Our work

Proof we build, not just pitch.

Real n8n + Claude systems we’ve shipped, each one you’d own outright. The missed-call recovery and lead follow-up we build for your front desk runs on the same patterns you see here.

Content & SEO

An evergreen content & SEO engine

The problem. Publishing consistently is the first thing that slips when you get busy. The blog goes quiet, and the inbound it used to bring in quietly dries up.

What we built. A multi-stage pipeline that pulls topic angles from live sources, scores each one for real search demand, drafts the post with images, and queues distribution, so a content calendar keeps running whether or not anyone has the time.

  • n8n
  • Claude
  • Firecrawl
  • Google Sheets

Built on tools you own, with a human approving the queue before anything publishes.

AI content engineautomation workflow

Daily schedule reads industry sources

Trigger

1

Scout topics

Claude proposes evergreen angles from RSS feeds

2

Validate demand

SERP + autocomplete signals, scored 0–100

3

Draft + images

Claude writes the post and generates images

4

Queue distribution

Builds social pins, marks ready to publish

A queue of demand-validated, publish-ready drafts

Document automation

Intake form to a traceable requirements doc

The problem. Writing a requirements doc by hand eats a day, and it still drifts from what was actually agreed in the room.

What we built. A short intake form becomes a structured, traceable document. It runs a gap-check that refuses to invent scope, stays grounded in your past docs, and waits at a human approval gate before anything publishes.

  • n8n
  • Claude
  • RAG (pgvector)
  • Slack
  • Confluence
  • Google Sheets

Every requirement carries a stable ID and acceptance criteria. It's auditable, not vibes.

AI requirements analystautomation workflow

Intake form submitted

Trigger

1

Gap-check intake

Asks for missing fields, never invents scope

2

Retrieve past docs

RAG pulls your closest prior examples

3

Draft the doc

Claude writes requirements + acceptance criteria

4

Approve in Slack

Human sign-off before anything ships

5

Publish

Pushes the doc + a traceability matrix to your tools

A published doc + traceability matrix, approved by a human

Ops automation

Meeting notes to tracked action items

The problem. Decisions made in a meeting evaporate by Friday. Action items live in someone's notebook and never make it into the system.

What we built. Drop in a transcript and it pulls out the committed actions (not the parking-lot maybes) and the decisions. You approve, and it opens a tracked task per action with a recap posted back to the team.

  • n8n
  • Claude
  • Slack
  • Jira

Human-in-the-loop: nothing is created until you say yes.

Meeting notes to action itemsautomation workflow

Transcript received

Trigger

1

Extract actions

Claude separates committed actions from maybes

2

Approve in Slack

You confirm before anything is created

3

Create tasks

One tracked issue per approved action

4

Post recap

Decisions and owners back to the channel

Every commitment becomes a tracked, owned task

Reporting

Weekly status reports, drafted for you

The problem. Status reports are an hour of copy-paste every week, and they still come out reading like a robot wrote them.

What we built. Every Friday it pulls the week's activity, drafts a status report in your voice with a clear red/amber/green rating, waits for your approval, then files it in the places people actually look.

  • n8n
  • Claude
  • Slack
  • Confluence
  • Google Sheets

Drafted, not auto-sent. You always approve before it goes out.

Weekly status reportautomation workflow

Friday schedule

Trigger

1

Pull activity

Reads the week's tracked work

2

Draft in your voice

Claude writes a structured status + R/A/G

3

Approve in Slack

You sign off or tweak first

4

Publish everywhere

Confluence + Slack + a R/A/G log in Sheets

A consistent weekly report, written and filed for you

How every build works

The same rules, every system.

Whatever the workflow, these four hold, because they are what makes an automation safe to actually run your business on.

A human gate before anything ships

Every system we build holds at an approval step. The automation drafts and proposes; a person says yes before it sends, publishes, or creates.

Grounded, not guessed

Gap-checks, retrieval from your own past work, and structured outputs keep the AI on facts. If the inputs are thin, it asks. It does not invent.

Owned, not rented

Built in your accounts on n8n and tools you control. If we part ways, the whole system stays with you, with no black box and no lock-in.

Every run logged

Error branches alert us when something drifts, and each run leaves a trail, so you can see what ran, what it did, and what it cost.

Free · no commitment · audit in ~2 days

See where your business is leaking booked revenue.

Tell us where to send your free Growth Leak Audit. It's a one-page read on your missed calls, follow-up, website, and local search, with a 30-day fix plan.

Free PDF. We only use your email to send the audit and follow up about your business.