How we deliver AI software projects
Six clear stages from first conversation to production deployment. No mystery, no scope creep, and a working prototype before you commit to a full build.
The six stages
Every engagement follows this sequence. Some stages overlap, and timelines vary by project size, but the order stays the same.
Discovery call
A 30-minute video call where you describe the problem, the data you have, and the outcome you want. We ask about your current tech stack, team size, and any compliance requirements. There is no cost for this call, and we'll tell you honestly if AI isn't the right tool.
After the call we send a one-page summary confirming what we discussed, along with an initial feasibility assessment. If the project looks viable, we move to a scoping workshop.
Scoping workshop
A half-day session (remote or in person) where we dig into the data. You share sample datasets, database schemas, or API documentation. We map the data pipeline from source to prediction or action.
The output is a scope document that specifies the model type, input features, target variable, success metric (e.g. F1 score above 0.85, or mean absolute error below 3%), integration points, and estimated timeline. This document becomes the contract's technical annex.
Proof of concept
We build a working prototype in two to four weeks using a representative subset of your data. The prototype runs in a notebook or a lightweight web app so your team can interact with it directly.
At the end of this stage you see real predictions or real automated outputs, not slides. If the results don't meet the agreed success metric, you can walk away. The proof-of-concept fee is fixed and quoted in the scope document.
Production build
Once the prototype is approved, we harden the code for production. That means containerised services, automated tests, CI/CD pipelines, logging, and error handling. We connect the model to your systems via REST API, webhook, or direct database integration.
Depending on complexity, the production build takes four to twelve weeks. You receive weekly progress updates with a demo of new functionality each Friday.
Launch and handover
We deploy to your chosen environment: AWS Sydney, Azure Australia East, GCP Sydney, or your own on-premise servers. We run the system in shadow mode alongside your existing process for one to two weeks so you can compare outputs before switching over.
Your team receives documentation, runbooks, and a recorded walkthrough of the codebase. If you have developers who will maintain the system, we pair with them during this phase.
Monitoring and retraining
Models degrade over time as data distributions shift. We set up automated monitoring dashboards that track accuracy, latency, and data drift. When metrics fall below threshold, we retrain the model on fresh data and redeploy.
Monitoring is available as a monthly retainer or on a per-incident basis. Most clients choose the retainer because it includes quarterly model reviews and priority support.
What you can expect from us
Two commitments we make on every project, and the tools and practices behind them.
Transparency
You own the code and the trained model weights. We use Git repositories hosted under your organisation's account, so you have full visibility from day one.
- Weekly written status reports
- Live demo every Friday
- Shared Kanban board with task-level estimates
- No lock-in: you can bring in another team at any point
Data security
Your data never leaves the agreed infrastructure boundary. We sign NDAs before the scoping workshop and can work within your VPN if required.
- Encrypted at rest and in transit
- Role-based access control on all environments
- Audit logs for every data access event
- Compliant with the Australian Privacy Principles