Why we publish ranges
Most operators have been burned by an agency engagement that started at "let us scope it" and ended at a number they could not afford or, worse, a number that ballooned through the build. Hiding the range protects the agency. It does not protect the client. We would rather you arrive at the first call with an idea of what realistic work costs and decide whether we are worth your time. If we are not the right fit, neither of us wants to discover that after three meetings.
The numbers below are the bands we have seen across our actual engagements. They are not aspirational. They are what real builds for real Australian operators have cost over the last three years.
What things actually cost
Floors below are the entry point. Upper bounds depend on scope, system count, compliance posture, and timeline. The figure your engagement lands at is set during scoping.
- Discovery sprint. From $15,000. Two-week scoped engagement to assess your environment, validate the workflow, and produce a recommended architecture plus a fixed-price quote for the build. Often pays for itself in saved scope mistakes.
- AI agent build. From $40,000. Highly dependent on system count, access control needs, and how mature the source data is. A focused single-workflow agent against three to five systems sits near the floor. Multi-step agents with workflow actions, identity-aware retrieval, and ten-plus integrations scale up from there.
- Bedstone OS deployment. From $80,000 for the first deployment, scaling with integration count and access requirements. Followed by a monthly operate retainer from $8,000 covering new integrations, retrieval tuning, and model upgrades.
- Custom software build. From $60,000. CRUD applications with light workflow sit near the floor. Multi-tenant SaaS with complex domain logic, hard integration requirements, and regulated data scale higher. Web and mobile builds, internal tooling, customer portals, and finance systems all sit in this category.
- Financial systems. From $50,000. Quote-to-cash, invoicing, reconciliation, board-pack automation. Scales with the number of finance and CRM systems integrated and the compliance posture.
- Security audits. From $15,000. Web and API audits sit near the floor. Full application plus infrastructure plus cloud configuration audit scales up. IRAP-aware audits price separately.
- Penetration testing. From $20,000. Scales with attack surface. Web-only, API-only, or mobile-only sit near the floor. Full external and internal red team scales higher. Re-test typically included.
- Cloud infrastructure (one-off). From $40,000. Greenfield AWS or Azure landing zone, migrations, IaC adoption, observability rollout. Smaller environments sit near the floor; regulated and IRAP-aware deployments scale higher.
- Monthly retainer. From $12,000 per month, scaling with engineering capacity reserved. Common for continuous build and operate work, ongoing Hub maintenance, or embedded senior engineering inside a client team.
- Embedded engineering. From $25,000 per engineer per month, three-month minimum. Senior engineering capacity inside your team. Common for organisations growing into in-house AI capability.
- Build-operate-transfer. Hybrid. The build is fixed-price. The operate phase is monthly retainer. The transfer to your in-house team is fixed-price for the knowledge transfer engagement.
If your project has scope beyond the floors above (large system count, sovereign or PROTECTED-classified deployment, multi-region rollout, multi-year operate horizon), tell us at the scoping call and we will quote against the actual shape. We do not cap engagements.
What shapes the price
The biggest determinants of cost in order:
- Number of source systems. Every additional system adds connector work, schema discovery, identity mapping, and rate-limit handling. The first three systems set the baseline. Each system after that adds approximately $8,000 to $20,000.
- Access control complexity. If everyone in the business can see all of the data, integration is fast. If access scoping has to mirror role-based, field-level, or document-level permissions, integration takes considerably longer.
- Data quality of the source systems. If your CRM is reasonably tidy, building against it is faster. If it has years of accumulated inconsistency, the project includes a data quality pass before the AI work starts.
- Compliance posture. Australian Privacy Act and OAIC obligations apply by default. APRA CPS 234, IRAP, sovereign data residency, and SOC 2 add evidence trails and architectural constraints that increase build time.
- Identity and SSO maturity. If you already run Entra ID, Okta, Google Workspace, or Auth0 cleanly, integration is fast. If identity is fragmented across separate user directories, we factor in a consolidation pass.
- Existing cloud account state. Greenfield deploys to a new account are fastest. Deployments into existing accounts with legacy IAM, networking, and observability already in place take longer because we work around what is already there.
- Model and provider choice. Commercial APIs (OpenAI, Anthropic, Google) are fastest to integrate. Open-weight models in your own VPC require GPU provisioning, inference serving, and ongoing operations work that adds to the build and the retainer.
- Workflow action scope. Read-only agents are fastest. Agents that take actions (create tasks, send emails, update records) require identity-bound API access and reversibility patterns that add to the build.
Engagement types
Bedstone uses four primary engagement types. Most clients move through several of them across the relationship.
Discovery sprint
Two-week paid scoping engagement. The output is a recommended architecture, integration plan, identified risks, and a fixed-price quote for the build. From $15,000, scaling with the breadth of systems to assess. Common starting point when the brief is broad or when the operational environment has not been audited recently. The sprint produces real work product, not just a slide deck. Most clients then commission a build engagement; some take the discovery output to a different agency, which is fine.
Fixed-price build
Defined scope, timeline, and price. Common for AI agents, Bedstone OS deployments, custom software builds, dashboards, and integrations under three months. Both parties know exactly what will be delivered and when. Change orders for material scope additions are agreed in writing. Most of our work fits this type.
Monthly retainer
Continuous build and operate work on a monthly engagement. Common after a successful fixed-price build, when the system is in use and the operator wants ongoing improvement, integration of new systems, retrieval tuning, model upgrades, and incident response. From $12,000 per month, scaling with engineering capacity reserved. Cancellable with 30-day notice.
Embedded engineering
Senior engineering capacity inside your team for a defined period. The engineer participates in your standups, attends your planning, commits to your repository. Common for organisations growing in-house AI capability or backfilling capacity during a critical build. From $25,000 per engineer per month, three-month minimum, scaling with seniority and reserved capacity.
Build-operate-transfer
We build, we run, we hand back. The build phase is fixed-price. The operate phase is monthly retainer (commonly six to eighteen months). The transfer to your in-house team is a fixed-price knowledge engagement covering documentation, walkthroughs, paired engineering sessions, and a final operational readiness review. Common where you want eventual ownership but lack the team today.
Engagement types compared
| Type | Best for | Trade-off |
|---|---|---|
| Discovery sprint | Unclear brief, ambiguous environment, or first engagement with us. | Two weeks before the build begins. Most clients consider that an asset, not a cost. |
| Fixed-price build | Defined scope, clear timeline. Most of our work fits here. | Change orders required for material scope additions. Both parties stay disciplined about scope. |
| Monthly retainer | Continuous improvement, new integrations, ongoing operate work. | Open-ended commitment. Cancellable with 30-day notice, so the lock-in is short. |
| Embedded engineering | You want senior capacity inside your team without a hire. | Higher monthly cost than retainer because the engineer is reserved for you. |
| Build-operate-transfer | You want eventual in-house ownership but cannot hire the team today. | Requires planned transfer engagement. Knowledge transfer is the work, not an afterthought. |
The discovery sprint
Most builds we run begin with a discovery sprint. It is two weeks of scoped work, paid, with concrete output. The sprint typically includes:
- Operational walkthrough of the systems, workflows, and stakeholders in scope.
- Technical audit of the source systems we expect to integrate against, including API surface, rate limits, identity model, and data quality.
- Identity and access mapping. Who can see what. What scoping the AI workflow must respect.
- Compliance posture review. AU Privacy Act, OAIC, sector-specific frameworks (APRA, healthcare, government). What constraints apply.
- Architecture recommendation. Where things sit. How they connect. Which model providers fit the data sensitivity.
- Risk register. What we believe will be hard. Where the unknowns are. What we will discover during the build.
- Fixed-price quote for the build, with assumed scope, deliverables, timeline, and change-order triggers documented.
The sprint is not a slide deck. It is a delivered work product you can hand to another agency if you decide we are not the right fit. The artefacts are yours.
How we quote
For most build engagements, the quote process runs:
- Brief from you via the contact form. As much as you want to share. We reply within 24 hours.
- 30 to 45 minute scoping call. No SDR layer. The conversation is with the engineer who would lead the work. We share by the end of the call whether we are the right fit and why.
- Either a direct fixed-price proposal (where the scope is clear enough from the call) or a discovery sprint proposal (where it is not). Both arrive within three business days of the call.
- Written engagement letter. Scope, timeline, deliverables, price, payment schedule, IP terms, change-order process, termination terms. Nothing material is missing.
- Signature and kickoff. First-week work begins within a week of signature.
Change orders and scope creep
Scope creep is the largest source of agency-client conflict. We handle it by being explicit:
- Small clarifications (a renamed field, a different output format, a minor design adjustment) are absorbed without a change order. We do not run a meter for trivial requests.
- Material additions (a new integration, a new workflow, a new compliance obligation) are quoted and agreed in writing before the work starts.
- Additional time required. If during the build we hit complexity that requires additional time, we tell you straight away, scope the variation, and only bill once you have agreed in writing. No silent overruns, no surprise invoices.
- Reductions (you no longer need a planned deliverable) reduce the engagement value proportionally. We do not pocket the saved time.
Payment terms
Standard payment terms for fixed-price builds. Paid upfront, on purpose: it lets us put senior people on your build from day one, locks in your engineering window, and removes the cash-flow friction that slows engagements down.
- 100% on signature is the default. Engineering capacity is reserved, infrastructure is provisioned, and the build is on the calendar the day payment lands. You get a faster start and our undivided attention.
- 50% / 50% on request. Where client procurement requires staged payment, we accept 50% on signature and 50% on delivery against the acceptance criteria in the engagement letter. Just ask during scoping.
Monthly retainer is invoiced on the first business day of the month, payable within 7 days. Embedded engineering is invoiced on the first business day, payable within 7 days. We are open to alternative payment cadences for client-side procurement constraints.
IP and ownership
You own the work product. All bespoke code, prompts, configurations, infrastructure-as-code, documentation, and operational runbooks produced for your engagement transfer to you on payment. We retain no licence on the bespoke deliverables.
Where we use our own libraries, frameworks, or patterns developed across multiple engagements (for example, the retrieval scaffold underlying Bedstone OS), you receive a perpetual, royalty-free, worldwide licence to use them inside the deliverable.
R&D Tax Incentive
Custom AI agent builds, novel software work, and the experimentation involved in production AI engineering commonly qualify for the Australian R&D Tax Incentive (RDTI). We document the technical experimentation, the hypotheses tested, the uncertainty resolved, and the experiments conducted in a form your RDTI consultant or accountant can use. This is not a side activity; it is built into how we work.
We do not provide tax advice. Whether a specific engagement qualifies depends on your circumstances and is assessed by your accountant or a registered RDTI consultant. What we can do is provide the technical narrative and the records.
Common questions
What does an AI agency in Australia actually cost?
Production AI agent builds start at $40,000, with the upper bound depending on system count, workflow complexity, and access control requirements. Bedstone OS deployments start at $80,000 for the first deployment. Custom software builds start at $60,000 and scale with scope. Security audits start at $15,000. The discovery sprint starts at $15,000. We do not cap engagements at the upper end; the figure for your build is set at scoping.
Do you do hourly rates?
No. Hourly engagements create incentive problems and produce unpredictable bills for clients. We quote fixed-price for defined scope and monthly retainer for continuous work. Senior engineering capacity through embedded engagements is also priced monthly.
What is the smallest engagement you take?
The discovery sprint starting at $15,000 is the smallest commercial engagement. It produces a clear plan and architecture. Many clients begin there and decide on a build engagement afterwards.
Do you work with startups?
Yes, where the work is suited to our practice. We focus on production systems, so we are less useful for very early-stage prototype work. We are well suited to startups that have product-market fit and are now building the operational backbone.
Are these eligible for the R&D Tax Incentive?
Custom AI agent builds and novel software work commonly qualify for the R&D Tax Incentive. We document the technical experimentation that the RDTI requires. We do not provide tax advice. Your accountant or RDTI consultant assesses eligibility against your circumstances.
Do you offer fixed-price or time-and-materials?
We default to fixed-price for defined scope. We use monthly retainer for continuous build and operate work. We do not do open-ended time-and-materials because it disincentivises both parties from being efficient.
What happens if scope changes mid-build?
We agree change orders before doing the work. Small clarifications are absorbed without a change order. Material scope additions are quoted, agreed in writing, and added to the engagement. If during the build we hit complexity that requires additional time, we tell you straight away, scope the variation, and only bill once you have agreed in writing. No silent overruns, no surprise invoices.
Who owns the IP at the end?
You do. All bespoke code, prompts, configurations, and infrastructure-as-code produced for your engagement transfer to you on payment. Where we use our own libraries or patterns developed across engagements, you receive a perpetual, royalty-free, worldwide licence to use them inside the deliverable.