You spent forty-five minutes on an estimate. You emailed it. They said they needed to talk to their spouse, or that they were getting a couple more bids. Then nothing.

You meant to follow up. You were on a job site. By the time you remembered, they had signed with somebody else.

This is the most common gap I see when I read a company’s actual email, and it is not a discipline problem. It is that the follow-up lives in one person’s memory, and that person has a day job.

The part nobody automates, and should

Most people think the automation here is the sending. It is not. The sending is easy. The hard part is knowing which quotes are still open, and that is a data problem.

If your quotes live in QuickBooks and your customer records live somewhere else, nothing knows the current state of a quote. So any follow-up system has to answer three questions first:

  • Which quotes went out, and when?
  • Which of those have been accepted, declined, or gone quiet?
  • Who is the right person to chase, and on what?

Until those are answerable from one place, an automated follow-up is just a way to email people who already bought.

Where the quote data actually lives

This is the part that decides whether the rest is a week of work or a month.

If you use Method CRM with QuickBooks Desktop, estimates sync between them two-way. Method documents this as a real-time sync that fires whenever a record is added or changed on either side, rather than a nightly batch (Method Sync in Detail). That matters more than it sounds: quotes created in QuickBooks appear in Method without anyone re-entering them, so the follow-up runs from Method while your team keeps working the way they already do.

Check the Desktop part specifically when you are comparing options. Plenty of CRMs advertise a QuickBooks integration and mean QuickBooks Online, which is a different product with a different API. Desktop support is the narrower list.

I would also not take any vendor’s sync documentation, including that one, as settled until I had watched it run against a copy of your own data for a week. Sync is the component most likely to be almost right, and almost right is worse here than obviously broken, because nobody goes looking for the quotes that quietly did not arrive.

If your quotes only exist as PDFs in a sent-mail folder, that is the thing to fix first, and it is a smaller job than it sounds. A single shared sheet with one row per quote, a date, a value, and a status field is enough to drive everything below.

What the sequence actually says

The mistake is writing four messages that all say “just checking in.” A sequence works when each message does a different job.

Day 2. A short note that assumes nothing and asks one question: is there anything in the estimate that needs explaining? Most non-responses at this stage are not price objections, they are one line the customer did not understand and did not want to look stupid asking about.

Day 5. Offer a specific, small next step rather than a decision. Not “are you ready to proceed”, but “want me to walk you through the labour line?” A five-minute call is easier to say yes to than a signature.

Day 10. The last one. Say plainly that you will stop following up, and ask whether their timing changed. This gets more replies than the two before it, because it removes the pressure and gives them a graceful way to say not yet.

Then stop. A fourth and fifth message trains people to ignore you.

What the clock actually counts from. Trigger the sequence off the date on the estimate itself, not off a “quote aging” field. Aging is a calculation, not a stored value: it is today minus the estimate date, and it is only correct on the day something recalculates it. Every system already holds the estimate date, it never changes, and it means the sequence keeps working if you later move the whole thing to a different tool. The one date worth adding is the date of the last contact, because that is what tells the system to shut up when the customer has already replied.

Where AI actually helps, and where it does not

It helps with drafting. Given the quote details, an AI model writes a message that references the specific job (the material chosen, the rooms, the concern they raised on the phone) rather than a template with a name merged in. That is the difference between a follow-up that reads as attention and one that reads as a mailing list.

You review before it sends. This is not a limitation to apologise for, it is the design. The model does not know that this customer’s brother-in-law is also bidding, or that you cut the price already as a favour. Ten seconds of review per message is cheap, and it is what keeps the automation from embarrassing you.

It does not help with deciding who to chase. That is the data problem from earlier, and no model fixes a missing status field.

The four cases where this makes things worse

Most writing on this subject stops before the failure cases, which are the part that decides whether you keep using the thing. These four are reasoned from how the pieces fit together rather than from watching them go wrong in production, so treat them as the things to watch for rather than as war stories.

  1. The quote was already accepted verbally and nobody updated the status. Now you have sent a chase to somebody who thinks they hired you. This is the most common way these systems lose trust in the first fortnight, and the fix is that a status field is only useful if updating it is easier than not updating it.
  2. The customer is in a slow-moving approval chain. A ten-day sequence aimed at a facilities manager waiting on a capital request reads as harassment. Long-cycle work needs the sequence stretched or switched off per quote, and somebody has to be able to do that in one click.
  3. The quote was a favour or a placeholder. Every shop has quotes that were never real. If those get chased automatically, you look like you cannot tell the difference.
  4. Two people follow up. If the automation sends on Day 2 and your salesperson also calls on Day 2, the customer gets a worse impression than if nobody had contacted them. Whatever tracks the quotes has to be the same thing your team looks at.

Every one of these is solved the same way: the automation writes to the place your team already looks, and any person can override it for one quote without asking anyone.

Is it worth doing

Work it on your own numbers rather than mine.

Take the quotes you sent last month, and the share that went quiet without a yes or a no. Multiply that count by your average quote value, then by whatever you think a follow-up would honestly recover. Ten percent is a defensible guess and you should substitute your own.

Then compare it to the cost, which for the drafting is a few dollars a month in API usage plus my time to set it up. The arithmetic usually decides itself, and if it does not, that is a real answer and you should not do it.

The number I cannot give you is what share of quiet quotes come back with a nudge, because that depends on your market and your prices and I have not measured it for you.

I am Kenyon Sutherland. I build the automation described above for small manufacturers and service companies: email and quote intake, screening, and reporting, on the tools you already run. If you want to know what applies to your operation, send me a week of your real emails and I will tell you.