Automation infrastructure is the totality of computing and networking resources that host and execute an organisation’s automation capabilities — RPA bot runtime environments, orchestrator management servers, AI model endpoints, integration platform runtimes, document processing services, and the network connectivity between these components and the enterprise systems they automate. Automation infrastructure decisions determine where bots run (on physical servers in the enterprise data centre, on virtual machines in a private cloud, or on cloud-managed compute), how they connect to target systems (directly on the same network, through a gateway, or via API over the internet), how their compute is managed (infrastructure the IT team maintains, or managed cloud services the vendor maintains), and how they scale (provisioning additional VMs when transaction volume increases, or auto-scaling managed cloud compute). These decisions have significant implications for security, data residency, maintenance cost, scalability, and the automation programme’s operational resilience.
Automation Infrastructure Options
| Infrastructure Model | Bot Execution Location | Orchestrator Location | Best For |
|---|---|---|---|
| On-premise (traditional) | Physical or virtual Windows servers in enterprise data centre | On-premise UiPath Orchestrator server | Legacy enterprise environments; SAMA/NCA strict data residency; no cloud connectivity |
| Private cloud (IaaS) | VMs in enterprise’s own OCI, Azure, or AWS tenant | Orchestrator on VM in same private cloud | Cloud-first enterprises with data residency compliance; scalable with cloud VM elasticity |
| Cloud-managed (SaaS orchestrator) | Bot VMs on-premise or in enterprise cloud; orchestrator managed by vendor | UiPath Cloud Orchestrator (SaaS); Power Automate Cloud | Enterprises that want managed orchestrator without infrastructure overhead |
| Fully cloud (cloud robots) | Vendor-managed cloud VMs for bot execution | Vendor cloud orchestrator | Automation targets are cloud-based (Salesforce, Microsoft 365, Oracle Cloud); no on-premise system connectivity required |
GCC Data Residency and Automation Infrastructure
GCC enterprises with data residency requirements — Saudi entities under NCA and SAMA requirements, UAE entities under UAE Federal Data Protection Law — face specific automation infrastructure constraints. When automation bots process financial data from Oracle EBS or ZATCA invoices, that data is present in the bot machine’s memory and potentially in log files during execution. If the bot machine is a cloud VM in a US or European region, financial data resides outside the GCC during processing — potentially non-compliant with data residency requirements. The compliant architecture for SAMA-regulated Saudi enterprises is bot execution on VMs within the Saudi Arabia Azure or OCI region, with the cloud orchestrator either on-premise or in the same region’s cloud, and network connectivity to Oracle EBS over private network (not public internet). For enterprises that mandate on-premise data processing, bot execution on physical servers in the enterprise’s Saudi data centre is the compliant approach.
Bot Machine Sizing and Scaling
Automation infrastructure sizing requires understanding the concurrent bot execution demand — how many bots run simultaneously at peak and how many VM resources each bot process consumes. An unattended RPA bot processing Oracle EBS AP invoices uses approximately 2 vCPU and 4 GB RAM during execution; a Document Understanding bot running computer vision models uses 4 vCPU and 8 GB RAM. For a finance automation programme with 10 concurrent unattended bots at month-end peak, the infrastructure must support 20–40 vCPU and 40–80 GB RAM — a VM sizing requirement that should be documented in the automation infrastructure architecture before procurement. Cloud infrastructure provides elastic scaling capability: additional VMs can be provisioned for month-end peak and deprovisioned after peak — a cost and resource efficiency not available with fixed on-premise server allocation.
What Goes Wrong in Practice
The most common automation infrastructure failure is a production bot deployment on a developer’s workstation that was never migrated to a proper server environment — because the bot was built, tested, and piloted on the developer’s laptop and the migration to production infrastructure was deprioritised. A bot running on a laptop is unavailable when the laptop is off, rebooted for updates, or taken to a meeting. Production automation infrastructure must be on dedicated, always-available server infrastructure — physical, virtual, or cloud — with the same uptime and monitoring requirements as production application infrastructure.
How Loop Wise Solutions Designs Automation Infrastructure
We design automation infrastructure as part of the automation programme architecture — specifying bot execution environments, orchestrator hosting, network connectivity to target systems, data residency compliance, monitoring, and VM sizing — before bot development begins. Infrastructure design is a governance deliverable, not a post-development afterthought.
