Foundations
A secure platform: the footing that makes automation safe to trust
Agents, knowledge systems and automations are only as trustworthy as the layer they run on. We build that layer: who may do what, what was done, where the data lives, how the systems connect, and how a bad change is undone. It is the same discipline our own products run on.
- Roles for people and agents
- Audit trail and data isolation
- Rollback on any change
In short
The secure AI platform, in our work, is the shared foundation beneath every agent, retrieval system and automation we build for a company: identity and role-based permissions for people and agents, an audit trail of every action, isolation of each organisation’s data, personal data masked where a model does not need it, private deployment where data must not leave, integrations with the ERP, accounting, devices and messaging you already run, versioned prompts and rules, tracing, and rollback. Build one feature on it and the next is an addition, not a rebuild.
What it provides
Identity and permissions set role by role, down to each screen, action and tool an agent may call. An audit log of who or what changed a record and when. Each organisation’s data kept apart from every other, with personal data handled deliberately: masked where a model does not need it, retained only as long as agreed.
Integration services for your ERP, accounting package, biometric devices, WhatsApp Business, telephony and email, so every system reads from and writes to one place. Versioning for prompts and business rules, tracing for every model call, and rollback for any deployed change.
- Role-based access for people and agents
- Audit trail and data isolation
- Personal-data masking and retention rules
- Integration layer for ERP, accounts, devices, messaging and telephony
- Prompt and rule versioning, tracing, rollback
How we build it
We start from the systems and the roles you already have and design the permission model first, because it shapes everything else. The platform runs in the cloud or on your own servers, containerised so the same build runs in testing and production, with tested backups and monitoring that raises an alert before users notice.
AurumOne and the CRM console are built this way: multi-language interfaces, permissions per module and action, versioned rules, audit logs. A platform project usually begins by putting your first agent or automation on this footing, then grows module by module.
- Permission model designed first
- Cloud or on-premises, containerised
- Tested backups and monitoring
- Multi-language interfaces
- Grows module by module
What it connects to
Your ERP and accounting, HR and payroll (including AurumOne), CRM, biometric and other devices, WhatsApp Business, telephony and email, identity providers where you have one, and the model providers or private model servers you choose.
- ERP, accounting and CRM
- HR, payroll and devices
- WhatsApp Business, telephony and email
- Model providers or private model servers
Where it shows up
Automation playbooks this service runs in
Secure AI Platform
Common questions
Do we need a platform before our first agent?
You need its basics: a role for the agent, logs, and a way to undo a change. We put the first agent or automation on that footing from day one, and the platform grows from there. Nobody should buy a platform before they have a use for it.
Can it run on our own servers?
Yes. Everything we build is containerised and can run in the cloud or on your own infrastructure, with the same backups, monitoring and access controls either way.
How is personal data handled?
Deliberately: each organisation’s data is isolated, models see only the fields a task needs, personal data can be masked before it reaches a model, retention periods are agreed in writing, and the audit log records access. We walk you through it before anything goes live.
Related