Security & data
Your ERP data never has to leave your environment
How we deploy AI on ERP data so that finance, IT and compliance can say yes: where data lives, which models are used, what AI may change and how every action is recorded.
Deployment options
| Option | Where the AI runs | Where your data goes | Best for |
|---|---|---|---|
| On-premise | Your own servers, next to your ERP | Nowhere outside your network | Strict data policies, legacy ERPs, no-internet environments |
| Private VPC | Your own cloud account, in a region you choose | Stays inside your cloud tenancy | Cloud-first companies with data residency rules |
| Cloud | Managed cloud services in a region you choose | Processed under business API terms, not used to train public models | Fastest start, lowest infrastructure effort |
In-region hosting is available for the US, the EU, the UAE and Australia.
Controls
Built in, not bolted on.
Human approval on every write
Agents read within their permissions but never change money, stock, master data or customer records without a person approving. Approval rules and limits follow your existing controls.
Role-based access
Each agent gets only the permissions it needs, and acts as the logged-in user where possible, so your ERP roles and segregation of duties still apply.
Audit logs
Every action is logged: who asked, what the agent read, what it proposed, who approved it and what changed. Logs are stored in your environment.
Data retention
Prompts, documents and logs are kept only as long as your policy says, then deleted. Personal data is excluded or masked wherever the task allows.
LLM options
Commercial models under business API terms, or private, self-hosted models running on your own hardware when data must never leave your network. We confirm each provider's terms in your project.
Security documents on request
Before any data is shared you receive our data processing agreement (DPA) to review, the list of sub-processors for your deployment (which cloud and which AI model provider, if any) and a data-flow diagram of the agent.
NDA on day one
We sign your NDA, or ours, before discovery starts, and a data processing agreement where you need one. Only named engineers access your data, and access is removed at handover.
GDPR and DSGVO approach
For clients in Germany and the EU we design for GDPR (DSGVO) from the start: data minimization, a documented purpose for each data flow, EU hosting, a data processing agreement, and the records your data protection officer needs. Where the EU AI Act applies, we document human oversight and transparency measures.
For the US, UAE and Australia we account for state privacy laws such as the CCPA, the UAE Personal Data Protection Law and the Australian Privacy Act and Privacy Principles.
General information, not legal advice. We work with your adviser on specific requirements.
Certifications and roadmap
We do not hold SOC 2 or ISO 27001 certification at this time. Because we can deploy entirely inside your own environment, your existing certified controls apply to the AI system too.
You own the code, prompts, configurations and cloud accounts, with documentation at handover.
FAQ
Security questions.
Can the AI run without internet access?
Yes. With private models on your own hardware, an agent can run entirely inside your network with no external AI calls.
Is our data used to train AI models?
No. We use business API terms under which your data is not used to train public models, or self-hosted models that never send data out.
Who can see our data during the project?
Only the named engineers on your project, under NDA, with access removed at handover.
Can an agent change ERP data on its own?
No. Every write goes through an approval step and is logged. You can switch any agent to read-only at any time.
Do you hold SOC 2 or ISO 27001?
Not at this time. We can deploy inside your own environment so your certified controls apply.
Works with leading ERPs, CRMs and helpdesks — cloud, hybrid or on-premise.
ERPCRMHelpdeskEmail inboxDocument storageDatabases & data warehousesCloud, hybrid or on-premiseSee an AI agent running on your ERP data in 4–6 weeks.
Fixed scope · fixed price. Human approval on every write. We reply within 1 US business day.