Home Quality Remodeling is a contractor in Concord, California. They handle structural work, HVAC, decks, and full-home remodels. They also run one of the deepest workflow setups we’ve seen on Niche.
It didn’t start that way. It began as basic lead capture and grew into a system that works leads from Yelp, Thumbtack, Google Local Services Ads, phone calls, SMS, and website chat, without anyone touching them until there’s a real reason to.
Here’s how it’s put together, and what it’s produced so far.
Leads everywhere, no time to work them
The problem is one most contractors know well. Leads show up on Yelp, on Thumbtack, through Google LSA, over the phone, in the website chat. Every channel has its own response-time expectations and its own way of formatting a lead.
Answering all of them by hand, fast enough to matter, isn’t realistic. Letting them sit is worse. A lead that waits until morning has usually already called the next name on the list.
Home Quality Remodeling didn’t fix this by hiring. They built workflows to handle the first stretch of every conversation, and to pull in a person once the lead is worth a person’s time.
The lead funnel: a 30-day follow-up sequence
The centerpiece is their lead funnel workflow. It’s on version 15, which tells you how much it’s been tuned.
It triggers on leads from three channels: Yelp (their largest by far), Thumbtack, and Google Local Services Ads.
The first thing it does is check scope. If the request is for something they don’t do, the AI declines politely and stops there. No time spent, no awkward follow-up a week later.
Next it checks for a phone number. If the lead didn’t come with one, the workflow opens an SMS conversation to get it, since a phone number is what makes everything after this possible.
Then the follow-up starts, and this is the part most businesses skip. Instead of one message and a shrug, the workflow runs 15 touches over 30 days, getting more direct as it goes:
- 5 minutes: a quick “let me know when you have a moment to chat”
- 30 minutes: a short circle-back to see if now works
- 4 hours: the first real push for an estimate, with a reason to move now
- Later that day: a bump with a phone number, then an availability check
- Day 1: a direct ask for contact details
- Day 2: a Calendly link so they can book themselves
- Days 3-5: progressively firmer follow-ups
- Day 7: the same ask from a different phone number, to break the pattern
- Days 14-15: a check on whether the project is on hold
- Day 20: a pricing angle
- Day 30: a last check with clear options, still looking, on hold, not ready, or have questions
Each message is written for the specific lead and reads like it came from someone in the office. The goal of every one is the same: get to a phone call or a booked appointment.
The conversation ends on its own terms. It stops when the customer hands over a phone number (routed to dispatch), when they ask for a quote (forwarded with the details), or when it hits the message cap and a person takes over. Every one of those exits fires a notification to dispatch over both WhatsApp and email, so a handoff never depends on someone watching one inbox.
What it’s produced:
- 2,095 Yelp leads processed, 505 turned into offers, 931 handed to the team
- 217 Google LSA leads, 178 handed to the team
- 19 Thumbtack leads, 15 handed to the team
That’s more than 1,100 qualified leads passed to the business without anyone working the early messages by hand.
Inbound calls: a voice agent that screens and routes
Their inbound call workflow (version 2) puts an AI voice agent on the phone. It answers questions, books appointments, and decides what to do with the call once it ends.
Every call gets sorted into one of three buckets:
- Callback requested: the caller left details or asked to be called back
- Qualified inquiry: a real question or genuine interest in the work
- Unqualified: wrong number, spam, robocall, or a hangup
That sort decides what happens next. Callbacks trigger an immediate notification. Qualified inquiries come through with a summary. Unqualified calls drop out quietly.
The workflow also watches for transfers to a human. If an appointment gets booked on a transferred call, dispatch gets the full details. As with the funnel, anything worth acting on goes out over both WhatsApp and email.
What it’s produced:
- 2,047 calls handled
- 2,042 of them legitimate, only 5 flagged as spam
- 1,026 calls ran longer than 30 seconds, the ones with a real conversation in them
- 526 workflow runs, every one completed
- 1,347 calls passed to the team for follow-up
Website chat: catching the visitors worth catching
The chat widget workflow handles people who land on the site and start a conversation.
When a chat ends, the AI labels it promising or junk. Promising means a real need, specific questions, contact info, or clear intent. Junk means spam, a competitor poking around, someone testing the widget, or a conversation that goes nowhere. The bar is set low on purpose: a borderline lead that gets a notification is a smaller mistake than a real one that doesn’t.
Promising chats come through with the customer’s name, email, and phone, a summary of the conversation, and context on what they’re after, which is enough for the team to follow up without re-reading anything.
Inbound SMS: triage before it reaches anyone
The newest piece handles inbound texts. The AI’s job here is narrow: decide whether a human needs to see the message.
It notifies on the things that matter, like scheduling, running late, quotes, complaints, billing, anything urgent, and callback requests. It stays quiet on the noise: one-time codes, spam, “ok” and “thanks”, STOP and START keywords, autoreplies. When a message is ambiguous, it notifies anyway. A couple of extra pings beat missing a real one.
Anything that clears the bar gets forwarded to the business phone with the customer’s number and their original message, so the team can reply straight back.
Why it works
A few things separate this from the usual auto-reply-and-hope setup.
It’s been iterated. Version 15 on the lead funnel isn’t a fluke. It’s the result of tuning timing, wording, and qualification over and over.
The follow-up is long. Most businesses quit after a touch or two. Home Quality Remodeling stays with a lead for a month, and the later messages get more direct instead of just repeating the first one.
It filters before it interrupts. The workflows don’t forward everything. They separate spam from real interest and only surface what’s worth a response, so the team’s attention goes to leads and calls that can turn into work.
And it hands off cleanly. The AI does the early back-and-forth, then steps aside the moment a lead shares a number, asks for a quote, or hits the message limit. Notifications go to both WhatsApp and email so nothing rides on a single inbox.
The results so far
Across the whole setup:
- 2,773 leads processed automatically
- 2,047 calls handled by the voice agent
- 1,124 qualified leads from the funnel handed to the team
- 505 Yelp leads turned into offers
- 859 workflow runs, with 5 failures
That’s thousands of leads answered and sorted, and more than a thousand real opportunities put in front of the sales team, without anyone chasing new leads or answering the same three questions all day.
If you want to copy this setup
Start small and keep changing it. You don’t need version 15 on day one. Start with capture and notifications, then add classification, follow-up, and routing as you learn what your leads actually do.
Handle every channel in one place. If leads come from several platforms, one workflow across all of them beats a great setup on one channel and nothing on the rest.
Don’t stop after one message. The 30-day sequence is doing a lot of the work here. Most leads don’t answer the first text, and staying in front of them with messages that change over time is what turns a cold lead into a booked call.
Sort before you notify. Not every message deserves an interruption. Let the AI clear the spam and the “thanks” so the team only sees what needs them.
Send important notifications two ways. Email gets buried and texts get missed, so two channels for the things that matter means fewer dropped handoffs.
What it adds up to
Home Quality Remodeling didn’t add headcount to keep up with their leads. They built a system that answers, sorts, and follows up on its own, and taps a person when there’s a decision or a deal to close. As they take on more volume, the system absorbs it.
The work they want to spend their time on is remodeling homes and closing jobs. Everything ahead of that now runs without them.
Niche Team
Product