Delete the manual. Execute automation.

n8n, GoHighLevel, HubSpot, Keap, Make and Zapier. Each one is a filter on the work page, so you can look at the projects instead of taking the list on trust.

n8n is where the depth is. The others I work in as the job requires, which is a real distinction rather than modesty, and it is worth knowing before you hire someone.

You will not be asked to change platform to suit me. Most of this work happens inside the tools a business already runs, and where a migration genuinely is the answer it becomes its own project with its own risks.

Yes, and one of the case studies here is exactly that. I mapped a live recruitment screening system end to end, audited it, and extended it.

The audit found four delay nodes with no incoming connections, so the intended human-like delay never ran; two email paths that could double-send; and a test spreadsheet wired in as production. None of that was visible from the outside. The system appeared to be working.

Getting out of a platform counts too. There is a migration here off a CRM with no automation export API, where 135 of roughly 178 automations were captured with hand-built tooling and analysed one at a time. Note the shape of that: 135 of roughly 178, not "all of them".

Every case study here ends with a panel headed "What can be verified", and it lists only things that can be checked inside the system itself: artefact counts, test counts, measured technical values.

On the deal-routing build that meant 115 automated tests across five workflows before it was switched on. The tests are as much the deliverable as the workflow is. They are what tells you it still works after the next person edits it.

Systems are built to fail closed, which means that when something goes wrong they stop and say so rather than carrying on with a half-finished result. The expensive failure is the quiet one: a workflow that looks like it ran, and did not. Most of the damage I get called in to undo started that way.

Because none were measured. Every figure on this site is an artefact count or a measured technical value.

No business-outcome metric, whether time saved, revenue or conversion, was captured on these engagements, so none is claimed. A number I cannot show you the derivation of is worth less than saying so plainly.

Statuses are held to the same rule. Where a project says "live", that reflects repository evidence and platform backups, not a check of the system today.

This site is the reference. The only client named anywhere on it is the one who volunteered a testimonial and agreed to be named. Everything else is described by sector.

The workflow screenshots are redacted from a written recipe rather than by hand, so the same crop and the same boxes apply every time and a re-export cannot quietly uncover something. No patient name, email address or phone number appears anywhere here. Several captures I would have liked to show are simply not published, because I could not clear what was visible in them.

The systems themselves run in your accounts, not mine. If the automation lives in my account you are renting access to your own process, and you should be able to end the engagement and still own the thing.