Direct answer
Budget depends on process clarity, data quality, integration complexity, risk, exception volume and the support model. Start with a bounded review, define the baseline and value logic, then price the smallest implementation that can be tested safely. Advery scopes pricing privately once those conditions are understood.
Cost follows scope, data and risk
Compare proposals on outcome, ownership and ongoing responsibility, not day rate.
Choose the right commercial scope
| Stage | Typical purpose | What should exist at the end | Commercial question |
|---|---|---|---|
| Opportunity Review | Baseline, workflow selection, risks and implementation plan | Evidence pack, priorities, scoped next step and decision | Is there enough value and clarity to proceed? |
| Value Sprint | One focused workflow with limited integration | Working process, tests, documentation, handover and measures | Can one bounded change prove the result? |
| Connected implementation | Several systems, teams or customer journeys | Integrated workflows, safeguards, adoption and support model | Which proven components should now scale? |
| Operating Partner | Monitoring, support, changes and continuous improvement | Service reporting, issue response and managed backlog | Who owns performance after launch? |
A smaller engagement may be the right decision when the workflow is narrow and the business can own more of the configuration. A broader implementation may be justified when failures are costly, several systems are involved or adoption work is material. Advery prices each engagement privately after the scope and evidence requirements are clear.
What changes the implementation cost
- Process clarity: unclear ownership and inconsistent rules add discovery and redesign effort.
- Data quality: duplicates, missing fields and inconsistent records require remediation.
- Integration method: stable APIs are usually easier than exports, browser automation or bespoke legacy access.
- Risk and safeguards: sensitive data, consequential decisions and audit requirements increase design and testing.
- Exception volume: the unusual cases often determine whether a workflow is genuinely usable.
- Change and training: several teams or locations need more adoption support.
Budget for operating cost, not just build cost
Include software subscriptions, model usage, integration platforms, hosting, monitoring, support, security review, staff administration and periodic change. Ask which costs are fixed, which scale with volume and which are paid directly by the client.
Build a return range before approving the work
Estimate current volume, minutes per item, loaded staff cost, delay, rework, missed revenue and service impact. Then apply a conservative achievable improvement rather than assuming every minute disappears from payroll.
Recovered time is capacity, not automatically cash. State how that capacity will be used: faster service, more throughput, reduced overtime, avoided hiring, stronger assurance or higher-value work.
Compare proposals properly
- Confirm the exact workflow and excluded work.
- Compare who owns accounts, configuration, documentation and data.
- Check testing, security, privacy, human review and failure handling.
- Identify third-party and usage costs over twelve months.
- Require a baseline, acceptance criteria and post-launch measurement.
The best proposal is the one that makes the commercial result, operating safeguards and ongoing responsibility easiest to evaluate.
Sources and guidance
- Australian Government: Guidance for AI adoption
- OAIC: Privacy and commercially available AI products
- ASD ACSC: Questions to ask managed service providers