The afternoon job that runs itself — on the tools you already have.
An n8n workflow is a node graph, not a black box. A trigger node — a webhook, a schedule, or a native app trigger like the HubSpot or Gmail trigger — starts the run. Everything after it is a node you can open and read: an HTTP Request node calling an internal API, a Set node reshaping a payload into the fields your CRM expects, an IF or Switch node branching on a condition. AutoJet builds these graphs to do the specific job someone on your team does by hand today — pull a lead from three sources, score it, write it to the CRM, notify the right owner — except at 2am, without skipping a step.
Most workflows we ship have more than one path. Branching lives in an IF or Switch node: a score above threshold routes one way, a missing field routes to a human-review step instead of failing the whole run. Transient failures — an API that times out, a rate limit — hit the node’s built-in retry with a configurable backoff before we ever call it broken. What retries can’t fix trips the workflow’s Error Trigger, which posts the node name, the error, and a link to the exact failed execution to Slack or email. Debugging starts from that link, not a guess about which of forty runs went wrong.
Before any of that gets built, we spend time on which workflow actually costs you the most hours — not the loudest complaint from this week’s standup. That’s usually a different first build than anyone expected, and it’s the one we scope.
Manual vs n8n, in practice
Someone copies, chases, and remembers. Works until they’re out, overloaded, or the sheet drifts from reality.
A node graph you can open and read: trigger, branch on an IF node, action, built-in retry on failure. Runs on a schedule or a webhook, day or night. A node that fails trips the workflow’s Error Trigger instead of failing silent.
We don’t replace your CRM. We build the missing ops layer between the tools you already run.
How a workflow ships
Discovery → fixed scope → wire live tools → guard failures → hand off ownership.
- 01MapThe real path, not the org chart
- 02ScopeFixed price before build
- 03WireCRM, inbox, finance, ops
- 04GuardRetry-on-fail, Error Trigger, human-review branch
- 05HandoffExports + docs, yours
Connects to the stack you already run
Typical glue — not a migration project. If it has an API, we can usually talk to it.
Anatomy of a solid workflow
Every build we ship has an IF or Switch node driving the branch, and a wired Error Trigger workflow watching for the fail path — not a workflow that quietly stops when an API times out.
Set node writes the result · status update · notify owner · exit
Node retry fires first · unrecovered errors trip the Error Trigger → Slack/email with the execution link
When this makes sense
n8n speaks HTTP. If a tool exposes a REST API, or even just accepts a webhook, we can wire it in — CRM, finance tools, Slack, a plain Postgres database. Self-hosted n8n or n8n Cloud, whichever your security team signs off on. Credentials live in n8n’s built-in credential store, scoped to the node that needs them — not pasted into a Set node as plaintext.
Where a step needs judgment instead of a rule — draft a first reply, decide if a message is urgent — we drop a Claude skill into a single node that calls the model. The workflow stays the backbone. The model only touches the one decision that actually needed judgment, not the whole path.
Workflows fail loudly by design. A broken node trips the workflow’s Error Trigger, wired to post the failure straight to Slack or email with a link to the specific execution. That’s the gap between a workflow your team trusts and one that quietly drops a lead for two weeks before anyone notices the pipeline went quiet.
From our work
Leads arrived from a form, LinkedIn, and three inboxes. Whoever saw one first replied — hot prospects waited two days. We shipped one path: score against ideal profile, route with context, follow up until a human took over. The weekly “who owns this?” meeting had nothing left to do.
Full case study: Stop losing hot leads to Slack and a Friday spreadsheet →This sounds like you if
- Someone copies between two tools every morning by hand
- Leads sit a day or two before anyone follows up
- Invoices get chased by memory, not status
- A request has to pass three people before anything moves
- You’ve outgrown simple Zapier chains but don’t need a full-time automation hire
- A “automation” got muted because it couldn’t handle exceptions
What you get
- Fixed-price proposal after one discovery call
- Workflow built and tested on your real tools
- Docs your team can follow
- Full ownership: exports, credentials, data
- Error handling and retries — fail loud, not silent
- Optional retainer once the first system is live
Pricing
$300 starting, fixed after a discovery call
Need more than one build? See the Custom plan for multi-system work.
Questions about n8n workflows
No. Most clients haven’t. We set it up, connect tools, and ship a working system — not a course.
Zapier is one trigger to one action, run in Zapier’s own execution environment. n8n workflows are a node graph you can open, branch with IF/Switch nodes, and self-host if that’s a requirement — so the retry logic, the credentials, and the Error Trigger wiring are all inspectable, not hidden inside someone else’s platform.
Workflows fail loudly: a broken node trips the Error Trigger and pings Slack or email with a link to the exact execution. On retainer we fix it. Off retainer, you have the workflow export and docs for any n8n-capable developer to patch.
Price locked before build. One person scopes and ships. Real handoff — source included — not an open hourly meter.
Yes. We’ll say if extending or rebuilding is faster — we don’t default to a rewrite for our convenience.
Only where judgment belongs. Rules stay in n8n. If a task doesn’t need a model, we’ll say so — a script is faster and cheaper.
Want this scoped for your stack?
Fixed-price proposal after a short discovery call — no open-ended retainers to start.