Every Australian mid-market company sits somewhere on a five-stage path from AI curiosity to AI-native operations. The question worth asking at your next board meeting is not whether to invest in AI. It is which stage you are actually in, and what the next investment looks like. Get that wrong and you waste capital on the wrong problem. Get it right and you compound advantage every quarter.
This piece sets out the five stages we see across our Australian client base, the signals that tell you where you sit, and the AUD investment range that typically fits each stage. We work with Claude as our primary model across these engagements, so the patterns below reflect what real AU companies have funded between Sydney, Melbourne, and Brisbane in 2025 and 2026.
Why a stage model matters for Australian boards
Most Australian boards we sit with have no shared language for AI progress. The CIO calls a single chatbot pilot a transformation. The CFO sees the line item and asks whether the spend belongs in IT or operations. The CEO reads about an APRA expectation and asks whether the company is exposed. A stage model gives the board a common frame so the right question gets asked at the right time, with the right person on the hook.
The five stages are Exploration, Pilot, Integration, Scale, and AI-Native. Companies do not skip stages. They can move through them quickly, and they often run two stages in different parts of the business at once, but the work of each stage has to be done. Skip it and the next stage collapses under its own weight inside twelve months.
Stage 1: Exploration
This is where most Australian businesses sat in 2023 and early 2024. A small group of curious staff use Claude or another assistant from personal accounts. The CIO knows it is happening and tolerates it. There is no policy, no procurement record, and no spend tracked against AI as a category.
Signals you are in Exploration
No dedicated AI line item in the budget; spend is hidden inside software or training codes
Shadow usage is the norm: employees pay for Claude or ChatGPT personally and reimburse via expenses
The board hears about AI from press headlines, not from internal reporting packs
No data classification work has happened to inform what tools are appropriate for what content
Risk, audit, and legal have not formed a written position on AI use
Investment range and the right next step
AUD investment range: $0 to $50,000 per year. Most of that is licences quietly expensed by department heads. The mistake at this stage is treating a single tool purchase as the strategic move. The right next step is to formalise discovery: name an executive sponsor, fund a two-week scan of where AI would deliver value first, and put a paid Claude team workspace in place so usage is visible.
Stage 2: Pilot
The CIO funds one or two scoped pilots, usually in functions where value is easy to measure: customer service deflection, contract review, sales research, or internal knowledge search. The pilots usually run for 90 days, with a small internal team and an external consultancy. Claude is an approved tool, typically through a paid team workspace and with the Privacy Act implications of pasted content written down for the first time.
Signals you are in Pilot
One named executive sponsor, usually the COO or CFO rather than the CIO
A 90-day scoped pilot with success metrics agreed upfront and a defined kill criterion
A small AI working group with representatives from IT, legal, risk, and the sponsoring function
First conversation about Australian data residency and Privacy Act obligations
First procurement review of an AI vendor with security questionnaires completed
Investment range and the right next step
AUD investment range: $80,000 to $250,000 across the year, including consultancy. The mistake at this stage is letting the pilot live forever as a pilot. Either decide to integrate the result into business as usual, or kill it cleanly and write up what you learned. Australian companies waste serious capital running stage-2 pilots that never graduate, often because no one wants to own the integration risk.
Stage 3: Integration
One or two functions now depend on Claude inside real workflows. Customer support uses Claude to draft replies that an agent reviews and sends. Finance uses Claude to read supplier contracts and flag risky clauses for a lawyer. The AI working group has become an AI council with a charter, and the board sees a quarterly update. APRA-regulated entities and AUSTRAC reporting entities have formal model risk documentation that ties to existing frameworks.
Signals you are in Integration
Two or three production workflows have AI inside them and have run for a quarter
A named AI council reports to the executive committee with a written charter
Vendor risk reviews are completed for primary models, including Anthropic and any others in active use
Privacy Impact Assessments are completed for any workflow that touches personal information
A defined incident response path exists for when a model produces a harmful or non-compliant output
Investment range and the right next step
AUD investment range: $400,000 to $1.2M per year. The mistake at this stage is treating AI integration as an IT project. It is not. It is an operating model change. The CFO and the COO should own the budget jointly, with the CIO accountable for platform and security. The next step is a deliberate decision about how to consolidate tooling and start building a small central platform capability.
Stage 4: Scale
AI is now in five to fifteen workflows across the business. The company runs a central platform team that manages access, monitoring, a shared prompt library, and an evaluation harness. Business units pay an internal chargeback for AI usage so cost lands where the value lands. The board has approved an AI policy that references the Australian Voluntary AI Safety Standard and any sector codes that apply.
Signals you are in Scale
A centralised AI platform team of four to twelve people in a typical AU mid-market shape
Internal chargeback or showback so every business unit sees its own AI spend each month
A model evaluation harness that runs before any prompt or agent change ships to production
Retention and audit logging policies aligned with Privacy Act and sector record-keeping expectations
Sector-specific compliance live: CPS 230 for APRA entities, ISM controls for public sector, Aged Care or Health Records Act provisions where they apply
Investment range and the right next step
AUD investment range: $1.8M to $6M per year. The mistake at this stage is letting the platform team become a bottleneck. The platform exists to make the safe path the easy path, not to act as a gatekeeper that slows every team down. The next step is to push routine builds back out to the business units while the platform team focuses on shared infrastructure and high-risk reviews.
Stage 5: AI-Native
AI is load-bearing for how the company operates. New products are designed assuming Claude or similar agents are part of the user experience. New hires are onboarded with an AI fluency expectation. Promotions reflect demonstrated ability to direct agentic work and to evaluate model output critically. The board uses AI-assisted briefing packs ahead of meetings, and the audit committee receives independent third-party assurance over AI controls.
Signals you are in AI-Native
Product and service design assumes AI in the delivery loop from the first whiteboard sketch
Hiring, performance, and promotion frameworks explicitly account for AI fluency at every level
The board receives independent third-party assurance over AI risk and operating controls each year
New business cases are written with an AI baseline rather than an AI add-on
The company contributes to AU public consultations on AI policy and sector codes of practice
Investment range and the right posture
AUD investment range: $8M and above per year, often split across product, platform, and operating cost. There are perhaps 30 companies in Australia we would describe as AI-native in 2026, and the count is rising fast. The mistake at this stage is assuming you have arrived. AI-native is a posture, not a destination, and the model frontier moves under your feet every six months.
How to read this model as an Australian board
Three questions sharpen the conversation in your next board meeting. First: which stage are we in right now, with what evidence? Second: what is the cost of staying in this stage for another year, both in missed value and in risk that builds up under the surface? Third: what is the minimum credible investment to move to the next stage, and who owns it across the executive team?
If you can answer those three questions with shared numbers, you have a working AI strategy. If you cannot, the maturity model has just done its job by exposing the gap.
We work alongside Australian boards and executive teams on exactly this conversation. If a 45-minute structured discussion with a Claude specialist would help your team locate its stage and choose the next investment, you can book a brainstorm session.
