AI workflow automation with n8n, Make and your existing tools
Lead routing, quoting, CRM updates and reporting on autopilot
Every business has work that follows rules and still gets done by hand. Enquiries copied into the CRM. Quotes rebuilt from a price list. Invoice chasers written from scratch. The Monday report assembled from four tabs. None of it needs a person, and all of it is currently costing you one.
Automation that saves hours somebody can name
We start by mapping the work, then automating only what earns it. Most of the build happens in n8n, self-hosted on your server or ours, which takes you off per-task billing entirely. Make is the alternative when a team wants to maintain the flows themselves and prefers the interface.
AI goes in where a rule cannot reach: reading an incoming email and classifying it, pulling six fields out of a supplier PDF that has never had the same layout twice, drafting a reply. Anything that reaches a customer passes through an approval step first, at least until you have watched it behave for a month.
What we do not automate
Work that changes every time. Decisions with real consequences and no clear rule. Anything where the failure costs more than the hours saved. The readiness audit says so plainly, and it is usually a third of what people initially ask for.
After it goes live
Every workflow ships with retries, error handling, alerting and a page of documentation, so it can be maintained by somebody who is not us. The monthly review covers what ran, what failed and why, and what is worth automating next. Where the flows touch customers directly, they usually end up connected to a WhatsApp or CRM agent.
What automation includes
Automation audit
The tasks worth automating, ranked by hours saved per month, alongside the ones that should stay in human hands and why.
n8n / Make builds
Workflows with retries, error branches and alerting, documented well enough for somebody else to maintain them.
AI steps
Classification, field extraction, summarising and drafting inside the flow, with an approval gate before anything leaves the building.
CRM & WhatsApp flows
HubSpot, Zoho, Salesforce and WhatsApp Business connected, with the deduplication rules that stop three records appearing per lead.
Explore →Zapier migration
Rebuilt in self-hosted n8n, usually in fewer steps, which ends the per-task bill that grows every time business improves.
Monitoring & review
Run logs you can read, failure alerts to both sides, and a monthly session on what to build next.
How a workflow gets built
Shadow the taskweek 1
We watch somebody do the job once, end to end, because the written process and the real one are never quite the same.
Map & costweeks 1–2
The flow drawn out with its edge cases, an estimate of hours saved per month, and what it will cost to build and to run.
Buildweeks 2–5
Workflow built with error handling from the start, running alongside the manual process rather than replacing it on day one.
Run in parallelweeks 5–7
Two weeks of the automation and the person doing the same work, comparing outputs, until the differences are boring.
Hand overongoing
Documentation, alerting to your inbox as well as ours, and a monthly review of runs, failures and the next candidate.
From $2,000/project
Builds from $2,000; free 60-minute audit; hosting from $30/month.
Automation, answered
What kinds of work can be automated?
Anything that follows rules and touches systems with an API: lead routing, quoting from a price list, CRM updates, invoice reminders, report assembly, onboarding sequences. The audit lists yours with hours attached, in your own vocabulary.
Why do you use n8n rather than Zapier?
Because n8n can be self-hosted, so the bill does not rise every time your volume does. It also has hundreds of integrations and a code node for the cases nothing covers. We host it for you or put it on your own server, and either way you keep the flows.
Where does the AI part actually fit?
In the steps that need judgement rather than logic. Reading an email and deciding what it is about, extracting fields from a document whose layout keeps changing, drafting a reply for a person to approve. The rest of the workflow is ordinary, reliable plumbing.
What happens when a workflow fails?
It retries, then alerts both you and us with the data preserved so the run can be repeated once the cause is fixed. Silent failure is the real risk with automation, so every flow we build is noisy when something goes wrong.
How much does workflow automation cost?
Builds start at $2,000 for a single workflow, and hosting from $30 a month. The first hour of audit is free, which is usually enough to tell whether the rest is worth paying for.
Can you migrate our existing Zapier setup?
Yes, and it normally comes out smaller. Zaps accumulate steps that exist to work around Zapier rather than to do anything. We rebuild in n8n, test both in parallel, and switch once the outputs match.
Get the repetitive work off your team.
Name the task that eats the most time each week. The first hour of the audit is free, and often enough to price the fix.