AI that can stay in a controlled environment.

Three places the model can run.

Keen Concepts designs AI and automation for Australian businesses, including work where the data cannot go just anywhere.

• A model on hardware you operate, so the job does not call a public model API.
• Microsoft Azure AI Foundry or AWS Bedrock, when you want a managed service inside an account you control.
• Claude, OpenAI or Copilot, when the material involved is allowed to be processed there and that is the straightforward place to start.

These can be mixed. A sensitive record can stay local while a separate, permitted step uses a hosted model. The split should be written down in the design, not discovered later. The longer note is local AI or cloud AI.

One way to keep it on premises.

The AI Box is a compact, Australian-assembled machine for local models and automations. We supply it configured, hardened and supported. The specification, including processor, memory and network, is on the situations page. It is one option, priced per build. It is not a requirement for every private deployment. Hardware you already own can be the right host.

Local equipment still needs a home, power, and someone to apply updates. Ongoing maintenance is part of the engagement for that reason. Handover includes the as-built documentation, the credentials and the keys, as described in the process.

What we need to know first.

• Which documents or records the job uses.
• Whether a provider is allowed to process them, and under which account.
• Whether the job has to run without sending that material off the premises.
• Who will own the environment six months after the pilot.

Cloud-first is often the right pilot. Moving the same job later does not have to mean rebuilding the process, if the process was scoped properly. The automation itself is described under AI automation.