← All posts

How a GC Stays in Control of Automated Follow-Up

The real question isn't "does it work?" It's "what happens when it gets it wrong?"

A GC in the commercial fit-out space — let's call him Marcus — came to us with 40+ quote requests a month and zero follow-up system. His estimator was buried. Leads were going cold at day 3. He'd heard about automated follow-up and his first question wasn't about open rates or sequences.

It was: "What stops it from sending something embarrassing to a client I've been chasing for two years?"

That's the right question. And the answer is the part most vendors skip over.


The approval step is a business control, not a feature

Here's how the system works in practice.

Every morning, Marcus opens a simple review queue — one screen, not a dashboard full of charts. The system has already done the work overnight: it pulled in new quote requests, matched each lead to the right follow-up message based on job type and days since contact, and drafted the next action for each one.

But nothing has moved yet.

Each item in the queue shows:

  • The lead (name, company, job scope)
  • What the system wants to do (advance to follow-up 2, send a check-in, flag as cold)
  • The exact message it will send, word for word
  • A two-button choice: Approve or Reject

Marcus reads down the list. Most days it takes him eight minutes. He approves 90% without changes. He rejects two or three — usually because he knows something the system doesn't. A client who called yesterday. A job that's on hold. A relationship that needs a personal touch, not a template.

When he approves, it sends. When he rejects, it holds and flags for manual follow-up. Nothing touches a client inbox without his eyes on it first.

That's the whole model. The system does the thinking and the drafting. Marcus does the deciding.


Why this matters more in construction than almost anywhere else

In construction, your reputation is a short list of names. One tone-deaf follow-up to the wrong contact — a project manager who just had a bad week, a developer you're trying to get on a preferred vendor list — and you've done damage that a good quote won't fix.

Automated systems fail in two ways. They either do too little (you still have to manually trigger everything, so what's the point) or they do too much (they fire off emails you didn't review, and you find out when a client replies confused).

The approval queue sits between those two failure modes. The system does the heavy lifting — tracking, drafting, timing — but it cannot act unilaterally. Every outbound action is a recommendation until a human confirms it.

This isn't a safety net. It's the operating model.


What Marcus's numbers looked like at week 6

Before: 40 quote requests a month, follow-up happening on maybe 15 of them, average first follow-up at day 6.

After six weeks with the approval queue running:

  • 40 out of 40 quote requests getting at least two follow-ups
  • Average first follow-up: day 2
  • 4 jobs re-engaged from leads he'd written off as cold
  • Zero client complaints about inappropriate or mistimed messages
  • 8 minutes a day of Marcus's time to run the whole queue

The volume didn't change. The coverage did.


The three things to check before you trust any automated follow-up system

If you're evaluating a system for your own business — or you're already running one and wondering whether it's safe — here's what to look for:

1. Can you see the action before it executes? If the answer is no, or "you can set it to notify you after," that's not good enough. You need a pre-send review step, not a post-send log.

2. Is the reject option as easy as the approve option? Some systems make approval one click and rejection a three-step process. That's not neutral design — it's nudging you toward approving things you should be questioning. Both options should be equally fast.

3. Does the system learn from your rejections? If you reject the same type of message three times in a row and the system keeps drafting it, it's not actually helping you — it's just creating a manual review burden. The queue should get smarter over time based on what you approve and what you don't.


This is what we built for Marcus. It's what we build for every contractor we work with.

If you want to see the queue in action before committing to anything, the 30-minute audit call is the right place to start. We look at your current follow-up process, show you where the gaps are, and map out what a review-first system would look like for your volume.


Samir Grouicha Skylab Innovations — AI systems built for founders