate the build, the care and the software

The build fee pays for the agreed process, testing and handover. Care pays for monitoring, fixing and improving the operating system after launch. Your n8n, Zapier or Make subscription is a separate operating cost in your own account. We do not resell platform subscriptions.

Other costs can include hosting, AI model usage, message delivery and a paid API feature in software you already use. We identify what the proposed system needs and put those costs beside the build and care figures. Existing software bills should only count as an additional automation cost where the project changes them.

Ask for a written list of inclusions and exclusions, the assumptions behind the quote and what would count as a change in scope. Tax treatment, any particular support coverage and the final commercial details should be confirmed in that written scope.

A payback estimate you can inspect

You get the payback period in writing before you sign. We start with the time spent on the current process and the time expected to remain after automation, including checking exceptions. Hours returned to your team are capacity. They become cash savings only when an actual expense falls, or value when that capacity is used productively.

A conservative comparison sets the build fee against the monthly value of time returned, less care and additional platform costs. Use your loaded hourly cost, your own volumes and a realistic adoption assumption. Keep revenue growth and avoided errors separate unless you have evidence to value them.

For a clearly hypothetical example, a $6,000 build returning 10 hours each week at $60 per hour gives $2,600 of monthly capacity value using 52 weeks divided by 12. Subtract $750 care and an assumed $150 of additional software costs. The remaining $1,700 per month gives about 3.5 months of simple capacity payback. These are scenario inputs, not our results or a forecast for your business.

If only five hours are returned, the same inputs leave $400 per month and a 15 month payback. That is why we test the assumptions before choosing the scope. Training, transition work, seasonal variation and tax effects are outside this simple example and need separate consideration.

You keep what you paid to build

You own the workflows, the documentation and the accounts. If we stop working together, you keep them and the system keeps running. Your continued use still depends on your own subscriptions, credentials and infrastructure being maintained. Leaving an agency does not remove those operating responsibilities.

Care is month to month and you can cancel without penalty. Before a handover to your team or another maintainer, make sure someone is taking responsibility for alerts, billing, access changes and recovery. Exported files are useful, but they are only part of the handover. The explanation of what each process does and why it stops matters just as much.

Budget for the day something changes

A failed run is the obvious problem. A run that completes with no output and a run that never starts can be harder to see. Our care work covers all three, with a monthly report of what ran, what failed, what we repaired and the improvement worth making next.

The care range is $750 to $2,500 AUD per month, priced on how much you run and what your business loses when it stops. We agree the actual coverage, escalation path and responsibilities in the written scope.

If a process affects customers, payments or sensitive records, the recovery plan should say when to stop it and let a person decide. Repeating a failed action without checking the destination can send a second message or create a duplicate record. Useful monitoring is connected to safe repair.