Research | AI Development in Australia | ComplxAI
Research programme

Research

Engineering-led reports on how AI is being built and paid for in Australia. Two were published in September 2026: what a production AI system costs to design, build and run, and where Australian organisations are with AI development. Every figure is cited to a published source or labelled as a ComplxAI indicative range, and each report prints its method and limits beside the findings.

Read the cost report Read the State of AI report
  • Every figure cited or labelled as indicative
  • Written by the engineers who build the systems
  • Method and limits printed with each report
The programme

Research from the build side of the table

Much of what is published about AI in Australia asks people how they feel about it or forecasts what will be spent. Far less describes what a working system costs to design, build and run, or which use cases reach production and which stay in the slide deck.

Both reports are desk research read by engineers. Published pricing and market data are gathered, cited and read against each other: provider price lists, the AWS Price List for the Sydney region (the AWS region for Australian data residency; systems can be deployed in whichever region you need), the RBA exchange rate table, salary guides, ABS and National AI Centre surveys, and the Australian cuts of the global consultancy surveys. Where no Australian source publishes a figure, as for what an individual AI system costs to build, the gap is filled with a ComplxAI indicative range from the pre-Discovery model behind our AI project cost calculator, and labelled as such.

Ground rules for every report

  • Every figure carries a numbered reference or is labelled a ComplxAI indicative range
  • Indicative ranges are rounded, pre-Discovery and never a quote
  • Method, sources and limitations published with the findings
  • No figure derived from client data; no engagement counts or survey results of our own
Reports

Two reports for 2026

Both published in September 2026, each with its method, sources and limits set out beside the findings.

Report · published September 2026

Australian AI development cost report 2026

What a production AI system costs to design, build and run in Australia, by type of system rather than as one figure: internal assistants, RAG knowledge systems, agents, AI features added to an existing product, AI-native SaaS and enterprise platforms. Each type carries a ComplxAI indicative range, a typical timeline and its drivers. Model API and AWS Sydney prices are converted to Australian dollars and cited, and monthly model and infrastructure costs are shown apart from the build, with upkeep, talent costs and how to stage the spend.

Report · published September 2026

State of AI development in Australia 2026

Where Australian organisations are with AI in 2026: how many use it and how deeply, what is being built, how the build, buy or defer call is made, which models can be run from Australia, who does the work, what it costs, what stops projects and how the rules are changing. Every figure comes from a published source and is cited; ComplxAI's own observations are qualitative and carry no numbers. Written for whoever makes that decision next.

Methodology

How the reports were put together

The same method for both, set out in full in each report's methodology section.

  1. GatherPublished pricing and market data: provider price lists, the AWS Price List for the Sydney region, the RBA exchange rate table, salary guides, ABS and National AI Centre surveys, and the Australian cuts of global surveys.
  2. CiteEvery figure carries a numbered reference to a source accessed on 16 September 2026. US-dollar prices are converted to Australian dollars at the RBA rate and dated.
  3. Fill the gapsWhere no Australian source publishes a figure, a ComplxAI indicative range from the model behind our cost calculator stands in, rounded and labelled as such.
  4. Read as engineersThe sources are read against what it takes to design, build and run these systems. Where ComplxAI adds a view, it is a qualitative observation and carries no numbers.
  5. PublishMethod, sources and limitations printed with every finding.

Three limits are worth stating here. Australian cuts of global surveys are small, so where a source states its Australian sample size the report repeats it and otherwise treats the figure as indicative. Prices and exchange rates move, so every converted figure is a snapshot. And the indicative ranges are a model, not measured engagement data: no organisation was interviewed for either report, and no figure is derived from client data.

Alongside the reports

From a published range to your own project

The reports give ranges and drivers. These take the same questions to your own numbers.

AI project cost calculator

The model behind the cost report's indicative ranges, applied to your project's shape, integrations and stages.

Estimate a project

ROI calculator

What an automation is worth before you commission it.

Estimate your return

Case studies

Client-reported outcomes from systems we have built.

See what we've built

Shorter reads sit in insights; the five-stage model that turns an indicative range into a written price is in how we work.

After reading

Talk through what the numbers mean for your project

The reports give ranges and drivers; Discovery turns them into a written price for your own project. The first two conversations are free: a 30-minute intro call and a one-hour pre-discovery session. A question about a source or a figure can go to info@complxai.com.

Book an AI strategy call Request a proposal