Business automation

Stop paying people to retype data between the software you already own

We connect the tools you already own (Jobber, QuickBooks, HubSpot, Gmail, Stripe) so information moves on its own. Most builds take 3–8 weeks and give the office back a day or more per week.

Built for

  • Contractors and home services with 5–50 trucks and one office holding it together
  • Shops where the same job gets typed into three systems
  • Firms whose invoicing waits on one person’s Friday
  • Teams paying for good software that doesn’t talk to their other good software
  • Anyone whose “process” is a spreadsheet plus a group text

Automation builds run $6k–$40k depending on how many systems connect and how many workflows you need. A single high-value flow, like job sync to QuickBooks or estimate follow-up, usually lands at $6k–$12k. A full intake-to-invoice rebuild across five systems runs toward the top. Ongoing support and changes are $2,500–$12,000/mo, or nothing if you’d rather run it yourself.

The problem

Nothing’s broken. It’s just all held together by one person and a spreadsheet.

The job comes in by phone. Someone writes it on a pad, types it into the scheduler, texts the tech, and later types it again into QuickBooks. Nobody planned that. It grew, one workaround at a time, and now it takes a full-time person to keep it from falling over. When she takes a week off, invoicing slips and nobody chases the quotes.

The cost is quiet, which is why it never gets fixed. Quotes that go out Friday night instead of Tuesday morning lose to the guy who answered first, and invoices sent late get paid late. Missed follow-ups don’t show up anywhere on your P&L. They just don’t become revenue.

What it buys you

The outcomes we hold ourselves to.

Quotes go out the same day
Job details flow from the intake form to the estimate template automatically. The estimator reviews and hits send instead of rebuilding the job from a text message. Shops that were quoting in 3–4 days typically get to same-day or next-morning.
Invoicing stops waiting on Friday
Completed jobs push into QuickBooks as they close, with the line items already in place. Invoices go out the day the work’s done, which pulls days out of your cash cycle without hiring anyone.
The office manager gets a day back
Killing the retype points on a typical intake-to-invoice flow gives back 6–12 hours a week. She doesn’t leave. She starts chasing receivables and unsold estimates instead of doing data entry.
Follow-ups happen whether anyone remembers or not
Unsold estimate at day 3, day 10, day 21. Text via Twilio, email via Gmail or Outlook, stops the moment they reply or book. It’s the cheapest revenue in the building and it currently depends on someone’s memory.
You can see the numbers on a Monday
Jobs booked, average ticket, quote-to-close, aging receivables: pulled from your real systems into a report that lands in your inbox before you get to the office. No one builds it by hand.
It’s yours, and you can read it
We build on n8n, which is open source and self-hostable. Every workflow is visual, so you can open it and see the boxes and arrows. There’s no proprietary black box and no per-task pricing that punishes you for growing.

Deliverables

What you actually receive.

Written, handed over, and yours to keep, whether or not we work together again.

Connected systems

  • Two-way sync between your field software (Jobber, ServiceTitan, Housecall Pro) and QuickBooks
  • CRM sync so HubSpot or Pipedrive reflects real job status, not stale notes
  • Payment events from Stripe posted back against the right invoice
  • Form and phone intake landing in one place with a source tag attached
  • Calendar and dispatch writes to Google Calendar or Outlook

Working automations

  • Lead routing by trade, zip code, and job value, with a fallback so nothing sits unclaimed
  • Quote follow-up sequences over SMS and email, auto-stopped on reply
  • Job-complete → invoice → payment reminder chain
  • Document processing: read a PDF PO or signed estimate and file the fields where they belong
  • Scheduling and reschedule notifications to the customer and the tech
  • Review requests timed to job completion, not to a monthly blast

Visibility

  • A daily or weekly ops email with the five numbers you actually check
  • A dashboard for job pipeline and receivable aging
  • Alerts when something stalls: quote untouched 48 hours, job closed but not invoiced
  • Error notifications that reach a human, with the failed record attached

Ownership and handover

  • n8n running on your infrastructure or a hosted instance in your name
  • Every workflow documented in plain English, one page each
  • A recorded walkthrough of each automation for whoever inherits it
  • Credentials in your accounts, not ours
  • Training for the two people who’ll touch it most

Process

How the engagement runs.

No surprises, no scope drift you didn’t agree to. Each phase ends with something you can read or use.

  1. 01

    Follow one job end to end

    We take a real job from the phone call to the paid invoice and write down every system it touches and every time someone retypes something. Usually a half day on site. The list is always longer than anyone expects.

    Week 1
  2. 02

    Build the first automation

    We start with one workflow, usually the worst retype point, and get it into production. You see a real result in the first few weeks, not a status update. And if we got it wrong, we found out cheap.

    Weeks 2–3
  3. 03

    Build out the rest in order

    Working in the sequence we agreed, one automation at a time, each one live before the next starts. You’re running on the finished pieces while we build the next ones.

    Weeks 3–7
  4. 04

    Break it on purpose

    We test what happens when QuickBooks is down, when a phone number is garbage, when a customer replies “STOP.” Every workflow gets retry logic and a human alert. This is the part that decides whether you trust it in month six.

    Week 7
  5. 05

    Hand it over

    Training with your team, plain-English docs, recorded walkthroughs, and 30 days of tuning while it settles. After that you can run it yourself, keep us on a retainer, or hand it to someone else. Your call, and your system either way.

    Week 8

Examples

The kind of work this turns into.

Intake to invoice for an 18-truck plumbing company
Calls hit Twilio, get logged and routed by zip code, and land in ServiceTitan. On job close, the line items post to QuickBooks and the invoice sends that afternoon. The office manager stopped retyping about 60 jobs a week and started chasing the aging report.
Unsold estimate follow-up for a roofing contractor
A three-touch sequence over 21 days (text, email, text) pulling the customer name and the roof detail from the estimate. Stops on reply. Runs on 100% of estimates instead of the ones someone remembered on a slow Thursday.
PO processing for a commercial subcontractor
Purchase orders arrive as PDFs in a shared Outlook box. An n8n flow with Claude reads each one, pulls the PO number, line items, and delivery date, checks it against the job in the CRM, and flags mismatches for a human. Clean ones file themselves.
Lead routing for a multi-trade home services group
Leads from four websites, Angi, and the phone all land in one queue, get tagged by trade and value, and route to the right team with a 15-minute escalation if nobody claims them. Nothing sits overnight because it came in on the wrong form.
Monday morning numbers for a 30-person HVAC shop
One email at 6am: jobs booked last week, average ticket, quote-to-close, receivables past 30 days, techs under target hours. Pulled live from ServiceTitan and QuickBooks. Replaced a spreadsheet that took the controller three hours every Monday.

Questions

Before you ask.

Will this replace my people?

No, and if you’re hoping it will, hire a different firm. It removes the typing, not the person. The office manager who spent nine hours a week retyping jobs spends it on collections and unsold estimates instead, which is work that makes money and that she was too buried to do. Where it helps with headcount is letting you go from 18 trucks to 25 without adding a second admin.

What if n8n goes down?

Two answers. First, it runs on your infrastructure or a dedicated instance we monitor, and every workflow retries failed steps and alerts a human with the record attached, so a stalled job shows up as a text rather than a customer complaint three days later. Second, and more important: n8n is open source. If we vanished tomorrow you’d still have the software, the workflows, and the docs, and any n8n developer could pick it up. That’s the reason we build on it.

We’ve got 20 years of junk in QuickBooks. Does that kill this?

Usually not. Automations work on new records going forward, so the mess sits behind you instead of blocking you. Where old data does matter, mostly matching customers and deduping, we handle it in the build and tell you upfront what the cleanup costs. Nobody has clean data, and we’ve never walked away from a job over it.

My software doesn’t have an API. Now what?

Most of the big field tools do: ServiceTitan, Jobber, Housecall Pro, QuickBooks. For the ones that don’t, there are usually workable routes like scheduled exports, email parsing, or a database read. We check this in week one and tell you honestly if a piece isn’t worth doing. That answer is free.

My guys won’t use anything new.

Good news: mostly they don’t have to. The best automations are invisible to the field, where the tech closes the job in the same app as always and the invoice happens on its own. Where we do change something, it’s one app fewer, not one more. If a build needs your crew to learn a new tool, that’s a strike against it and we’ll say so.

How fast do we see something real?

First automation in production in 2–3 weeks. We deliberately start with your worst retype point so the first thing you see is the pain going away. The whole build is typically 6–8 weeks, but you’re running on finished pieces the entire time.

Let’s find the work worth automating.

A 30-minute discovery call. We map where your time actually goes and tell you what’s worth automating, and what isn’t. No deck, no pressure. You leave with a written summary either way.