AI workflow automation with n8n, Make, and Zapier
Real sequences, one per function, each showing where AI does the reading and judging and where fixed rules take over. Some of this will sound familiar.
n8n, Make, and Zapier — and how we choose
n8n
Self-hosted or cloud. The right choice when everything must run on infrastructure you control, when volumes make per-task pricing expensive, or when a workflow needs logic that visual builders handle awkwardly. We set it up on your own server where that matters — common in legal, healthcare, and finance.
Make
Quick to build and readable by your team afterward. Strong for multi-step sequences with branching, and good value at moderate volume. Often the right balance when nobody internally will maintain code.
Zapier
The widest app coverage of the three. Where a system you depend on connects to only one platform, it is usually this one. Simple to hand over.
How we choose
On what you already run, the volume involved, where your data is allowed to be processed, and who will maintain it after handover. If you already use one, we build there rather than moving you. If you are on the wrong one for what you are trying to do, we will say so — and migration between them is work we do.
What we add on top
These platforms move information between systems reliably. They do not read a supplier's email and work out what it means. That is the AI step, built with Claude or OpenAI, sitting inside the sequence at the point where judgment about language is genuinely required.
The processes we automate
Each is a real sequence, not a task — with the step that needs AI named, the platform it typically runs on, and the tier it maps to.
The technology we build with
For the AI steps
Claude (Anthropic) and OpenAI models. Selected per workflow on accuracy for the specific task, cost per run, and where the data is processed.
For connecting systems
Make, n8n, or direct connections. n8n where a business wants everything running on infrastructure it controls.
Systems we commonly connect
| Category | Examples |
|---|---|
| Email and documents | Microsoft 365, Google Workspace, SharePoint, Dropbox |
| CRM | HubSpot, Pipedrive, Salesforce, Zoho |
| Project and work management | Asana, ClickUp, Monday, Jira, Trello |
| Finance | Xero, QuickBooks, Sage, FreeAgent |
| Data and records | Airtable, Notion, SQL databases, Google Sheets |
| Communication | Slack, Teams, WhatsApp Business |
| Industry systems | Practice management, applicant tracking, policy administration, and job management systems, connected where a documented connection exists |
Where your data goes
Named per workflow before the build starts, in writing. Processing region, retention, and whether anything is used for model training — the answer to that last one is no, and it is confirmed in the contract.
You keep your systems
Nothing moves to a platform of ours, and nothing runs on a credential only we hold.
AI workflow examples
Five systems we designed and built, each showing the sequence, the step where AI is required, the systems involved, and what happens when it is unsure.
Which of these is costing you the most?
If one of the above took a name and a face as you read it — the person who does it, the morning it takes — that is the place to start. Tell us which one, and we will give you an honest view on whether AI is the right answer or whether something simpler would do.
Half an hour, no preparation, no obligation.
Book a 30-minute call