The SMB Operator

DEEP DIVE

What I'd Actually Automate Inside a $5M Business

Not the front door. The handoffs between people, ranked in the order I'd fix them, and the part where each one breaks.

It leaks in the handoffs. Ranked: 01 approved to job packet, 02 field to office, 03 done to invoiced, 04 what's stuck.

Everyone automates the front door now. Missed-call text-back, instant lead replies, quote follow-ups. Fine. That's not where a $5M shop bleeds.

Quick intro, since a lot of you are new here. I'm Drew. I run SMB Ops, where we build automations and back office systems for $3M to $50M service businesses, and every Wednesday I write this about what actually works after the AI demo ends. Not for you? Unsubscribe is at the bottom, no hard feelings.

In the shops I've worked with, it leaks in the handoffs. Every spot where work moves from one person to the next and the only system is somebody remembering.

Drew

The Happy Path

THE DEMO

Missed call โ†’ AI texts back โ†’ job booked โ†’ owner goes golfing
Front door handled. Everything after it is somebody's problem.

If you don't have missed-call text-back and quote follow-ups yet, set them up this week. It's cheap, often just a setting. Table stakes. But that's not where a $5M shop leaks.

Then Reality Hits

REALITY

The money leaks between people, not before them.
Quote approved, nothing happens. Notes stuck in a tech's phone. Invoice out Thursday for a job done Monday. Nobody can say what's stuck.

One job from start to cash, with the handoffs where work stalls. Front door (call, form, quote sent) is table stakes. 01: after approval, nobody builds the materials list, crew notes or pre-arrival text. 02: on site, photos, notes and change orders stay in the tech's phone. 03: after completion, the invoice goes out days later. 04: nobody can see what's stuck or whose name is on it.

Here's the order I'd fix them. For each one, the build is the easy part. The part where it breaks is the job.

The Order I'd Build It

1. Quote approved. Then nothing.

The leak: the customer says yes. Then nobody builds the materials list, crew instructions or pre-arrival checklist until the day before. The crew shows up short a part, and the customer didn't know to clear the garage.

Built: approval triggers a job packet drafted from the quote line items: materials, crew notes, a pre-arrival text. The ops manager reviews it instead of building it.

An approved quote with three line items turns into a job packet: materials list, crew instructions and a pre-arrival text get drafted. A flag pops up: tank size not on the quote, sent to the estimator, not guessed. The ops manager reviews and approves.

Where it breaks: the quote says "replace water heater" and nothing about tank size. No gate code. The lazy version guesses. The good version flags it to a person by name. A guessed materials list is worse than no list, because everyone trusts it. Same idea as the Approve button: the system does the 90%, a human spends 30 seconds on the part that matters.

Tools: your field service software already knows the moment a quote is approved. Jobber, for example, notifies admin users with an activity feed and push notification. That's your trigger. Zapier, Make or n8n builds the packet, AI can draft the crew notes, a person approves.

2. The field-to-office gap

The leak: the tech changed scope on site, snapped photos, got a verbal yes on an extra. It all lives in his phone. The office invoices off the original quote, wrong or late.

Built: a job can't close until it has photos, notes and any change from the quote. Changes get approved by the customer on site and land on the job record, not in a text thread.

Where it breaks: techs hate typing. Make closing a job a five-minute form and you'll get "done." Ask only for what the invoice needs, take voice notes, and send incomplete jobs to a needs-info list with the tech's name on it. Skip this and your CRM is lying to your AI. The record says one thing, the job site said another.

Tools: your field service app's notes and photos, plus an automation that holds incomplete jobs back from billing.

3. Job done. Invoice goes out days later.

The leak: the job wraps Monday. The invoice goes out Thursday, because the office batches them. Nobody tracks the gap, so nobody feels it.

Illustrative example: a $5M shop bills around $20k per working day. If invoices run three days behind, roughly $60k is sitting in customers' pockets instead of yours, all the time.

Built: job marked complete, invoice drafted the same day from the job record, office reviews and sends. Complete with no invoice after 24 hours? It shows up with a name next to it.

Where it breaks: don't auto-send everything. Warranty work, disputed jobs, open change orders and special terms get held for a human, with the reason attached. A wrong invoice sent fast costs more than a right one sent late.

Tools: Jobber, for example, puts invoice reminders on the schedule for completed jobs with no invoice yet. Start by not ignoring them.

4. The Monday "what's stuck" list

The leak: ask the owner what's stuck right now. The honest answer is "let me check with three people." Or it's the spreadsheet the office manager keeps that's quietly the real system.

Built: every Monday at 7, one list. Every quote, job and invoice waiting on a human, how long it's waited, and a name next to it. Sorted by dollars. Not a dashboard. A list somebody works.

Where it breaks: "Ops" is not a name. If an item has no owner, that's the finding. And when the list is wrong (a job showing open that finished weeks ago), fix whatever made it wrong, not the report.

Tools: your field service and accounting tools, a scheduled automation, and email or Slack. A spreadsheet is a fine start.

What I Wouldn't Automate Yet

  • Pricing judgment on weird jobs. AI can draft the standard quote. The job with three surprises needs someone who's been burned before.

  • The first reply to an angry customer. Draft it, sure. A human sends it. Fast matters less than not making it worse.

  • Writing off an invoice or firing a customer. Anything with relationship consequences stays human.

  • Any process three people do three different ways. Automating it just makes the mess faster. Agree on the process first.

The Play

โ

Automate the handoff. Flag the gap. Never let the system guess.

Three Takeaways

  1. The front door is table stakes. Set it up this week, then move on.

  2. Every handoff needs a trigger, a packet and an owner.

  3. Missing info goes to a person, by name. Guessing is how automations quietly cost you money.

Steal This

STEAL THIS

The One-Job Walkthrough

About 20 minutes. Pick one job that closed last month and walk it from approved quote to paid invoice with whoever touched it. At every handoff, ask:

  • Who passed it, and who caught it?

  • What did the next person need that wasn't there?

  • How long did it sit?

  • How did anyone know it was their turn?

Wherever the answer is "they just know," that's your first automation.

Close

None of this is as fun to demo as an AI receptionist. That's the point. The money in a $5M shop is in the boring middle, and the middle is all handoffs.

YOUR TURN

Where does work get stuck in your shop? Hit reply and tell me the one handoff that drives you nuts.

One line is plenty. I read every reply.

WANT A SECOND SET OF EYES?

Free Handoff Check. A 30-minute call. We walk one real job from approved quote to paid invoice and mark where it stalls. You keep that list either way. If nothing's worth building, I'll say so.

FREE PROMPT OF THE DAY

Handoff 03 is the invoice that goes out days after the job. The free Invoice Mailing SOP prompt lays out sending and chasing an invoice, and splits what a person does from what your accounting tool should do.

I send one free prompt like that every weekday morning over at SMB Automations. Get the weekday prompts โ†’

Drew Reynolds
The SMB Operator

P.S. Know an owner who's the only person who knows what's stuck? Forward this to them.

Stop typing AI prompts. Start talking.

You think 4x faster than you type. So why are you typing prompts? Wispr Flow turns your voice into ready-to-paste text inside any AI tool. Speak naturally, tangents and all, and Flow cleans it up. Available on Mac, Windows, iPhone, and Android.

Keep Reading