AI Developers for Australian Businesses
We build AI agents, assistants and automations that plug into your existing systems. We build on Claude, ChatGPT, AWS Bedrock or Azure AI Foundry.
AI development services
We connect AI to the systems and data your business already uses, with access controls, testing and cost limits built in. Our AI development services include:
AI assistants for your business data
Ask questions in plain English and get answers from your own systems, with the same numbers as your dashboards. The assistant is read-only and limited to what each user is allowed to see.
Connect Claude or ChatGPT to your systems
We build secure connectors (MCP servers) so your team's AI tools can look up records and run reports in your software. Staff sign in with their existing Microsoft or Google account.
Private AI for your team, hosted in Australia
Roll out Claude to staff on your own AWS account. No data leaves Australia, each user signs in individually, and spending is capped and audited.
Document and data extraction
Turn contracts, spreadsheets, PDFs and photos into structured data. Fewer forms to fill in and less re-keying.
AI features for your product
Add drafting, summarising, explaining and smart intake to your SaaS or app. Where it matters, a person approves the result.
Customer chat and intake agents
A chat assistant works out what the customer needs, asks the right follow-up questions, and hands your team a request that's ready to act on.
Search and Q&A across your documents
Ask questions across thousands or millions of pages and get answers that cite the source.
AI MVPs for startups
Take an AI product from idea to paying customers. View an example startup AI product we built, AusPol AI.
Need a custom machine-learning model? See Machine Learning. Want ChatGPT and other AI search tools to recommend your business? See AEO.
AI solutions that we've built
-
Private-credit lender An AI assistant and Claude connector for the loan book
Staff wanted fast answers about loans, investors and balances without waiting for a report. We built an in-app assistant, plus a secure connector that lets staff use Claude directly with their own Microsoft login. It can produce IRR, LVR and funding reports on request. Answers use the same calculations as the dashboard, and all processing stays in Australia.
-
Construction procurement marketplace An AI buyer assistant
Buyers describe what they need, or upload a bill of materials or a photo. The assistant checks whether the marketplace can supply it, asks for missing details, and passes a ready-to-quote request to the team. Before choosing a model, we tested the candidates against sample buyer requests. Spend caps and abuse protection keep running costs predictable.
-
International education marketplace Contract data extraction
Agencies and institutions upload commission contracts as PDFs. AI reads each one and pre-fills the parties, territories, commission terms and dates, including contracts that cover several institutions. Processing runs in Australia for Australian users and in the EU for European users. Contract pre-fill has been live since September 2025, and the platform has used AI on AWS in Sydney since April 2024.
-
Decision-support software AI inside an established SaaS product
The product helps organisations prioritise investments. We added AI that drafts a first value model from a short brief, rewords comparison questions so they're neutral, and explains inconsistent answers in plain English. A person reviews the AI's work before it's used.
-
AusPol AI Hansard podcasts and cited Q&A
Every sitting day, AusPol AI turns Australian parliamentary transcripts into podcasts and personalised briefings. Users can ask questions across Hansard back to 1901 and get answers with citations, drawn from more than 4.4 million indexed passages. It has subscriptions, team accounts and usage-based billing.
Keep your data in Australia
Most of our AI solutions run on Claude through AWS Bedrock or Azure AI Foundry in Sydney and Melbourne. Your prompts and documents are processed in Australia, and AWS doesn't use them to train models. Where you need it, we lock this down with AWS policies that block AI use outside Australia.
We can also run EU deployments for European customers, or in other regions where data sovereignty is critical.
- AI sees only what the signed-in user is allowed to see.
- It is read-only by default, and any change needs a person to confirm it.
- Text from customers and documents is treated as untrusted, which protects against prompt injection.
- We build in spend caps, rate limits and budget alerts.
- Per-user sign-in gives you an audit trail of who used AI and when.
Why Launch Lab
- Experience. While AI is still nascent for many companies, our AI work is live in client products, and our work in this space is accelerating.
- A senior Australian team. Based in Sydney and Canberra, with 11+ years of experience. We've shipped 50+ startups as well as solutions for Government, Universities and large organisations. We also work with multiple Fortune 500 companies alongside our clients.
- Deep ML roots. Our CTO, Dave Pretty, started working with neural networks during his PhD research more than 20 years ago.
- AI-led delivery. We use AI coding agents every day, which keeps builds fast and affordable.
Some of the tech we use
- Claude
- ChatGPT
- AWS Bedrock
- Azure AI Foundry
- Bedrock Knowledge Bases
- MCP (Model Context Protocol)
- pgvector
- Python
- TypeScript
- AWS Lambda and Step Functions
- TensorFlow
- PyTorch
- scikit-learn
After meeting with several development teams, it was clear that Launch Lab not only understood our vision but could also deliver a top-quality product. Launch Lab consistently delivers high-quality work and is responsive to any issues that arise. They communicate directly and don't waste time.
I had the pleasure of working with Shaun and the Launch Lab team for 2 years in my time as Head of Marketing at Immutable. I highly recommend Shaun and his team, I can attest to the fact that he will always have the client's best interests in mind and give his honest thoughts and ideas, even if it is at the expense of banking less hours on his end.
Launch Lab were a pleasure to work with, extremely competent and have provided a great outcome for this project. Project delivery was professional, efficient and effective.
Launch Lab are extremely competent and professional and are also pleasure to work with. I highly recommend Launch Lab!
Amazing team, really helped our business in the early stages. Shaun and David were great to work with. 10/10. Safe pair of hands and very cost effective.
It's the second time I've worked with them on launching a website (for different companies). When I got the opportunity to build a new website, my first thought was 'speak to Shaun'. I trust their recommendations completely and their work is of the highest quality. From beginning to end everything has gone smoothy and professionally. Highly recommend!
Artificial Intelligence development FAQs
What AI development services do you offer?
We build AI assistants over your business data, connectors that let Claude or ChatGPT work with your systems, document and data extraction, customer chat and intake agents, and AI features inside existing products. We also run private, Australian-hosted Claude rollouts for staff, and build custom machine-learning models where a general AI model isn't the right fit.
If you're unsure if your requirements fit within these services, please reach out to us and we can let you know if our skills and experience are a good fit for your project.
Do you build AI agents?
Yes. Our agents use tools to look things up, run reports and pass work to your team: for example, a loan-book assistant for a private-credit lender and a buyer assistant for a construction marketplace. They are read-only by default, see only what the signed-in user can see, and need a person to confirm any change.
Is our data kept in Australia?
Yes. We build on AWS Bedrock or Azure AI Foundry in Sydney and Melbourne, so prompts and documents are processed in Australia. AWS doesn't use them to train models. Where needed, we add AWS policies that block AI use outside Australia. For European customers we can run the same setup in the EU. If a feature needs a model that isn't available in Australia, we'll tell you before using it.
Which AI models do you use?
Mostly Anthropic's Claude, plus OpenAI and Amazon's own models where they fit better. We pick models by testing them on your real questions for accuracy, speed and cost, and re-test when a new model comes out. You aren't locked in to one provider.
Can you connect AI to our existing software?
Yes, and that's most of our work. We build secure connectors (MCP servers) and API integrations so AI can read your data and act within your rules. Staff sign in with their existing Microsoft or Google accounts, so their AI tools only see what they're allowed to.
How do you stop the AI making things up?
We ground answers in your own data, show citations back to the source, and reuse your system's own calculations instead of letting the model do the maths. Before launch we test against real questions with blind grading. For anything that changes data, a person approves the action.
How much does AI development cost?
It depends on how many systems are involved and the unique requirements of your project. We're happy to provide a development cost and the running cost per use on your data, so you know the full cost before committing to a build.
How long does it take?
Usually weeks, not months. Our loan-book assistant for a private-credit lender went from first line of code to live (as a beta) in about three weeks.
Do you work with startups?
Yes. We've helped 50+ startups ship products, and we run our own AI startup, AusPol AI. We can take an AI product from prototype to paying customers, including subscriptions, usage-based billing and cost controls.
Are your AI developers based in Australia?
Yes. Our team is in Sydney and Canberra, and we don't offshore development. You work directly with the senior people building your system.
Have an AI idea, or a process that needs automating?
Tell us what you'd like AI to do. We'll come back to you on whether it's feasible, the likely approach and a rough cost.