Follow up on the quotes you already wrote

Most service businesses we work with have a quote problem they do not call a quote problem. A tech walks a job, writes up a number, and sends it. Then nothing happens. The customer does not say no. They just go quiet. A week later the office manager scrolls past the quote in the CRM, means to call, gets pulled into a scheduling fire, and forgets. A month later the job is gone, usually to whoever called back first.

When we ask owners how many open quotes they have right now, most guess low. When we pull the number from their system, it is often two or three times the guess. Roofers, HVAC shops, lawn care companies, and remodelers all have the same pattern: plenty of quotes going out, very little structured follow-up after the first send. The work to win those jobs was already done. The truck rolled, the estimate was written, the price was set. What is missing is a few short, well timed touches afterward.

That is why quote follow-up is one of the first systems we suggest when a client asks where AI can pay for itself. It is small, it touches revenue directly, and it is easy to measure. It is also easy to get wrong in ways that annoy customers, so the design matters more than the model.

Decide the rules before you write the messages

The first version of a follow-up system is a spreadsheet, not a prompt. Before anything sends, we sit down with the owner and write out the rules a good office manager would follow if they had unlimited time.

Start with timing. For most residential work, the first follow-up should go out within one business day of the quote, while the customer still remembers the tech's name. A second touch around day four or five. A last one around day ten to fourteen, framed as closing the loop rather than chasing. After that the quote goes to a human list, not another automated message. Three touches is usually plenty. Five starts to feel like a collection agency.

Then decide who the message comes from. Customers respond better to the tech or the office manager by name than to the company. If the tech who walked the roof was Marcus, the text should sound like it came from the office on behalf of Marcus, and replies should land somewhere a person will see them the same day. A follow-up that invites a reply and then ignores it is worse than no follow-up.

Next, write the stop rules, because they protect the business more than anything else in the system. Stop the sequence the moment any of these happen:

• The customer replies at all, even with "not yet."
• A job gets booked or a deposit comes in.
• Someone on staff logs a call or note on that quote.
• The quote is marked lost, or the customer asks not to be contacted.
• The quote is edited or reissued, which starts a fresh clock.

Most of the bad experiences we hear about come from missing stop rules. A customer signs on Tuesday and gets a "just checking in" text on Wednesday. That single message makes the whole company look disorganized. So the system checks the quote's current status right before every send, as well as when the sequence starts.

Finally, decide what the messages are allowed to say. We keep a short list of things the system can never do on its own: change a price, offer a discount, promise a start date, or answer a warranty question. If a customer asks about any of those, the reply goes to a person with the quote attached. The automation's job is to restart the conversation, not to negotiate.

With those rules written down, the messages are the easy part. A model can draft a short, specific note for each quote using the job details: "Hi Dana, this is the office at Ridge Roofing following up on the estimate Marcus wrote for the back slope repair. Happy to answer any questions or get you on the schedule. Just reply here." We review a sample of drafts with the owner before turning anything on, and we keep a plain template as a fallback for any quote missing the details a good draft needs.

Measure it like a sales channel

Once it is running, treat follow-up the same way you would treat any channel that is supposed to bring in revenue. That means a few numbers checked every week, not a dashboard nobody opens.

The core number is the close rate on quotes that got follow-up compared with the close rate before the system existed. Pull ninety days of history first so you have a baseline. If your shop closed thirty percent of quotes before, and quotes in the new sequence close at thirty eight percent, that difference times your average job value is what the system is worth. For most of our clients that number is large enough that the conversation about cost ends quickly.

Next, look at reply rate per touch. If the first message gets most of the replies and the third gets almost none, the third can probably go. If the second message outperforms the first, your timing is likely too early. Small changes here move results more than rewriting the wording.

Watch the complaint signal too. Count opt outs, angry replies, and any customer who mentions being contacted too much. A healthy sequence should see very few. If that number climbs, slow the cadence before you touch anything else.

Then check the handoff. How many replies needed a person, and how long did they wait? A reply that sits for two days loses most of what the follow-up earned. We usually route replies into the same place the office already works, whether that is the CRM inbox, a shared text line, or a team chat channel, and flag anything that mentions price or timing so it gets answered first.

Last, audit a handful of sent messages every week. Read them as if you were the customer. Did the name match the right tech? Was the job description accurate? Did anything go out after the job was booked? This takes ten minutes and catches problems long before they show up in reviews.

None of this needs a large project. A typical build for us connects to the client's existing field service or CRM tool, reads open quotes, applies the rules, drafts and sends messages, and writes every action back to the quote record so staff can see what happened. Most of the effort goes into the rules conversation and the stop checks, not the AI part. That is the right balance. The model makes the messages sound like a person wrote them. The rules make sure the business only says what a careful person would say, at a time a careful person would say it.

If you have a pile of open quotes and a team that is too busy to chase them, this is a good place to start. You already paid for the hard part. Follow-up is how you collect on it.