AI for business · Private AI hosting

The same AI models your staff already use, hosted in Australia.

By default, the AI tools people reach for process your data overseas, on personal accounts, with no record of what went in. We set up Claude from Anthropic and OpenAI models in a dedicated AWS account in Sydney that we manage for you, through Amazon Bedrock, with single sign-on, access rules, logging and guardrails, then keep it running. Your data stays in Australia, stays yours, and never trains anyone's model.

Talk about a secure AI setup How it works
  • Australian regionModels served through Amazon Bedrock in AWS Sydney (ap-southeast-2). Prompts and files are processed onshore.
  • Managed by CodiumA dedicated AWS account for your organisation. We set it up, secure it, run it and bill it through us.
  • No training on your dataBedrock does not use your prompts or outputs to train models, and neither Anthropic nor OpenAI ever sees them.
  • SSO, rules and logsSingle sign-on from Microsoft 365 or Google, access rules per team, and a log of who used what.
Why a private setup?

Your staff are already using AI. The question is where your data goes when they do.

Most organisations handling client files, health records, legal matters or government work have quietly ended up with two policies: the official one that says don't, and the real one where people paste things into a personal ChatGPT or Claude account because it saves an hour. A hosted setup gives them the same models, inside your boundary, with the controls you'd expect of any other business system.

"Can I use AI on this client's file?"

  • ×Data processed overseas under a consumer agreement

  • ×Each person's own login, their own history, no oversight

  • ×No record of what was shared or what came back

  • ×"Don't use it" as the policy, used anyway

Data residency you can state

Prompts, files and responses are processed in the Sydney region. When a client, auditor or regulator asks where their data went, there's a plain answer.

The same Claude and OpenAI models

Claude from Anthropic and OpenAI models, served through Amazon Bedrock. The same capability as the public tools, under AWS's enterprise terms rather than a consumer one.

Controls you already understand

Sign in with the accounts your staff already have. Decide which teams can use it and for what. Remove access the day someone leaves, like any other system.

Run by a hosting team

Patching, monitoring, cost control and model updates are handled by the same team that runs our managed AWS hosting, so it stays secure after launch day.

What's included

A complete setup, managed for you after launch

We provision it, lock it down, bring your people onto it and keep it running. You get a tool your staff can use for the client and internal work they've been holding back on.

01

Hosted in Australia, in a dedicated AWS account

A dedicated AWS account for your organisation in the Sydney region, provisioned, secured and managed by us, with Amazon Bedrock serving the models inside it. The data in it is yours. Nothing is shared with other customers, and nothing leaves the region to be processed.

How we run AWS environments
02

Claude and OpenAI models

Anthropic's Claude family and OpenAI models available through Bedrock, chosen per use case. Change or add models later without rebuilding anything.

03

Single sign-on and access rules

Microsoft 365 or Google Workspace login, access by group, and permissions that follow your organisation chart.

04

Logging and monitoring

Who used it, when, and for what, retained to your policy. Usage and cost dashboards so there are no surprises on the bill.

05

Guardrails and policies

Bedrock guardrails and system policies that set what the tool may be used for, filter sensitive content, and block the uses your sector forbids.

06

Staff onboarding and ongoing management

Hands-on sessions so your people know what it's for and what it isn't, plus the patching, model updates, monitoring and support that keep it that way.

Pair it with a support SLA

Want the models inside your own software, not just a chat?

The same hosted endpoint can power AI features in the systems you already run: reading bills, answering from your data, drafting from your records. That's our AI for business work, and it runs on this foundation.

AI for business Need it on your own hardware instead?
How it works

From a conversation to a tool your staff can use, in weeks

There's no software to build. The work is in setting it up correctly for your organisation, bringing your people onto it, and running it well afterwards.

01

Scope

A short conversation about who will use it, what for, which data is involved and what your sector requires. We agree the models, the policies and the access groups.

02

Provision

We set up a dedicated AWS account for you in Sydney, enable the models in Bedrock, and configure networking, encryption, logging and retention to match your policy.

03

Connect and guard

Single sign-on is connected to your identity provider, access groups are mapped, and guardrails are written and tested against the uses you've allowed and the ones you haven't.

04

Onboard and run

Your people are shown how to use it well. From then on we monitor, patch, manage costs and keep models current, with a support line when someone has a question.

Under the hood

Amazon Bedrock, in Sydney, in a dedicated account we manage

Bedrock lets AWS serve Anthropic and OpenAI models inside a private cloud boundary set aside for you. The model providers never receive your data, AWS doesn't train on it, and everything is processed in the Australian region you choose.

  • Onshore processing. Models run in ap-southeast-2. Prompts, files and responses don't leave Australia to be processed.
  • Private connectivity. Traffic to Bedrock stays on the AWS network through private endpoints, not the public internet.
  • Encryption and keys. Data encrypted in transit and at rest, with keys that belong to your account alone and are never shared with other customers.
  • Logging to your policy. Invocation logs and audit trails are kept in your dedicated account, retained for as long as your policy says, and shared with you.
  • Guardrails at the platform. Content filters, denied topics and sensitive-information rules enforced by Bedrock, independent of the model.
Our managed AWS hosting
Diagram of Codium's AWS environment with load balancing, containerised services and managed data stores Dedicated AWS account, Sydney region
Who it's for

Organisations where "where did that data go?" has to have an answer

From a ten-person practice to an enterprise with thousands of staff. If your work involves other people's confidential information, this is how you use AI on it.

Legal

Summarise matters, draft correspondence and search precedents with client files that stay under privilege and in the country.

Health

Clinical notes, letters and admin with patient information handled to Australian privacy obligations, not a consumer app's terms.

Financial services

Client analysis, reporting and drafting with the access controls, logging and retention that your compliance team will ask about first.

Government and NFP

Data sovereignty, auditability and policy enforcement for departments and funded organisations that cannot send records offshore.

Good questions

Things people ask before they move AI onshore

Can't see yours? Ask us directly.

Is it really the same Claude and OpenAI models?

Yes. Amazon Bedrock serves Anthropic's Claude models and the OpenAI models AWS makes available, under AWS's terms. The capability is the same as the public tools. What changes is where it runs, who can use it and what gets recorded.

Does our data leave Australia?

Prompts, files and responses are processed in the Sydney region. AWS does not use Bedrock inputs or outputs to train models, and the model providers don't receive them. Logs and any stored conversations stay in your dedicated account. If even that is too far, we also set up local LLM hardware on your premises.

Who owns the AWS account and how is usage billed?

Codium does. We run a dedicated AWS account for your organisation inside our managed environment, so there is nothing for you to set up or administer. Model usage is passed through on your Codium invoice at AWS's published rates with a full breakdown, alongside the management fee, as with our other hosting. The data in the account is yours, and we can hand the account over to you if you ever want to run it in-house.

Can staff use it for anything?

Only what you've allowed. Guardrails set permitted uses, filter sensitive content and block denied topics, and access groups decide who can use which capabilities. The onboarding sessions cover what it's for and where the lines are.

Can it connect to our own systems and documents?

Yes, as a second step. The hosted models become the foundation for AI features in your own software: answering from your data, reading documents, drafting from records. That work is scoped and built separately, with its own permissions, and runs on the same account.

How long does it take?

A standard setup is live in a few weeks: scoping, provisioning, SSO and guardrails, then onboarding. Sector-specific compliance configuration or integration with your identity platform can add time, and we'll tell you up front.

Ready to give your staff AI you can stand behind?

Tell us who would use it and what for. One of our digital strategists will walk through what a hosted setup looks like for your organisation.

Get in touch Back to AI for business