Automation Build.

Built to run without us.

A build is finished when you can run it, hand it on, or hand it to someone else and the work continues. Five to six weeks for a single workflow, broken into five phases. The running system, the source configurations, the documentation, and the source code all sit with you when we are done.

How a build is shaped

Phase one is scoping and design. We agree what the automation will do, where it sits in the existing process, what triggers it, what it touches, and what “done” looks like. Edge cases get named in writing now, not discovered on launch day. The design is signed off before any infrastructure work begins.

Phase two is infrastructure setup. Self-hosted n8n on your own server, NocoDB on the same infrastructure for tabular data, your own database where you need one, your own API keys. No automation runs on a Perkins SmartOps account where the system would go dark the moment you stopped working with us. Where personal data is involved (customer records, employee records, financial data, anything that identifies a real person), we document the processor relationships, the retention behaviour, and the data flow in writing as part of this phase. That record is the input your data-protection assessment needs. We are not data protection specialists, so the assessment itself sits with you or your adviser.

Phase three is the build. Workflows are built incrementally with checkpoints where you can see the state of the work, raise concerns, and redirect if the design needs adjustment. The approach aligns to internal audit and process optimisation discipline: assume any step can fail, design what happens when it does, and rank every control by what its absence would cost.

Phase four is testing. The automation is run against real data, real edge cases, and the failure scenarios named in scoping. Test results are documented. Anything that does not behave as designed is fixed before handover, not after.

Phase five is handover. A team walkthrough, written technical documentation covering every node, the source configurations exported and stored where you can reach them, and a short post-launch support window for any issues that surface in live running.

Why ownership matters

Most automation that goes wrong in the wild goes wrong because the people who could fix it have left or can no longer be reached. The platform is fine. The credentials are fine. Nobody knows what the automation was supposed to do, where the source lives, or how to make the change the business now needs. Months of work becomes a black box.

We avoid that by treating ownership as a deliverable, not a courtesy. The infrastructure is yours, the configurations are yours, the documentation is yours, and the build follows conventions any competent n8n developer could pick up and extend. If you stop working with us, the system keeps running. If something later needs changing, you have the source, the design notes, and someone else can do the work. This is part of why we recommend self-hosted n8n over closed platforms where the workflow would live behind a vendor account: the risk of an automation provider going bust is not theoretical.

What happens after handover

Three paths after the short post-launch support window. You run it yourself, with the documentation and source. You take a Managed Automation arrangement, where we monitor, maintain, and respond to issues on the automation you continue to own and host, with a fixed response window when something needs changing. Or the build becomes the foundation for an Ongoing Partnership, where each new workflow is cheaper than the last because the context, the access, and the design language are already in place.

If the build inherits an existing automation stack, a Health Check usually runs first. We want to know what is already there, what is working, and what would conflict, before adding a new layer to a system whose history nobody can fully reconstruct. We are independent AI and automation specialists. No vendor pays us for the platform recommendation, so the choice of where the build runs is the choice that fits the work, not the choice that pays us.